BBM draft errata

From OpenCog
Jump to: navigation, search

Please list errata and comments about the BBM draft here.

Please insert the errata/comments in the wiki page section corresponding to the relevant chapter. If no section on this wiki page exists for that chapter yet, just create that section using wiki syntax.

When relevant, please indicate the specific page to which the correction or comment refers.

THANKS!!!

Contents

Global Edits

global search for "straight quotes" and replace them with “curly quotes/smart quotes”. Also search for:

  • " ’" i.e. space plus right "smart apostrophe" at beginning where it should be a left "smart apostrophe". This won't catch all of them, since beginning of line will be missed.
  • two consecutive "smart apostrophe" characters (both right and left) that typically are found in code examples and should be replaced with "straight quotes"

With respect to code:

  • all quotes and apostrophes should be "straight", with exceptions for semantic distiction, e.g. the right "smart apostrophe" for "prime".
  • adopt a standard font style for all code examples throughout the book. Consider using a smaller font, perhaps 2 pt less than the paragraph text style.

With respect to formulas:

  • some are numbered, most are not. It would be better to be consistent (no recollection of any reference to the formula numbers)
  • p722 has the correct "prime" character - " " (not sure this character will come out on wiki page); this character should be used throughout the book in formulas and variable expressions

global search for "above" and "below" for references to other parts of the book and replace with specific locations


check all bullet lists and check for colon, check that capitalization of the first word is consistent across all bullets in the group, ensure that periods are used for bullets that are sentences and omit if not a sentence


1st: manually global search "CogPrime s" and replace as appropriate - e.g.:

  • p133 "difference between CogPrime s cognitive schematics"
  • p757 "natural dynamics of the CogPrime system"
  • p787 "exists to mediate communications between other CogPrime s, and provide CogPrime s with information."

then: global search for "CogPrime " to remove extra space (sometimes paragraph break) before punctuation (comma, period, apostrophe, quotation make, hyphen, etc).


verify all figures, tables are referenced in the text


replace all "I" with "we" when referring to authors. (e.g. p7,54,206,209,210,211,430,453,640,791,832,1003,1011


replace all "my" with "our" when referring to authors (e.g. 53)


remove spaces from all space padded slashes, i.e. replace " / " with "/" to be consistent with the vast majority of usage, specifically: p43, 47, 169, 191, 217, 222, 225, 227, 342, 534, 589, 592, 598, 800, 817, 826, 827, 933, 964


ensure that all em dashes (or en dashes) are space padded one each side (for consistency), i.e. replace "xxx–yyy" with "xxx – yyy". This, of course, does not apply to hyphenated words, which use a simple hyphen or dash character. Conversely search for all " -" instances and replace with an en or em dash instead of the hyphen (as appropriate)


Resolve inconsistent use of 4th section level numbering, e.g. Chapters 9 and 11 use numbering e.g. "11.4.3.1 Systematic Experimentation", but others do not, e.g. Chapter 1 p8 "Build AGI from Narrow AI"


inconsistent presentation of urls - most in text (usually in parentheses) others in footnote. some with http, others just the domain name. Some in monspaced font, others in default font

excluding url in References section, pages with:

  • url in text: 9 16 269 270 271 340 342 342 347 350 350 379 385 388 521 689 703 738 1025 1037 1037
  • url footnote: 5 29 83 551 551 581 827 847 850 864

--- inconsistent use of "AtomSpace" - all variations (e.g. "Atom space", "Atomspace") should be changed to "AtomSpace"


global search for "WIKISOURCE:" - perhaps place in footnote after the first sentence of the following Secion - perhaps a url

e.g. p418: "WIKISOURCE:AtomNotation" = "http://wiki.opencog.org/w/OpenCogPrime:AtomNotation"?


search for each usage of "appendix" to make sure the proper reference is made


since it is not at all clear that all missing references (some to papers, some to other chapters, some to missing tables/figures), it is suggested to perform a global search for "?" to catch most of them.

Preface

p v,vi: notes about being a preliminary draft.

Don't forget to remove these notes :)

p v-vi: "CogPrime" is always followed by a new paragraph.

Remove the paragraph break after each occurance of "CogPrime"

Contents

p xxv: Missing space after chapter titles.

44.10Limiting the Degree of Disambiguation Attempted . . . . . . . . . . 821 through to 44.17Which Path(s) Forward? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 840


p xxx: "References" should be a top level heading and not appear to be part of Appendix H

Chapter 1: Introduction

p3: missing word "is"

An outsider to the AI field might think this sort of book is commonplace in the research literature, but insiders know that’s far from the truth.


p4: missing word "of"

understanding of particular specialized aspects of intelligence; and the creation of AI systems displaying intelligence regarding specific tasks in relatively narrow domains.


p6: missing word "to" (I know it's debatable...)

Due to their close interoperation they give rise to the overall systemic behaviors that characterize human-like general in- telligence.


p7: missing word "Chapter"

We’ll pursue this direction briefly in Chapter 7 and other chapters,


p8: strange character "É"

“Hmm. You know, that just might work.” É and a


p9: missing word "being"?

or a typical chatbot, due to using Wikipedia and other online resources, but still it’s far from being an AGI.

p9: dot should be out of quotes '... Ray Kurzweil's online chatbot "Ramona." Our ...'


p11: "a" -> "an"

this logic to be the crux of the system and the source of its potential power as a an AGI system.


p19: forth bullet point, unnecessary apostrophe in the last word "... design pattern'."

Chapter 2: What is Human-Level General Intelligence

p23. add "the"

Perhaps in the future


p23. reverse order 7 and 9:

(followed up in Chapters 97 and 79 with more rigor)


p24. second paragraph, third line, missing "one" in the sentence: "... similar to the above, ..."


p35. not a valid sentence

One of the central ideas underlying Cog’ is that with an integrative cognitive architecture that combines multiple aspects of intelligence, achieved by diverse structures and algorithms, within a common framework designed specifically to support robust synergetic interactions between these as- pects.


p38. Figure 2.2 wrong caption - s/b "The Robokind humanoid robot"?

The Nao humanoid robot

Referred to from p32 "...robot platforms such as the Nao (see Figure 2.1) or Robokind (2.2)..."


Chapter 3: A Patternist Philosophy of Mind

p42-45 Inconsistent syntax of bullet points. Some start with a comma, some with a dot and some with ":"


p47 Bullet point number 2. Missing 's', in "... different sorts of pattern(s) that ...".

p44. missing period (end of page)

and vice versa.


p51. - "analysis: Iteratively search" inconsistent layout(indent/spacing) with "synthesis: Iteratively build" above (p50)

"or in more detail" be consistent with about - i.e. "Or in more detail:" - also in both cases, the line spacing (double) would be better

no comma needed:

1.Begin with some components (the initial “current pool”), and a goal function


p54. captialize analysis in title

Self and Focused Attention as Approximate Attractors of the Dynamic of Iterated Forward/aAnalysis


p54. remove 'd' from based or rephrase. "various cognitive processes acting within a broad based of knowledge"


p55. The word 'agentive' feels out of place. (There's no other use of 'agentive' in the whole book)


p56. Rephrase: "Finally, we turn to the notion of an "attentional focus" is similar to Baars' notion of ..."


p57. wrong reference format (Boden2003; Goertzel1997)

mixed (em) dash styles:

However, in human minds and functional AI minds, the attentional focus will not be a complete chaos with constant turnover: sometimes the same set of ideas - or a shifting set of ideas within the same overall family of ideas will remain in focus for a while.

p57. Capitalize first word of sentence "Synthesis":

"...dynamic. synthesis occurs in the attentional focus when two or more items..."

And "Analysis" 7 lines below.



p.58 inconsistent used of "forward-analysis" and "forward/analysis"

3.4.4.1 also uses "forward-analysis",

3.4.4.2 uses "forward/analysis",

also 3.4.4.1 used "forward-synthesis activity"


p58. replace apostrophe with accented character

as not to end the section on too blase’blasé


p59. change "in in" to "it in"

will restrict ourselves to giving int in Appendix [?]"


p59. delete second "and"

"consciousness theory and and patternist"


p59. delete second "the": "the the brain's"


p59. fix quote symbols : ÒbackgroundÓ / Òunity of consciousness.Ó


p63. ( In the pattern intensity formula, is superscript '+' a standard mathematical syntax? If not, add explanation "a negative value results in a zero pattern intensity".)


Chapter 4: Brief Survey of Cognitive Architectures

p66: add "Figure" to disambiguate from formula (4.1)

symbolic, emergentist and hybrid – as broadly indicated in Figure 4.1.


p71. Text within figure 4.2 is not readable.


p71. There's a single line of text in the page, floating between two figures.


p72. Missing 'e' in "CogPrim's"


p72. missing reference

now building a system called Texai


p74. missing 't' in "... but in its logical aspec ..."


p75. overflow of margin (rearrange the wording if necessary to fix?)

SAL [JL08], based on the earlier and related IBCA (Integrated Biologically-


p75. missing space before have Ben Kuipers and his colleagues [?, MK08, MK09]have


p75. something wrong with formatting here: does "Our general perspective" start a new paragraph? If so, then indent, otherwise remove CR.


p77: "the" instead of "he"

In the context of perceptual data processing, the practical result of this is the need to couple each shallow learner with a pre-processing stage,


p79: Capitalize sentence and add period.

Eeach node also receives as input the belief state of the node above it in the hierarchy (which constitutes “contextual” information).


p 79: remove redundant "predictive"

Each node then contains the predictive task of predicting the likelihood of a certain centroid being most apropos in the near future, based on the past history of observations in the node.


p80: Fig 4.6 caption - wrong "prime" character following "S"

S’S‘ denote state variables pertaining


p82: remove comma

“cognitive synergy,


p82: replace "or" with "of"

“degree orof success or failure”


p83: add comma after learn, or remove "etc"

(and learn how to learn etc.)


p83: margin overlfow

through both unsupervised and teacher-driven processing of high-dimensional


p83: redundant and ambiguous

- "directly closely" redundant

- "related to the present one" is ambiguous -- perhaps "related to our own" as in the next bullet point?


p83: not a sentence - remove "which"

which is focused on creating software enabling an iCub


p84. Rephrase: "Like DeSTIN considered as a standalone architecture IM-CLEVER does entail ..."


p86: add "and"


Polyscheme has been used to model infant reasoning including object identity, events, causality, and spatial relations.


p86: inconsistent lead in with other bullets

James Albus’s 4D/RCS robotics architecture shares a great deal with

4D/RCS is James Albus’s robotics architecture that shares a great deal with...


p89: Fig. 4.8 : font should be darker.


p89: send -> sending


p89: Add 'not' , before 'vice versa'.

"send a lot of influential information to the other, but vice versa."


p92: confusing - did he or did he not add; is this a quote or not?

As Restak did not added, "...he is also not so effective at relating these emotional functions to straightforwardly implementable algorithms or data structures."


p95: add full stop to end of second dot point, at end of 4.5.5 PolyScheme

"...some language-specific processes"


p97: the time of what writing?

At time of writing (200912)


p97: expand the acronym BICA

Acronym was last used on page 4 "BICA (Biologically Inspired Cognitive Architectures)"


p98: replace double quotes with a single quote inside a qouted sentence.

A cognitive cycle can be thought of as a moment of cognition, a cognitive “moment.”'moment'.”


p99: replace "fo" with "of"

hat interact in the “global workspace” to create a model foof the agent’s current situation.


p:100 the "2" for footnote 2 preceded by a line break (remove) - The LIDA Cognitive Cycle in More Depth


p103: Talbe 4.1 - no glossary entry for "AT"; no glossary acronym associated with "STI"



Figures 4.6, 4.11, 4.12 are not referenced in the text

Chapter 5: A Generic Architecture of Human-Like Cognition

p111: fix reference: given in Chapter 6 belowFigure 6.1,


p112: archtiecture -> architecture


p112: "which is strikes me" -> "which strikes me"


p113: remove redundant words:

And one could view the integrative diagram as consistent with consistent with the deep learning paradigm


p113: "action and language. Etc." -> "action and language, etc."


p114: there are 7 figures, not 5

presented here in a series of five F seven figures.


p116: change 'parts' to 'portions'

In text: "Other parts of working memory"

In Fig 5.3 "other portions of working memory"


p117: lost somewhere between an emdash and endash - probably add space after the dash:

CogPrime is mostly like this it implements working memory as a subset


p119: redundant

Albus [AM01] has elaborated this kind of hierarchy quite elaborately.


p120: inconsistent with prior use on p 114

split among 7 different picturesseven different figures


p121: imporper, inconsistent form

(okay, 7seven)


p121:

The paragraph on this page asks: "Is anything critical left out of the diagram?"

Clarify where emotion appears in the diagram.

Consider reviewing the models to include comment on all elements listed in "2.3.1 Competencies Characterizing Human-like Intelligence"?

Chapter 6: A Brief Overview of CogPrime

p126: either use plural "diagrams" or be specific "Figure 5.1".

Comparing these diagrams to the integrative human cognitive architecture diagram given in Chapter 5


p126: add apostrophe to "OpenCogPrimes"

where OpenCogPrime's PLN


p128: change to correct "it's" apostrophe

Learning to build structures resembling structures that itÕs it's shown


p128: remove 'D' symbol

"...but that is not the inten of the project D the goal is..."


p128: change to correct "team's" apostrophe

"...focus of this project teamÕs"


p129. "proposal" WTF? - cut and paste error?

(a topic to be explored in more depth later in this proposal Chapter [?])


p129: "pattern" should be plural

should be considered as specialized ways of storing particular types of pat- terns, optimized for spacetime efficiency.


p133: missing apostrophe in "CogPrime's"

Finally, the biggest difference between CogPrime's cognitive schematics


p133: Include Economic Attention Networks (ECAN) in Glossary


p136: insert space before dash

proceed to Part 2 – after completing Part 1


p141: Presumably Table 6.1 is a mess, overflows the page, needs caption, needs clarification. Suggestions:

  • Have caption explain the what "Int. Diag. Sub-Diagram" means and perhaps maps each of the middle column entries to the corresponding Figure 5.X
  • use footnotes in columns that include "no (clear) correlate"
  • Perhaps break up into three tables for the boxes, hexes, octagons

Chapter 7: A Formal Model of Intelligent Agents

p150: redundant: "the agents sends information to the environment by sending symbols"


p157: missing "in": "we give a semi-formal sketch of some ideas on this (in) Chapter 9 below"


p162: weird looking commas in sentence:

"This conceptual approach may seem a bit artificial, but, we don't currently see a better alternative, if one wishes to ..."


p162: "left-semi-formal" s/b "left semi-formal" ?


p163: Definition 7 uses "=" instead of "≡" as in prior chapter 7 Definitions?

Chapter 8: Cognitive Synergy

Chapter 8 and sections 8.1 and 8.2 have the same name: "Cognitive Synergy".


p168: plural missing?: "may be viewed as different sorts of pattern(s) that a system..."


p169: FYI - Fig 8.1 is identical to Fig 6.7 - inconsistent with other places in book where references to figures in other chapters are made.


p168, 171 (Fig 8.2): (Editorial errata) there is inconsistency in the enumerated types of memory the bullet points that include intentional memory, it should be explicitly presented as a separate box/sub-box in the Figure on 8.2 (as well as in the caption - "In order to make these judgments it must effectively manage declarative, procedural, episodic, sensory, attentional and intentional memory, each of which is associated with specific algorithms and structures as depicted in the diagram.".

It would make more sense to break out bullet "Sensory and episodic memory" into separate bullets

  • Episodic memory
  • Sensory memory

The same holds for p172: "That is: each of these cognitive structures or processes deals with one or more types of memory – declarative, proce- dural, sensory, episodic, attentional or intentional".


p172: Table 8.1 - 8.6 missing; well, at least there is something resembling a Table on p175, which is labeled Fig 8.3


p172: Last paragraph is too long and confusing: "Attentional knowledge..."


p173: table reference problems

"The synergies between CogPrime ’s cognitive processes are well summa- rized in Table 6 below,"

"To illustrate this point further, consider one of the synergetic aspects de- scribed in Table 8.4 below"


p173: two missing apostrophes: "OpenCogPrimes PLN inference framework", "OpenCogPrimes declarative memory"


p174 extraneous period

"higher-confidence next step.."


p174 Duplicate section - "Section 8.5 The Cognitive Schematic" is a duplicate of "Section 6.6 Analysis and Synthesis Processes in CogPrime"


p174: Missing "be": "this implication may (be) interpreted to mean"


p175 Improve quality of fig 8.3 (continued on p176)


p180: wrong right quote character - “demes“


p182: duplicate section: "8.6.2 Cognitive Synergy in PLN" is almost identical to "6.4.1 Cognitive Synergy in PLN" with some corrections to one or the other, e.g.:

  • CH 6: "and CogPrime’s internal simulation engine handles" versus Ch 8 "and OpenCogPrimes internal simulation engine handles"
  • Ch 6: (p131) "This is just one example of the multiple ways in which the different cognitive processes in CogPrime interact synergetically; a more thorough treatment of these interactions is given in [Goe09a]." versus Ch 8: "This is just one example of the multiple ways in which the different cognitive processes in CogPrime interact synergetically; a more thorough treatment of these interactions is given in Chapter 48."

p182: Missing "t": "is likely to be a regular visitor to i(t)s preschool or not"


p183: footnote 1 is floating in an empty line


p184: "IF" should not be capitalized, perhaps emphasis? - "with the property that IF a proto-AGI system passes the"


p184: comma missing between "Turing Test" and "the video Turing Test,"


p185: "as in the OpenCog design" s/b "as in CogPrime"


p186: insert space before left paren: "better for the task than this subset of 6 components(assuming the latter are"


p187: replace at->as: "Just at 2/3 of a human brain may not be (of?) much use..."

Chapter 9: General Intelligence in the Everyday Human World

p191: replace question mark with a period

than an AGI that embodies and hence automatically utilizes these principles?.


p192: all bullet lead-in headings have extra space before comma. E.g.

  • Linguistic communication , in a

p193: "better" inconsistent with following bullet point, awkward

For instance, an experimental chemistry pro- cedure may be much better more easily communicable procedurally, whereas an allied piece of knowledge from theoretical chemistry may be much better more easily communi- cable declaratively;


p193-5: Section Numbering inconsistent with rest of book, not included in TOC (or just use a plain heading like in previous chapters, e.g. p8: "Build AGI from Narrow AI")

9.3.0.1 Naturalness of Knowledge Categories

9.3.0.2 Cognitive Completeness

9.3.13 Generalizing the Embodied Communication Prior


p194: remove extra space after bulleted bold and punctuation (: or ,)

  • At least one learning type : either analysis, or synthesis, or both
  • At least one interaction type , for each (knowledge type, learning type)

pair it handles: “isolated”


p194: "Then, intuitively, it seems to follow from the ECP with" should not be a new paragraph, use lower case and no indent, e.g. as in p204.


p195: remove line break before bolded hypothesis - or perhaps add a colon? (i.e. "Thus we arrive at the hypothesis that:"


p195: missing period

interaction terms in the simplicity measure.


p195: replace hyphen with en dash

serve as the seed for a novel sort of theory of general intelligence - a theory of


p195: missing closing parenthesis: "(e.g. at the Piagetan formal level"


p196: add periods to 7 & 8

7. When running downhill, one must lift one’s knees up high. 8. When looking at something that you just barely can’t discern accurately, squint.


p197: remove period from 2

2. Identifying beginnings, endings and crossings of processes


p199: add period to 1

There is an area inside the band and an area outside the band.


p199: remove apostrophe after phenomena

the “fundamental” aspects underlying these phenomena’ are.


p201: add apostrophe to AGIs - "but the issue of the AGI's body also merits consideration."


p204: remove "long and" since this is shorter than the median chapter length to this point.

Happily, this long and diverse chapter supports a simple, albeit tentative conclusion.

Chapter 10: A Mind-World Correspondence Principle

p205: in this chapter there are three capitalized titles without number. In two of them (2nd & 3rd) the following paragraph starts with ":", in the first one the following paragraph starts with "."


p205: strange characters at "and worlds — where a “world” is conceived"

and worlds where a world is conceived"


p205: missing 's' in "thee" at "are intelligent in these worlds."

full stop in front of new sentence without prior sentence at ". For a mind to work intelligently toward"


p205: strange characters at "where “nice” means that a"

where nice means that a


p206: strange characters at "It’s not clear, at this point, what a real “general theory of general intel- ligence” would look like"

It’s not clear, at this point, what a real general theory of general intelligence would look like


p206: reverse words "it gives" from "that approach currently it gives useful conclusions"

For all its beauty and theoretical power, that approach currently gives it useful conclusions --- p206: strange characters at "the notion of “general” intelligence"

the notion of general intelligence


p206: strange character at "given enough resources and time — but"

given enough resources and time and time but, even when this is true


p207: strange characters at "these “state-sets” from here on"

these state-sets from here on


p207: are A and B in the right place at "P(o(S, A, t(T ))|o(S, B, T ))"? if T = t_i and t(T) = t_i+1 then is this a link from A -> B here, not the other way? if A and B should be switched, do so on p211 too.


p207: strange characters at "time-distribution conceived as “after” T."

time-distribution conceived as after T.


p208: strange character at "the composition P ∗Q — one follows P"

the composition P ∗Q one follows P


p208: strange character at "thinking here — we want to be thinking"

thinking here we want to be thinking


p208: unnecessary parens, character '?' at "(Here nodes? obviously means the set of nodes of the path P .) "

Here nodes(P) obviously means the set of nodes of the path P.


p209:

colon in front of sentence: ": For an organism with a reasonably high"

missing full stop after "is a goal-weighted approximate functor"


p210: colon in front of sentence: ": For a mind to work intelligently toward certain" - make formatting here consistent with to p205 and p209


p210:

strange characters at "mind-states, where “nice” means that"

strange character at "above — so what?"

space before closing parenthesis at "(such as CogPrime ),"

strange character at "understanding, and so forth — until"

missing space between "O has" at "organism Ohas limited resources"


p211:

missing "to" at "we’ve tried to do here is to outline a clear" ?

strange character at "best way to do so — the details would"


p211: missing space: "If an organism Ohas limited..."


p212:

strange character at "infinite capability É rather with biases"

space before apostrophe at "CogPrime 's overall design"

missing semicolon at the first bullet point item (syntax-semantic correlation)

replace the full stop at the seconds bullet point item with a semicolon

Chapter 11: Stages of Cognitive Development

p216: remove apostrophe in "differences between humans and AI’s"


p216: add comma after "broadly"

Sketched so broadly, these are not precisely defined categories but rather heuristic, intuitive categories.


p218: remove "is", duplicate "as"

Indeed, Piaget viewed is develop- mental stages as as emerging from general “algebraic” principles rather than as being artifacts of the particulars of human psychology.


p219: add space between right quote and em dash

One of the crucial learning achievements here is “object permanence” –


p219: likewise insert space padding on each side of em dash in following bullet point (to be consistent with the preceding chapters of the book)

More abstract logical thought is applied to the physical world at this stage. Among the feats achieved here are: reversibility – the ability to undo steps already done; conservation – understanding that properties can persist in spite of appearances; theory of mind – an understanding of the distinction between what I know and what others know


p220: inser space on each side of em dash

relationships between phases of developmental ability – an approach which we

previous stage – so that the onset of a given stage represents the mastery of a


p223: insert comma after CogPrime

Obviously the main focus is CogPrime, but the essential ideas apply to any other uncertain inference centric AGI architecture as well.


p223: Fig 8.1 inconsistent (non-US) spelling for "recognise" s/b "recognize" for label of Infantile stage


p224: be specific

components, which work together in a feedback-control loop (Figure 11.3)


p229: fix proper title capitalization in "SpongeBob" and "SquarePants"

However, when her parent plays the disc, it’s “SpongeBob SquarePants.” If you then ask her what show her friend will expect when given the “Dora” DVD box, she will respond “SpongeBob”


p230: symbol "|-" is undefined (at least not until p417), perhaps a footnote is needed here

Chapter 12: The Engineering and Development of Ethics

p237: consider rewording "and thus the chapter on ethics occurs near the beginning rather than the end."

Is page 237, chapter 12 really 'near the beginning'?


p238: replace "SF" with "science fiction"

One SFscience fiction film after another has highlighted these dangers.


p239: remove first "to"

suggested to answers to each of them


p244: education->educated: "a CogPrime system if appropriately situated and education should be able to..."


p255-261: (odd numbered pages) Page heading too long "12.5 Clarifying the Ethics of Justice: Extending the Golden Rule in to a Multifactorial Ethical Model"

overwrites page numbers.


p255: add CEV, CAV or CBV to glossary


p256: remove apostrophe from "AGI's"

do is an extremely powerful heuristic, and while AGIs may eventually grow


p256: either completely rewrite the sentence, or at least remove "is that"

The truth of the matter, it seems, is (as often happens) not all that is that simple or elegant.


p259: "doubtless" s/b "doubtlessly" - can't see where the "and..." ties in to what precedes it. Perhaps adding for - "and for (effectively applying?) ..."

Applying these principles internally appears critical to effectively applying them externally, for reasons that are doubtlessly bound up with the interpenetration of internal and external reality within the thinking mind, and for the “distributed cognition” phenomenon wherein in- dividual mind is itself an approximative abstraction to the reality in which each individual’s mind is pragmatically extended across their social group and their environment


p259: wrong ordering: "logical coherence and the categorical imperative (imperatives 4 and 5)" are actually imperatives 5 and 4.


--- p262: change "paper" to "chapter"

Our goal in this paper chapter is not to enumerate a full set of complex networks of interacting ethical formulations as applicable to AGI systems


p262: be consistent with use of American English - change "programme" to "program"

out that this program must be undertaken in order to facilitate a grounded


p269: "see http:\metagovernment.org" s/b "see http://metagovernment.org"


p270: "And to i" incomplete second bullet.


p271: "Relationship with Coherent Extrapolated Volition" sub heading - would be 12.7.1.1 if numbered


p271: CEV is defined here (I believe for the first time) but is used several times previously


p271: quoted text not symmetrical in before and after line spacing


p275: remove space before comma - "considerations ,"

within a population with multiplicity of approach, and similar nuanced ethical considerations, combined with a sense of empathy,


Chapter 13: Local, Global and Glocal Knowledge Representation

general edit: this chapter is particularly egregious is placing paragraph breaks after "CogPrime". Mostly causing the following text to appear as a new paragraph which can be very confusing and hard to detect.


p286: remove comma after "knowledge"

’s explicit knowledge, representation, however, are carefully tuned to match


p287: "an Link" -> "a Link"


p291: bad line-break hyphenation for "Exist-sLink"

are taken into account in conducting the averaging. ForAllLink and Exist-sLink may


p301: margin overflow:

(the amount of energy flowing through the bundle from A1 to A2 at time t). So


p304: remove " e.g"

accurate remembering in many contexts [BF71, RM95] e.g. We believe the


p304: rework, remove bullet points and introducing sentence (the first two bullet points are only covered in this section):

The remainder of this section presents a fuller formalization of the glocal memory concept, which is then taken up further in Section 23.7's presentation of Glocal Economic Attention Networks (ECANs) – rough analogues of glocal Hopfield nets that play a central role in CogPrime.


p306: remove capitalization of first word in bullet points:

  • Thatthat is reasonably well-aligned (i.e. the correlation between

and basic distance is significantly greater than random)

  • Inin which most memory items come in pairs, consisting of one

p307: remove paragraph (ad pointed out on p304, glocal is not discussed in later chapters). Any reference here is erroneous or redundant.

We will see in later chapters how glocal memory can improve the performance of Hopfield networks, explain observed data about human neural memory, and serve as a useful principle for structuring attention and memory in an AGI system (CogPrime).


p309: "Figure ??" s/b "Figure 13.1"


p310: Figure 13.1 is blurry and hurts my eyes.

--- p311: Improve quality of Fig. 13.2

Chapter 14: Representing Implicit Knowledge via Hypergraphs

p317: remove paragraph break after "CogPrime" and "uses to represent knowledge explicitly"


p318: replace "Ñ" with endash

involving these probabilistic weights becomes quite involved Ñ particularly


p320: mismatched brackets - what is it "{}" or "()"?

ExOut add_numbers {3, 4)


p321, 323: overflow of section title in page heading "Implications of Patternist Philosophy for Derived Hypergraphs of Intelligent Systems"


p321: references that may be missed (Wang, 2006, 1995 TODO make ref)


p322: subject-verb disagreement, better to make "system" the subject.

The basic idea is that, even if a system – through its underlying dynamics – has no explicit connection to probability theory, nevertheless it still must behave roughly as if it does, if it is going to be intelligent.


p323: spelling error "abovee" - better: remove the word

The Cognitive Equation Principle, briefly discussed in Chapter 3 abovee,


p323: poor reference - this is described three paragraphs ago - at least the "above" word suffix is absent. Better: "This relates to the Cognitive Equation Principle defined above"

This is closely re- lated to the Cognitive Equation phenomenon described in Chapter 3

Chapter 15: Emergent Networks of Intelligence

p325: drop "we"

In line with these general observations, most of this book we will focus


p328: "Figure 15.4" s/b "Figure 15.3"

English-language concepts), and in Figure 15.3 for a specialized knowledge


p328: add "as illustrated in Figure 15.4"

In these cases, as illustrated in Figure 15.4, the hierarchies are normally spatiotemporal in nature – with lower layers containing elements responding to more localized aspects of the perceptual field, and smaller, more localized groups of actuators.


p333: Improve quality (make larger to fill the page more?) of Fig. 15.7

Chapter 16: AGI Preschool

p341: replace comma with colon for third bullet point

  • a manipulatives center: with a variety of different objects of different

p341: add colon for last bullet point

  • a dramatics center: where the robot can observe and enact various movements

p346: incomplete sentence in first paragraph

The design of the training and testing needs to account for potential


Chapter 17: A Preschool-Based Roadmap to Advanced AGI

p360: "work" missing period.

detailed roadmap based on the ideas outlined here is left for future work.


p360: remove space between "AI" and close quote

and Metrics for Human-Level AI” organized by John Laird and Pat Langley


p360: replace "Õ" with apostrophe

conference workshop of the AGI-09 conference, triggered by Itamar Arel’s


p360: Everything after "AGI Roadmap Workshop" belongs in a footnote. Better:

The train of thought presented in this chapter emerged in part from a series of conversations preceding and during the 2008 "AGI Roadmap Workshop"1. Although the focus of the workshop was considerably more general than the present chapter, here we focus on preschool-type scenarios, whereas at the workshop a number of scenarios were discussed, including the preschool scenarios but also, for example:

1 The "AGI Roadmap Workshop" was held at the University of Tennessee, Knoxville in October 2008. Some of the ideas also trace back to discussions held during two workshops on "Evaluation and Metrics for Human-Level AI" organized by John Laird and Pat Langley (one in Ann Arbor in late 2008, and one in Tempe in early 2009). Some of the conclusions of the Ann Arbor workshop were recorded in [LWML09]. Inspiration was also obtained from discussion at the "Future of AGI" post-conference workshop of the AGI-09 conference, triggered by Itamar Arel’s [ARK09a] presentation on the "AGI Roadmap" theme; and from an earlier article on AGI Roadmapping by [?].


p365: add qoutes around "yes"

is less likely to say “yes” if he’s asked too many times, because Ben


p367: example is wrong for "Empathy", you are providing an example of sympathy, not empathy (a critically important distinction). Better (borrowing/extending from the Theory of Mind example):

the robot is asked to identify and then display these same emotions.


p369: there is only one chapter between Chapter 17 and Part 2 (chapter 19); adjust accordingly, e.g.:

The next two chapters, intervening between this one and Part 2, explores

Chapter 18: Advanced Self-Modification: A Possible Path to Superhuman AGI

p373: margin overflow at: In this chapter we consider some mechanisms for "advanced self-modification"


p374: margin overflow at: We begin with a relatively near-term, down-to-earth example of self-modification:


p374: two occurrences of emdash formatting error - replace "—" with "–":

form this sort of tool-building may take is cognitive schema learning —

not use these — they may embody entirely different cognitive processes


p376: out of date (by a decade) information:

language Java. It is a large project and completion is scheduled for early 2003. But even at this stage, their partially complete Java supercompiler has had some interesting practical successes – including the use of the supercompiler to produce efficient Java code from CogPrime combinator trees.


p377: remove garbage meta-text surrounding "function"; insert "Figure 18.2" as described in "[NOTE...]"

119469711function119469711BNGBen Goertzel119469711651906446[NOTE: Give a graphical representation of the decision graph corresponding to the supercompiled binary search program for L=4, described above.].


p379: define an acronym on its first use - suggestion: replace "pdf" with "pdf (probability distribution function)"

schemata have complexities within these bounds, and the input pdf (probability distribution function) is between

Chapter 19: The OpenCog Framework

general chapter edit - rerun spell checker on entire chapter


p388: second bullet - suggestion, reference needed "Hugo de Garis’s Parcone"


p390: paragraph 3, sentence one. "important" should be "importantly"


p391: bullet 2 "Save and restore to (and from) disk *clarity


p391: spelling error "dataabases" s/b "dataabases"

criteria for databases. It is a very specialized database. Among the factors


p393: spelling error "reloade" s/b "reloaded"

forgotten Atom is persisted to disk so it can be reloade. Otherwise, the Atom


p393: spelling error "bove" s/b "above"

of its LTI. When the proxy’s LTI increases above a second threshold i2 , the


p394: bullet 3 "sped" should be "spend"


p395: paragraph 1 "will" may be better than "would" *wording

edge needs to be executable by the system. While it would be possible, and


p395: paragraph 4 "program state" should be "program states"?

dure repository should also facilitate the storage of program state during the


p396: inconsistent use "inserts and deletes" or "insertions and deletions".

quick insertions and deletes.


p397: spelling error "provice" s/b "province"

How is that knowledge used and useful? That is the provice of cognitive


p399: spelling error "is" s/b "in"

actions is atomic, incremental steps, where each step should be relatively


p399: either remove "The" or make CogPrime an adjective, e.g. approach, design, or architecture

cate with other MindAgents through message queues. The CogPrime has, in


p400 remove extraneous "t"

tion algorithms, which are ideally suited t to guarantee that, over the long


p400: poor hyphen break for "MindA-gents"

In essence, this is a ticketing system, designed to handle cases where MindA- gents or Schema spawn one-off tasks to be executed – things that need to


p401: spelling error "Wthin" s/b "Within" Wthin each Unit we have one or more MindAgents, a Task queue and, op-


p401: 1st bullet - incomprehensible - perhaps inject ("embodied by" or "within") between ("processes" and "MindAgents"), or use the expression from the previous section "MindAgents embodying various focused cognitive processes"

, and how these determine how much attention various focused cognitive processes MindAgents pay to them.


p402: spelling error "reaosnably" s/b "reasonably"

This is an area in which we can safely borrow from reaosnably mature


p409: remove extra text in second bullet:

Rule of Choice (see Chapter decides what to do, using the PLN Rule of Choice (see Chapter 34).


too many clauses remove ", " and "contains" to make it clearer

a subtler design, wherein each globally distributed MOSES subpopulation contains a pool of largely similar genotypes, and contains a cache of relevant parts of the AtomSpace,

Chapter 20: Knowledge Representation Using the Atomspace

p415: add "mean" between "we" and "that"

"interoperate" we mean that e.g. a fragment of a chunk of declarative knowledge


p416: (in 20.2 Denoting Atoms) "sorts of Atom" s/b "sorts of Atoms"

First we describe the textual notation we’ll use to denote various sorts of Atom throughout the following chapters.


p418: add reference to "PLN Book" i.e. [GIGH08], perhaps on 398 (if first usage) instead.


p418: delete "WIKISOURCE:AtomNotation" or perhaps place in footnote after the first sentence of the Secion 20.2.2 - perhaps a url "http://wiki.opencog.org/w/OpenCogPrime:AtomNotation"?


p420: replace code curly quotes with straight quotes e.g. replace ‘‘c’’ with "c"

ListLink
   CharacterNode "c"
   CharacterNode "a"
   CharacterNode "t"
   corresponds to the character string

("c", "a", "t")

p430: replace "I" - suggestion: "I have written" s/b "are expressed"

icateNode for f. These f-internal variables, which I have written using


p403: (same sentence) - "version" is disconnected from preceding sentence structure - no suggestions.

and #2, so the above grounding would be denoted

f -> (#1 > #2 - 1)

version, it is assumed that type restrictions are always crisp, not proba- bilistically truth-valued.


p431: "This page deals" s/b "This section deals" or "The following sub-sections deal"

This page deals with the representation of variable expressions in a


p432: provide specific chapter - "in the chapter" s/b "Chapter 33"

As will be discussed later (e.g. in the chapterChapter 33 on evolutionary learning,


p432: reference needed for "Moshe Looks’ PhD thesis" - add "[Loo06]"

Holman’s Elegant Normal Form (which is discussed in Moshe Looks’ PhD thesis [Loo06])


p432: margin overflow

But, putting computer programs into a useful, nicely-hierarchically-structured


p432: "So...so...so" redundant - better: remove "So:", replace "so we can" with "in order to"

So: In large part, we want to remove variables so we can use functional programming tools to normalize programs into a standard and pretty hierar- chical form, so we can in order to mine patterns among them effectively.


p433: inconsistent use of mono-spaced fonts for code expressions in paragraph text (especially better to use mono-spaced here for "->")

e.g. mono-spaced


p434: remove space in "specific )"

(general -> specific ) inherits from (specific -> general)

Denote the type of a mapping from type T to type S, as T -> S. Use the

e.g. not mono-spaced

The reason is that, if R1 = T1 -> S1 is to be a special case of R2 = T2 -> S2


p434: remove spaces immediately inside parentheses:

( INT, VOID ) inh ( Xv1 , Xv2 ) 

s/b

(INT, VOID) inh (Xv1 , Xv2) 

p435: use mono-spaced for code in paragraph

just assign Xv1 = INT, Xv2 = VOID.

is false because the restriction Xv1 inh Xv2 is imposed, but it’s not true that INT inh VOID.

The following list gives some examples of type inheritance, using the el- ementary types INT, FLOAT (FL), NUMBER (NO), CHAR and STRING (STR), with the elementary type inheritance relationships

  • INT inh NUMBER
  • FLOAT inh NUMBER
  • CHAR inh STRING
  • (NO -> FL) inh (INT -> FL)
  • (FL -> INT) inh (FL -> NO)
  • ((INT -> FL) -> (FL -> INT)) inh ((NO -> FL) -> (FL -> NO))

again, remove spaces immediately inside parentheses for last three bullet points above


p435: remove spaces immediately inside parentheses:

( INT, VOID ) inh ( Xv1, Xv2 ), [ 1 inh 2 ]

s/b

(INT, VOID) inh (Xv1, Xv2), [1 inh 2]

p437: replace hyphen with endash

slightly in this case, but only slightly - an increase that is well-justified by


p437: italicize "pow" (be consistent with previous usage)

This rendition uses alternate variants of - and pow schema, labeled


p437: remove spaces immediately inside parentheses:

(P p f) (v1, ..., vn ) = f (p (v1, ..., vn ))

s/b

(P p f) (v1, ..., vn) = f (p (v1, ..., vn))

Chapter 21: Representing Procedural Knowledge

p441: indecipherable references to sections. Suggestions:

  • replace "next two sections" with Sections 21.3 and 21.4
  • replace "fourth section" with "Section 21.6"
  • replace "Lastly, the fifth section" with "Section 21.7"
  • what about Sections 21.5, 21.8, 21.9?

Our goal in this chapter is to provide practical techniques for approximat-

resented is of paramount importance, as we shall see in the next two sections,

fourth section delves into effective techniques for representing programs. A

Lastly, the fifth section changes direction a bit and discusses the translation


p. 450 paragraph 1 "with all arguments are in normal form" remove are

p. 454 paragraph 2 "the number of neutral transformation" make plural

Chapter 22: Emotion, Motivation, Attention and Control

p. 461 paragraph 2 "Sectionis" missing space


p465: remove extra period after i.e.

of multiple Ubergoals; and choices here, i.e.. different specific ways of


p465: replace "with" with "will"

goal corresponding to external novelty with will have low STI but high

p465: put "also" after will

but there will also be many other Goals also


p. 466 last sentence "experience" should be past tense


p468: "Atop" s/b lowercase in heading


p. 469 paragraph 3 "thee" extra e

same paragraph "others" should be singular


p. 473 paragraph 5 "militate" should be mitigate?


p474: "RFSs" inconsistent with all other uses, so either make this "RFS's" or change all the others (starting in Ch24 to "RFSs"). It really should not have an apostrophe. Case in point: p546 uses HPANs. However if the intent of "RFSs" is to imply "requests for service" then it would be ok. This assumes changes to Section 24.2 heading to align with Glossary term. Otherwise, it should not have a "s" suffix at all since it is already defined as "Requests For Service".

Atom is found to consistently receive a lot of RFSs, and has a long time-scale


p. 475 paragraph 3 "Context node Chas" add space?

Chapter 23: Attention Allocation

p. 487 first sentence: remove "with"

Similarly to with STI, with LTI we will define theoretical LTI values, and


p.493 last sentence: remove extra "Prime"


p.501 paragraph 3 "Attention spread" should be "Attention was spread using"

paragraph 4 "If all the retrieved pattern" should be plural


p503: replace "a process to be discussed in a later chapter." with specific reference to chapter and section.


p503: replace "as shown in Table 23.9" with correct table reference, i.e. "Table 23.2"


p.505 paragraph 5 "Another issue that arises..." C should be plural


p506 "F" should be italicized - consistent font and style as prior use in p505 bullets

The difficulty is that the predicates F involved may be


p506 remove space after "relationship" and before comma

as a virtual predicate evaluation relationship , i.e.


p.507 first sentence extra "Prime"


p508 replace "outlined" with "outlines"

Appendix B outlined some very broad ideas regarding the potential utiliza


p509: replace "present paper" with "this Chapter" or "this Section 23.13"

However for the concrete technical work in the present paper, the traditional formulation of information geometry will suffice.


p510: place formula on separate line

Dabak [Dab99]...
is given by the exponential weighted curve
[formula here]
under the weak condition that the log-likelihood ratios

p510: replace "foinformation" with "of information"

Now we move on to discuss the practicalities of foinformation-geometric learn-


p512: reference to figures are suspect. Perhaps explicitly refer to Experiment 1 and Experiment 2 in the text.

The results are shown in Figure 23.3. We also ran several experiments comparing the sum of squares of the errors to the input training noise as measured by the value of σ.; see Figures 23.4 and ??.

Chapter 24: Economic Goal and Action Selection

p516: replace "differential" with "different"

is achieved via giving them differential amounts of currency. STICurrency is


p516: Change 24.2 Heading from “Requests for Service” to “Request for Services” to be consistent with Glossary definition



p516: It's rather strange to define it after it has already been used in a previous chapter (see Ch22 p474 note). In any event the close apostrophe belongs should be changes to double quotes and the close quote should appear before the comma. The same applies to the abbreviation. If the heading change is made the corresponding change should be made here as well.

promises of funding (or “request for services”, to be called “RFS’s” from here


p516: replace "a" with "an" (and insert where ommitted)

say that an Atom “issues” an RFS or “transfers” currency while what we really

mean is that some MindAgent working on that Atom issues an RFS or transfers

The logic of these RFS’s is as follows. If agent A issues an RFS of value x


p516 replace apostrophes with qoutes and place period outside.

of RFS’s possessed by an Atom may be referred to as its “promise”.


p518: replace "Whomever" with "Whoever" (bad grammar)

Whoeverasks for a redemption of the promise first.

Alternatively: The money will be paid to whomever asks for a redemption of the promise first and then the promise will be rescinded from the other one.


p518: "<" should be "<=" and add space to it

tagged with the constraint (B+C <= x). This means that in order to redeem


p518-9 First paragraph of 24.2 has different font characteristics (kerning, line spacing, etc)


p520 Min Jiang - no reference


p521: text overflow

mate Nash equilibria, e.g. [?] [ http://www.springerlink.com/content/p08jqj7w7rklb3r5/


p522: use explicit chapter - i.e. change "in the following chapter" to "in Chapter 25"; remove commas

We will have more to say about schemata and their modularization in Chapter 25.

Chapter 25: Integrative Procedure Evaluation

p523: place comma outside of qoutes - replace “mechanical,” with “mechanical”,

While this may seem somewhat of a “mechanical”, implementation-level


p524: described where? - be specific.

The SimpleComboTreeEvaluator simply does Combo tree evaluation as de- scribed earlier.


p524: "Chapter 19" should be "Chapter 20", perhaps refer to specific section

not defined by a ComboTree as defined in Chapter 19


p524: missing period at end of paragraph

PredicateNode (which is defined as the mean truth value of the predicate across all arguments known to CogPrime )


p536: text overflow

MindAgents, particularly those concerned with PLN inference. WIKISOURCE:ProcedureExecutionDetails

Chapter 26: Perceptual and Motor Hierarchies

p532 replace "an" with "a" 2x

by hierarchical structure, an hierarchically structured perceptual system is

advantageous. To control a body marked by hierarchical structure, an hier-


p532 paragraph 3 "There seems to be no good reason to complicate an AGIs" add apostrophe; same sentence "fact" should be "facts"; same sentence "it's" should be "its"

good reason to complicate an AGI's early development phase by forcing it to learn the basic facts of the world’s and its body’s hierarchality.


p542: replace "this document" with "this section" (or something else to that effect

fully resolved: this documentsection just gives a conceptual overview.


p545,47 top page heading overflow, runover page number


p548 better to start 3a and 3b on new indented lines


p548 use parentheses instead of square brackets (to be consistent)

[including sensory inputs and actions; and also the states of the most important items in memory during the experience]


Chapter 27: Integrating CogPrime with a Compositional Spatiotemporal Deep Learning Network

all even pages: overflow of Chapter title in page heading --- p552: unmatched double quote (i.e. left only before "action"

action and goals/reinforcement [GLdG+ 10]. The "action CSDLNs discussed


p556: recommend moving Figure 27.1 to previous page


p556: break up 1) 2) 3) 4) onto separate line:

1. the standard CSDLN is formed via processing input

2. frequent pattern mining on the standard CSDLN is used to create subnetworks and corresponding parent nodes in the semantic CSDLN

3. the newly created nodes in the semantic CSDLN get their internal clusters updated via standard DeSTIN dynamics

4. the clusters in the semantic nodes are used as seeds for frequent pattern mining on the standard CSDLN, returning us to Step 2 above.


p557: awkward, suggest reorganizing wording:


Finally, we note that the nodes and networks in the semantic CSDLN may be either:

  • linked into the nodes and links in a semantic network such as CogPrime’s AtomSpace
  • actually implemented in terms of an abstract semantic network language like CogPrime’s AtomSpace (the strategy to be suggested in Chapter 29).

Chapter 28: Making DeSTIN Representationally Transparent

p563: remove "while"

For instance, while olfaction does not lend itself well to hierarchical mod- eling, but audition and haptics (touch) do:


p567: add apostrophe in "OpenCogPrimes" 2x

ships have been stored between concepts in OpenCogPrime's memory and grammatical patterns between state-subtrees, then whenever concept C be- comes important in OpenCogPrime's memory, this can cause a top-down in-


p571: avoid "A" at the beginning of the sentence to avoid confusion with the other uses of "A" - Perhaps:

This is difficult to calculate in general, though – a rough proxy, which it seems will often be acceptable, is to measure the distance between A and B in terms of both


p572: add "an" (second bullet)

Suppose one has an embodied agent using DeSTIN for visual perception,


p572: add period to end of third bullet

amples than classic DeSTIN.


p573: replace "ÒkeypointsÓ" with double quotes

ematically isolates certain points in a visual scene as “keypoints” which are


p574: insert apostrophe in "itÕs"

evolved), but it’s also somewhat complex, requiring the use of subtle tem-


p575: more of the same - replace "Ò" and "Ó"

3. Wherever the vision system makes a decision based on “the most prob-able choice” from a number of possibilities, change the system to make a decision based on “the choice maximizing the product (expectation * confidence)”.


p575: space pad "*"

the highest probability * confidence.


p575: more of the same - "systemÕs"

accuracy of the system’s judgments.


p575: replace "Ð" with emdash

percepts that it intakes – i.e. to use what it thinks and knows, to modify


p576: replace "Ð" with emdash - 2x

cally embody this – a higher confidence being sensibly assigned to percepts

body the semantic meaning of the visual scene – which is exactly what we’re


p577 Fig 28.1 caption - remove hyphen/space in "co- ordinate"

region sampled relative to its scale-space coordinate frame. The features achieve partial

Chapter 29: Bridging the Symbolic/Subsymbolic Gap

general notes: This chapter really smack of what it really is - a separate paper injected into the book. This is glaring in the Introduction, when describing "two separate AI systems" - both of which have been thoroughly discussed by this point in the book. So perhaps, edit the p581 to something like this:

The approach presented here begins with two separate AI systems, OpenCog (introduced in Chapter 6.3) and DeSTIN (introduced in Chapter 4.3.1) - both currently implemented in open-source software. Here are the relevant features of each as they pertain to our current effort of bridging the symbolic/subsymbolic gap:

OpenCog, an integrative architecture for AGI [Goe10d] [GPW+ 11], which is centered on a "weighted, labeled hypergraph" knowledge rep- ...

DeSTIN [ARK09a],[ARC09], is a deep learning system consisting of a hi- ... being perceived. The core concepts of DeSTIN are similar to those of Jeff Hawkins’ Numenta system [HB06] [GH09], Dileep George’s work (http://vicariousinc.com) and work by Mohamad Tarifi [TSH11], Bundzel and Hashimoto [BH10], and others. However, The specifics of ...


p579: add hypens to "pre" and "post"

and can be thought of as pre- or post-processors to symbolic representation


p581: add apostrophe to "OpenCogPrimes" 2x - also consider just using "OpenCog's" since it gets confusing between referring to the framework and the implementation of the framework. It appears the chapter always uses "OpenCog" when not in the possessive case and "OpenCogPrime" exclusively in the possessive case.

2. Utilizing one of OpenCogPrime's cognitive processes (the "Fishgram" fre-

DeSTIN states, and then recording these patterns in OpenCogPrime's


p582: more of the same

3. Utilizing OpenCogPrime's other cognitive processes to abstract concepts


p583: misplaced/incorrect character "É" - parhaps an emdash? (see p587)

3 above, wrapped up in the phrase "É and other methods act on these per-


p583: more of the same

perception and dealing with text directly. Integrating OpenCogPrime's cur-


p584: remove reference to "companion paper" and add both "Uniform DeSTIN" and "classic DeSTIN" to the Glossary

From here on, unless specified otherwise, when we mention "DeSTIN" we will refer to "Uniform DeSTIN" as presented in Chapter 28.3the companion paper and [Goeon], an extension of "classic DeSTIN" as defined in [ARK09a].


p585: more of the same

OpenCogPrime's pattern mining component, Fishgram, exists precisely for


p587: replace "É" with " –"

case and the sunglasses case – unless the DeSTIN pattern library contained a


p588: replace "Part I" with "Chapter 28.2.1"

In Chapter 28.2.1 we have briefly indicated how DeSTIN could be extended beyond


p588: more of the same

and building HebbianLinks in OpenCogPrime's Atomspace between these Per-


p590: more of the same

everyday tasks. OpenCogPrime's Hebbian learning process can be used to find


p593: and one last time

out everyday tasks, but will also provide raw material for OpenCogPrime's

Chapter 30: Procedure Learning as Program Learning

p597: which chapters? - be consistent/specific - replace "(to be discussed in later chapters)." with ", discussed in Chapters 34 and 38." (assuming those are the correct chapters).

example PLN inference and concept predicatization (to be discussed in later chapters)discussed in Chapters XX and XX.


p597-99: drop the bold-italic style on the symbols and words - this is inconsistent with the rest of the book.

  • the numbered list bold lead-ins are ok
  • Use plain italic instead of bold-italic
  • use "quoted word" instead of bold (or perhaps just italic again, or both), e.g.:

In evolutionary learning terms, the program space is the space of “geno-types”, and the behavior space is the space of “phenotypes”.


p598: use numbered footnote

the complex∗ program space and chaotic execution function will often quickly


p599: use numbered footnote - 2x

ture† – they are not “needle in haystack” problems or uncorrelated fitness

superset of the latter’s desideratum, according to the fitness function.‡


p599: add these to the glossary:

  • hierarchical Bayesian optimization algorithm (hBOA)
  • Bayesian optimization algorithm (BOA)

p600: used numbered footnote

this is certainly contingent on the existence of some compact§ and reasonably


p600: Capitalize start of sentence "in"

on which modeling operates (introducing additional inductive bias). In Cog-


p600: replace "as discussed above" with "as discussed in Chapter 7.3"

problem-solving is indeed possible, e.g. AIXI tl as discussed in Chapter 7.3.

Chapter 31: Learning Procedures via Imitation, Reinforcement and Correction

p603: replace ?? with chapter number (32??)

text and affordances available. While Chapters ?? and 33 focus on procedure


p603: add apostrophe to "OpenCogPrimes"

experiments done with OpenCogPrime's predecessor system, the Novamente


p605: close quotes on item#'s: 2, 11, 16


p607: remove emdash at end of #5

5. tell the dog “I’m done teaching you to stay”


p608: remove "need"

from a human-controlled avatar as the fitness function, would have needed need an excessive number of reinforcement trials to learn relatively simple


p608: italicize or quote "I"

the procedure and see how it works. This is where the "I" part of IRC learning


p609: use numbered footnote

Chapter 40 ∗ . CogSim can be visualized using a simple testing UI, but in the


p609: "SL" is undefined in the paragraph - elaborate ('Second Life' is relegated to the footnote)

certain procedure would be likely to do if enacted in the SL virtual world. Of


p609: try using a different font for the "prime" part of the expression - otherwise they will appear as an undesirable "curly" apostrophe.


1. The episodic memory is mined for pairs (P’, E’) that are similar to (P,E)

2. The fitness of these pairs (P’, E’) is gathered from the experience base

Also, it would be more consistent to italicize P and E which is also used on:

  • p610: “Find a procedure P” that is as similar as possible to procedure P, but that
  • p611: series of N actions that it carried out prior to getting fed. It then must figure out how to set N

add Behavior Description (BD) to glossary

Chapter 32: Procedure Learning via Adaptively Biased Hillclimbing

p614: poorly placed comma, poor phrasing. Alternative:

The other, MOSES, is used for problems that are sufficiently difficult that the right way to solve them is to progressively build a model of the program space as one tries out various solutions, and then use this model to guide ongoing search for better and better solutions.


p614: poor phrasing: "seems not adequate", try "does not seem adequate" or "seems inadequate"

nates at a solution that seems not adequate, then the search can "backtrack"


p616: italicize the "t" in "time t"

Pt returns 1 when the perception is true at time t or 0 otherwise, where t


p617: user numbered footnote

probability of the desired behavior knowing p, then Bayes rule∗ tells us that


p623: do not italicize "Section 32.5" (inconsistent), replace "looses" with "loses"

for the reasons explained in Section 32.5 , that is the search looses itself in large regions of sub-optima.


p624: "takes a very cut" - incomprehensible

That is because in this circumstance the search is mislead by longer candidates that fit and takes a very cut before it can reach the optimal more compact solution.


general: the use of times in the tables and text is rather bizarre, like using Roman Numeral notation. Better to translate everything into seconds - e.g. 135s instead of 2min15s; 6.02s instead of 6s20ms

Also, is would look better to use the courier font (or something else to set it off) for double_dance, triple_kick, fetch_ball, etc

Chapter 33: Probabilistic Evolutionary Procedure Learning

p625: Co-authored with Moshe Looks∗ - use numbered footnote, or rather do something more in line with Chapter 32 (Primary author: Nil Geisweiller)


p. 625 paragraph 1 "The requirement is not that complex human behaviors need to be learnable via program induction alone, but rather than" should be that

p. 626 paragraph 2 first sentence is 7 lines long - clarity

"We also suggest that the majority of procedure learning problems arising in everyday human situations *missing plural* can be solved via programs *missing plural*

bullet 1 "Evolution is the right approach to the learning of difficult problems "

p. 627 paragraph 2 "he has articulated ways in which, during neurodevelpment difference (should be different)"

p. 628 paragraph 1 last sentence "it" should be "them"

p. 630 paragraph 3 figure?? missing


p630: use numbered footnote

or combination of knobs is undesirable.† Correspondingly, our set of knobs


p631: missing figure

Considering the space behaviorally (right of Figure ??),


p632: Section 33.3.1.3 is structurally unappealing, suggestion: make bullet points with bold lead-in heading; also kill the "om this step," e.g.

...is based on three conceptual steps:

  • Reduction to normal form. Redundancy is heuristically eliminated by ...
  • Neighborhood enumeration. A set of possible atomic perturbations is ...
  • Neighborhood reduction. Redundant perturbations are...

At the end of this process, ...


p. 633 paragraph 4 "centered around a single program of family of closely related programs"

same paragraph "This is consonant constant" (or perhaps, consistent)


p633: unbold "n"

grams containing less than n symbols), or spanning an arbitrary collection


p634: use numbered footnotes - 2x a deme;‡ a set of demes (together spanning an arbitrary area within pro- gram space in a patchwork fashion) will be referred to as a metapopulation.§


p. 634 paragraph 2 "via a representation-building process such as the one shown above" no diagram above


p635: missing figure

The preceding algorithmic sketch of MOSES leads to the top-level archi- tecture depicted in Figure ??.


p636: use numbered footnote

conditional (if-food-ahead ),¶ which takes two arguments and evaluates one


p. 636 paragraph 2 "It's It can successfully navigate the first two turns of the placeSanta" (missing space)


p637: use numbered footnote

by a movement knob, followed by a rotation knob, etc.||


p. 637 footnote "That there be is/exists"


p638: no sure reference "[23]"

based on the size of the space based on a model derived from [23] that is


p638: use numbered footnote

new representation is constructed centered around this program.∗∗ Additional


p640: no sure references "[27, 28]"

However, the results past attempts at introducing program simplification into genetic programming systems [27, 28]


p641: use numbered footnote - 2x

lation size†† ). In 50 independent runs of MOSES without model-building, a

sorts of linkages between knobs are being learned.‡‡ For the 6-knob model


p. 641 footnote 1 "The fact that reduction to normal form" (missing word)

p. 642 paragraph 2 "...,MOSES explicitly addresses on the underlying (semantic) structure..."


p645: change in font style characteristics for the PLN code that that used on previous page.


p646: Section 33.5.1 should use italics for "P" and "x"


p. 647 bullet 2 "position are" should be "is"


p647 replace "an" with "a"

of CTs that doesn’t explicitly have an hierarchical structure, via introducing


p. 648 paragraph 2 and 6 CT's should be CTs


p652: use double quotes instead of (wrong-way) apostrophes

We have left open the all-important question of how to find the "specifica- tion approximating predicate" Q.


p652 replace all "Q's" with "Qs"

Chapter 34: Probabilistic Logic Networks

p655: remove apostrophe on "Ikle’" (or rather be consistent across all uses (Matthew versus Matt and Ikle versus Ikle') p285, 479, 1047, 1073, 1074

Co-authored with Matthew Ikle


p656: paragraph break needed between "systems" and "In" - i.e. In practice... should be a new (non-bullet paragraph); (of course, the global rule to standardize use of end periods on bullets applies here as well)

  • ...natural language process-ing software systems

In practice, PLN consists of


p656: sync introduction of two acronyms (FOPLN and HOPLN) to glossary entries for "Higher-Order Inference (HOI)" and "First-Order Inference (FOI)" (acronyms FOI and HOI are never referenced in the text - although "first-order inference" is referenced on p650 and "higher-order inference" on p1055. Suggestion: modify these glossary entries to add the acronym:

  • PLN, First-Order (FOPLN): See First-Order Inference
  • PLN, Higher-Order (HOPLN): See Higher-Order Inference

vided into first-order and higher-order sub-theories (FOPLN and HOPLN).


p659: abduction is incorrect - replace "→ C" with "→ B" (inconsistent with the glossary: "If A implies C, and B implies C, then maybe A is B" and Fig 34.1)

abduction ((A → C) ∧ (B → C) ⇒ (A → B)),


p660: Missing appendix?

The various PLN rules have been gathered in appendix [TODO APPENDIX]


p666 inconsistent MindAgent name - "ConfidenceDecay" versus "ConfidenceDecayUpdater"

The ConfidenceDecay MindAgent must take care of applying the confidence- decaying predicates to the Atoms in the AtomTable, periodically.

The ConfidenceDecayUpdater MindAgent must take care of:

Chapter 35: Spatiotemporal Inference

p682: replace "which extend" with "what extent" (use which for predefined options, what as open ended); replace "same" with "safe"

dow and we want assess to which extend what extent the laptop is far from the window, therefore same safe from the rain.


p682: horrible grammar

We want to assess how far much the Laptop is far from the window


p683: same as above - replace "which extend" with "what extent"

and want to assess to which extend what extent climbing on the pillow is going to bring


p683: replace "bad" with "bag", fix bad grammar

2. The toy is near the bad bag inside the cupboard, how near is much the toy is near to the edge of the cupboard?


p684 replace "from" with "to"; question mark, double quotes.

4. If I am on the pillow near the edge of the cupboard how near am I from to the toy?

The target theorem is “How near I am from to the toy if I climb on the pillow?

3. Step 2 with axiom 4, target theorem: How near I am I from to the toy if I climb on the pillow?”


Chapter 36: Adaptive, Integrative Inference Control

p688: remove first "on the other hand"

The findExamples operation, on the other hand, sometimes reduces to a simple search through the AtomSpace. On the other hand, it can also be done in a subtler way.


p692: remove second "alternate"

An alternate solution for inference control... By structuring the rule sets into such a pipeline, one has an alternate, and very attractive solution to the inference control problem.


p692: remove first "most"

Most pProcessing for most situ- ations would follow well-established, previously-learned chains of thought.


p692: remove "then"

Provided that chains of thought are not heavily branched (i.e. for any given rule in set A, only a small number of rules in set B follow, etc.), then the ’logical deduction’ or data processing of input can be performed quite rapidly.


p692: lots of grammatical errors:

Finally it is interesting to note that the format of these chains of though may entirely be coded as Nodes and Link thought may be entirely coded as Nodes and Llnks


p693: replace "but also" with "as well as"

mainly successful ones but also as well as a sampling of unsuccessful ones for purpose


p694: remove these sentences (Ch 23 is the chapter on attention allocation)

The formation of HebbianLinks will be discussed in the following chapter, on attention allocation. Here it will suffice to get across the basic idea. The discussion of HebbianLinks here will hopefully serve to help you understand the motivation for the HebbianLink formation algorithm to be discussed later.


p699: lowercase "into" in heading

Incorporating Other Cognitive Processes into Inference

Chapter 37: Pattern Mining

General note: code often overflows margins or even page: p705-7


p703: capitalized relim - or be consistent with use on p694

and an alternative is Relim [Bor05] which is conceptually similar but seems


p706: change plural "variables" to "variable"

there is also a variable for the agent itself. Many interesting patterns involve


p706: consider different formatting for the Fishgram pseudocode, similar to 33.3.2.1 Algorithmic Sketch


p709: replace bogus "Ð" character, perhaps endash.

properties Ð predicates about a single object. So the relations are effectively

Chapter 38: Speculative Concept Formation

p712: replace "PLN book" with "book Probabilistic Logic Networks"

Among these are the node logical operators described in the book Probabilistic Logic Networks,


p712: replace "an" with "a" hypothesis

a hypothesis that, in fact, the everyday human world tends to assign util-


p712: reverse "Chapter in" to "in Chapter"

As noted in The Hidden Pattern and hinted in Chapter 3 of Part 1,


p716: first bullet, replace colon with "that" (be consistent)

topology principle that relations


p717: first bullet replace square brackets "[]" with parentheses "()"

(i.e. the concepts should have something to do with each other, but not be extremely similar; blends of extremely similar things are boring)


p719 (last para before 38.4) replace bogus "É" character and unclosed left paren "(" and , reword, perhaps with endashes as well:

However, if one views each feature of the blend as a probability distribution over concept space – for instance, indicating how closely associated each concept is with that feature (e.g. via HebbianLinks) – then we see that the mutual information


p721: change "isame" to "insane" (above formula)

insane (an attribute), one would calculate


p723: "add_"" s/b "append_"

S = append_concepts(S, form_concepts(S,b))

S = append_attributes(S, form_attributes(S,b))

Chapter 39: Dimensional Embedding

p727: add "adds" or rework "complexities" to verb form.

as well as plusses – one suspects it adds complexities and constrains the brain,


p729: add comma after links

and links, but the discussion applies much more generally than that.

Chapter 41: Integrative Procedure Learning

p745: replace "PLN book" with "book Probabilistic Logic Networks"

hard problem in CogPrime , discussed in the book Probabilistic Logic Networks


p746 be specific, replace "above" with "in Chapter 19.1"

to analogous aspects of the Webmind AI Engine (one of CogPrime ’s prede- cessor AI systems, briefly discussed in Chapter 19.1).


p749: add "PA-series" to glossary

Chapter 40: Mental Simulation and Episodic memory

p737: capitalize memory chapter title

Mental Simulation and Episodic mMemory


p737: Replace "This" with "These"

This These are components that are likely significantly


Chapter 41: Integrative Procedure Learning

p745: replace "PLN book" with "book Probabilistic Logic Networks"

hard problem in CogPrime , discussed in the book Probabilistic Logic Networks


p746 be specific, replace "above" with "in Chapter 19.1"

to analogous aspects of the Webmind AI Engine (one of CogPrime ’s prede- cessor AI systems, briefly discussed in Chapter 19.1).


p749: add "PA-series" to glossary


p758: poor hyphenation on "PredicateN-odeMiner"

Next, given high-truth-value Atom structure templates, the PredicateN- odeMiner


p762: replace "&mdash;" with a real endash or emdash

the category of “CogPrime Procedure learning” — it’s not actually a

Chapter 42: Map Formation

p765: remove "Abstract" below chapter heading


p769: sentence ends abruptly - missing formula/expression/reference? - perhaps Table 42.1

version, takes the form


p770: Tables 42.1 and 42.2 - what's with the middle column? -- judging from the source wiki page, at least replace the "?" with ellipses


p770: the "see MapMining" - fix it - it is a vague reference, only apparent due to the "WIKISOURCE:MapMining" extraneous text prefix for Section 42.4

(see MapMining) - the more advanced the algorithm, the fuller the transfer


p773: no table

Table ?? presents a typical example of data prepared for frequent item-


p774: "relational databases" - add acronym "(RDBs)" and remove apostrophe is subsequent reference.

are intended for relational databases (RDBs); but the information about co-activity in a CogPrime instance is generally going to be more efficiently stored as graphs rather than RDBs.


p774: remove "an" add endash

From an ActivityTable stores like this, the way to find maps is not frequent itemset mining but rather frequent subgraph mining – a variant of FIM that is conceptually similar but algorithmically more subtle, and on which there has arisen a significant literature in recent years.


p776: bad hyphenation

pands its internals into a collection of appropriately interlinked ProcedureN- odes.


p780: lowercase "and" in heading

42.8 Recognizing and Creating Self-Referential Structures


p781: kill the ellipses

erwise further refined, which will then cause them to exist more definitively and hence to be explicitly recognized as yet more prominent patterns, ... etc.


p781: awkward, consider rewording, e.g. "If so, then this willMap may be encapsulated in a single new Node"

If so, then this map willMap, may be encapsulated in a single new Node


p782: poorly worded, perhaps: "The next question, though, is how do we calculate the truth values of structures like those above.

The next question is how to we calculate the truth values of structures like the above, though.


p782: remove "a"

be an infinite order probability distribution, which is a funny and complex


p782: change "papers" to "paper" (there is only one cited)

of infinite-order distributions defined in the above-referenced papers actually


p782: lame - delete "and will be presented later"

but some additional hyperset op- erations may also be introduced; these are currently being researched and will be presented later.


p782: incoherent "[?]"

is wired to carry out matrix manipulations, and [?] reduced infinite-order

Chapter 43: Communication Between Artificial Minds

p786: add period.

to set up a CogPrime representing “collective knowledge” of multiple virtual pets.


p786: add space after comma "form,or"

CogPrime (in binary form,or in an XML representation, etc.),


p788: bullet points mismatch with "are" lead in, remove "is" and "is associated via"

ed are:

  • is a ConceptNode, linked via a ReferenceLink to the WordNode for “crazy”
  • is associated via HebbianLinks with ConceptNodes linked via ReferenceLinks to the WordNodes for “insane” and “impractical”

p788: replace with correct left single (instead of right single quote) quote on ’crazy’ ’insane’ ’impractical’

by the word ’crazy’ and is associated with ’insane’ and ’impractical’. ”


p788 footnote should use a "1" not an asterisk (be consistent with other chapters, which renumber each chapter)


p788: add space after right paren "later)as"

we may denote crudely (a better notation will be introduced later)as


p790: remove colon after "is" (just above bullets)

The general definition of a psynese expression for CogPrime is: a Set of Atoms that contains only:


p790: remove colon in last bullet following "than"

Predicates or Schemata involving no relationships or nodes other than: the ones in the above three categories, or in this category


p791: remove comma after "Psynese" (or add "one" before "looks")

And Lojban may be used to create a linearized version of Psynese, that looks

Chapter 44: Natural Language Comprehension

general edit: change all RelEx and AtomSpace code font size to match that on p813 - this may eliminate many of the code margin overflows without further edits as noted below.

general edit 2: in rule code, inconsistent use of "/" line continuation and "#" prefix


p797: insert space - "atthe" s/b "at the"

next will be focused on the engineering approach, but at the end of each chap-


p797: either drop comma after "approaches" (better) or add comma after 43:

using a combination of experience-based and engineering-based approaches along the lines described in Chapter 43 may be the most practical option


p 798: rework last sentence of paragraph beginning "Another key dichotomy related"

Then, in Chapter 47 we discuss the use of CogPrime cognition to regulate the dialogue process. In Appendix E, we present a complementary approach to linguistic interaction with AGI systems based on the invented language Lojban.


p 798: replace hyphen with endash

these tasks in a much more interconnected way - using separate procedures language - specifically because human language syntax is complex and am-


p800: margin overflow, missing text, extra spaces:

T extT okenizerLink P arser → Syntactico−Semantic Relationship Extractor (RelEx) → Semantic R

T ext → T okenizer → Link P arser → Syntactico−Semantic Relationship Extractor (Link2Atom) → Seman


p802: figure (link grammar parse structure) is wrong. it is also unlabled, unnumbered - "The person with whom she works is silly" s/b "The cat chased a snake"


p803: why is "ran_N" included in the second table? (remove)


p805: remove space in "O +"

ple, the link structure for “chased” is “S- & O+.” In categorical grammar,


p807: replace superscript "1" with "^1"

The ^1 indicates the data source, where 1 is a number indicating that the


p807: remove space after "CogPrime 's" (line break, so may be hard for search and replace)

which defines an inheritance relationship: something that is part of CogPrime


p808: Begin new sentence ", the" s/b "The", add colons:

The relevant FrameNet Mapping Rules are:

Finally, the output FrameNet Mapping is:


p808: use numbered footnote for consistency: "S *" s/b "the character S1", also add double quotes.

words or may be variables in which case their names must start with the character “$”1.


p809: margin overflow of RelEx code in 44.5.3

# IF on(put,$var1) & _obj(put,$var0) \
  THEN ^1_Placing:Goal(put,$var1) \
       ^1_Locative_relation:Figure($var0) \
       ^1_Locative_relation:Ground($var1) [1]

/code


p810: margin overflow of RelEx code

# IF for($var0,$var1) ^ {present($var0) \
     OR past($var0) OR future($var0)} \
  THEN ^2_Benefit:Benefitor(for,$var1) \
       ^2_Benefit:Act(for,$var0)

p811: margin overflow of RelEx code - use smaller font?? perhaps abbreviate "AsymmetricEvaluativeComparison" to "AEComparison"


p816: missing number (1?)

(recall, [2] is the instance of eating involved in Example 1;


p816: remove elipses, replace first with endash, other two with periods. Stick it in a footnote as well.

(In principle, some instances of eating might not be instances of ingestion – or more generally, we can’t assume that all instances of a given concept will always associate with the same FrameNodes. This could be assumed only if we assumed all word-associated concepts were disambiguated to a single known FrameNet frame, but this can’t be assumed, especially if later on we want to use cognitive processes to do sense disambiguation.)


p816: there is no reference to "sentence ([4])"

This says that the specific instance of Ben observed in the example sentence “Ben slowly ate the fat chickens.” served the role of Ingestion:Ingestor in regard to


p816: missing apostrophe "OpenCogPrimes"

Right now, OpenCogPrime's RelEx2Frame rulebase is far from adequately


p817: "will" s/b "well"

world interpretation, but matches well with things the system has seen before.


p817, 818: remove "to" from bullet points and append to "capability"

an OpenCog NLP system with the capability to:

  • support simple dialogue in a video game like world, and a robot system

p818: remove extra space "( the" - in second bullet

another set of rules (the RelEx2Frame rules) mapping rel-sets into


p818: remove last pair of bullets. These should be a continuation of the same paragraph. Perhaps rework to put all the "In the old approach" first, then start the "In the new approach"

In the new approach nothing needs to be hand-coded...In practice, some hand-coded rules are being created to map RelEx...


p821: add periods to bullets:

To illustrate the difference, consider the sentences

  • Love moves quickly.
  • Trains move quickly.

p823:

code following "On the other hand, an Atom2Link rule would look like the reverse:" is identical to the code above (i.e. not "like the reverse") - perhaps the last two lines need to be switched.

Implication
   Atom-set representing portion of link parse
   Atom-set representing portion of semantic interpretation

p823: add apostrophe to "OpenCogPrimes"

OpenCogPrime's frequent subgraph mining algorithm in this context. MOSES


p824: code overflows the margin


p827: used numbered footnote for consistency - "of Word Grammar are as follows  :"


p827: spelling error (carried over from the source url referenced in the footnote - i.e. "trense" s/b "tense"

ment rules - e.g. number but not tense or transitivity.


p828: either remove "in" from "similar to in a neural network" or add "that"

Activation-spreading, similar to that in a neural network or standard semantic


p828: define term before using acronym "dag"

in a directed acyclic graph hierarchy (dag) of isa links. The basic idea is as follows. Suppose one


p831-2: the "P" for parent in the code looks like the alignment is a bit off.


p833: no figure number or caption for the Link Grammar diagram

Improve quality of diagram.


p833: incomplete parenthetical remark "(i.e. no situations where A is a Before landmark of B, and"


p834: no figure number or caption for the Link Grammar diagrams

Chapter 45: Natural Language Generation

p844: add figure label to replace "??" and no need to be so specific for year range.

The sentence generation approach currently taken in OpenCog (2009-2012), which we call SegSim, is relatively simple and is depicted in Figure 45.1 and described as follows:


p845: improve quality of fig 45.1


p846: remove space in "involvi ng"

is used for relative clauses involving “whose”, to connect to the previous noun


p847: use numbered footnote instead of "*"


p849: Figure 45.4 makes no reference to "mathematics" in the Example 1 text: "I study mathematics at Xiamen\_University."


p850: use numbered footnote instead of "†"


p850: replace “clearupcleanup” with “cleanup”

“cleanup” step. For example: the substructures we extracted for the


p852: correct spelling of "parse-able" to "parsable"

responds to a complete sentence, legally parsable via the link parser. The

Chapter 46: Embodied Language Processing

p855: remove space between and comma - "iconic ,"

hand, one might argue that the thumbs-up signal is very weakly iconic, in


p856: file left single quotes in Ôhideous’ and Ôvery ugly’

this via explicit verbal instruction, e.g. one may tell it "The word ‘hideous’ means ‘very ugly’."


p859: replace "paper" with "chapter"

final part of the paperchapter, which is a bit more speculative and adventuresome.


p859: replace "Novamente’s" with "OpenCogPrime's"

such as Novamente’sOpenCogPrime's PLN inference engine.


p861: change "point 1" to "first bullet point"

With the exception of point 1first bullet point (entity extraction) above


p861: remove "now", perhaps make past tense

corrective learning mechanisms such as we are now utilizing utilized for our


p862: remove space in "OpenCog -based"

with the OpenCog-based PetBrain software, involving the use of virtually


p863: fix punctuation by adding period inside quote and add comma afterwards.

For example, if the user says “I like the red ball. Grab it.”,


p863: change double single quotes to a double quote in examples at the bottom of the page

‘‘The ball is red.’’ s/b “The ball is red.”

“The stick is brown.’’ s/b “The stick is brown.”

“Grab it.’’. s/b “Grab it.”


p864: capitalize the first word "the"

The anaphor resolver will build a list containing two options for the pronoun


p864: change double single quotes to a double quote in examples near top of the page - similar to the prior page


p864: replace "*" with numbered footnote


p865: change double single quotes to a double quote in examples on the page - similar to the prior page


p868: remove space before semi-colon at end of page

“I know the answer, but I don’t know how to say it”;


p869: remove hyphen after endash

spread across different online virtual worlds – all communicating in simple

and habits and likes and dislikes – but there’s also a common knowledge-base

food incoherently, it’s less likely to get food – and since the virtual parrots

hat please” incorrectly, then it won’t get positive feedback from the avatar – and it will be programmed to want positive feedback.


p870: remove hyphen after endash

many other purposes as well. Humanoid avatars – both human-baby avatars

Chapter 47: Natural Language Dialogue

p871: remove "the" preceding NLP

of ChatPrime sheds much light on the conceptual relationship between the NLP and other aspects of CogPrime .


p872: margin overflow

Dialogue control utilizes hand-coded procedures and predicates (SpeechActSchema


p873: fix apostrophe in "IÕm"

  • Expressives: The speaker expresses some psychological state. I’m

p876-881 Section 47.3.1 formatting is weird. It appears that an unintentional paragraph break occurs between a subheading, lead-in, whatever and a colon. Regardless, it is ugly and distracting. e.g.

Trigger context

: The trigger is that the conversation partner asks a wh- question

Goal

: Similar to the case of a TruthValueAnswer, discussed above

Emphasis of the lead-in text would also help, (so would making each one a complete sentence or at least being consistent with the use of periods when it is a sentence, but this crosses the line into style) e.g.

Trigger context: The trigger is that the conversation partner asks a wh- question

Goal: Similar to the case of a TruthValueAnswer, discussed above


p882: top sentence is incomplete

moderately near each other); and, predicates involving


p882: add "e" to "th"

learning is much easier than the procedure learning required to learn new

Chapter 48: Summary of Argument for the Cog’ Approach

p890: a bit light on references in this chapter, perhaps a reference of wiki url footnote for Big Dog?

become more robust at moving and grabbing – e.g. with Big Dog’s move-


p892-3: again no reference?

AGI researchers such as Pei Wang, but we are swayed by the theoretical ar-

arguments of Pei Wang, Peter Walley and others that using single-number


p898 unclear "and it may also provided increased efficiency" perhaps:

and it may also provided actually provide increased efficiency p899: change from question to statement (remove trailing "?") - "What about the three mind-geometric principles outlined in Appendix B: ... ?"

The three mind-geometric principles outlined in Appendix B are:


p900 change hyplens to endashes

learning algorithm – one that was still susceptible to an exponential increase

component algorithms are clever on their own – they are less susceptible to

– are all critical as well, but none of these will yield the critical emergence of --- p904 "it" s/b "is"

if the AGI system is placed in an environment which is itself hierarchically


p904: ellipse should be endash

this is indeed the case with CogPrime – biases toward hierarchical, heter-

Chapter 49: Build Me Something I Haven’t Seen: A Cog’ Thought Experiment

general edit - almost entire chapter uses a right curly quote (”) instead of a left curly quote (“), like this: ”quoted text”


p911: add period to heading "Action Selection."


p913: add period to heading "HebbianLink formation."


p915: add period to heading "Perceptual Pattern Mining." - more importantly, either add missing text description or drop entirely


p916: replace "an" with "a"

stance of information iteratively passing up and down an hierarchical


p919: add period to heading "Evolutionary Pattern Mining."

Also either drop the orphan "But", or complete the sentence/paragraph.


p920: remove space between heading and period for:

Dimensional Embedding.

Embedding Based Inference Control.

Fitness Estimation via Simulation.

Concept Formation via Simulation.


p921: remove space between heading and period for:

Episodic Memory.

Concept-Driven Procedure Learning.

Predicate Schematization.

Map Formation.

Map Encapsulation.

Experience Based Disambiguation.


p921: add period to end of sentence.

Concept-Driven Procedure Learning. Learning the concept of ”horse”, as discussed above in the context of Chapter 30, is an example of this.


p921: remove space between "25" and comma and add period to end of sentence.

The synthesis of a schema for man- building, as discussed above in the context of Chapter 25, is an example of this. p921: add period at end of page

based on many possible mechanisms, but most simply importance spreading.


p922: remove space between heading and period for:

Experience-Based Knowledge Selection for Language Generation.


p923: remove space between heading and period for:

Experience-Based Guidance of Word and Syntax Choice.

Adaptation of Dialogue Controllers.


p924: add period to end of sentence

Action selection must tell it when to continue building the same structure and when to try a new one, as well as more micro level choices.


p924: replace improperly quoted text:

After a failed structure (a Òfalse startÓ) (a “false start”) is disassembled, the correspond-

may get linked to the Goal Atom representing ÒgoodÓ “good” in this context. This


p925: not a sentence, replace with an endash

knows it should probably try to formulate some new concepts. Because concepts – because it


p925: replace improperly quoted text:

comes up with Òcar-manÓ “car-man”).

When the notion of ÒcarÓ “car” is brought to mind, the distributed map of

nodes corresponding to ÒcarÓ “car” get high STI. When car-man is formed, it


p925: replace improper "Ð" endash:

something matching a certain specification Ð e.g. specification – e.g. how to build a chassis


p926: replace improper "Ð" endash:

Coordinating different parts of the body Ð say body – say an arm and a hand Ð will hand – will

called its creation beautiful, which is recognizes as a positive evaluation Ð so evaluation – so

Appendix A: Glossary

remove opening "::", remove paragraph breaks begore first bullet entry:

This glossary clarifies the meaning of various terms as they are commonly used in the context of OpenCog & CogPrime. It doesn’t attempt to provide general definitions of the terms involved; it is intentionally highly OpenCog/CogPrime-specific. Where terms are used in multiple ways within the OpenCog/CogPrime sphere, more than one meaning is given.


Consider adding a glossary entry for Artificial General Intelligence (AGI).


AGI Preschool: Missing full stop at entry's end.


Atom, Realized: Missing full stop at entry's end.


Atom, Saved: Missing full stop at entry's end.


Atom2Link: Formatting error at OpenCogPrime's


Attention: Missing full stop at entry's end.


Attentional Focus: Missing word "a"

i.e. it’s a moving bubble of attention.


Backward Chainer: Missing full stop at entry's end.


Cognitive Synergy: word "overcome" separated:

synergy involves coupling them together in such a way that they can help one another over come each other’s internal combinatorial explosions.


CogPrime: space before closing parenthesis; missing full stop at entry's end:

(and this implementation is OpenCogPrime )


Cognitive Equation: space at wrong place:

”Chaotic Logic, ”that


Consciousness: strange character "—":

rest of the Atomspace as ”unconscious” but this is just


Consciousness: formatting error at "CogPrime design".


CSDLN: missing opening parenthesis:

Compositional Spatiotemporal Deep Learning Network):

Declarative knowledge: glossary term "knowledge" in lower case, should be upper case?


Declarative knowledge: strange character "—".


DeSTIN: Define acronym first - Deep SpatioTemporal Inference Network


Demand: space before apostrophe in "CogPrime ’s OpenPsi".


Dialogue: Formatting error at "CogPrime context".


Efficient Pragmatic General Intelligence: Closing parenthesis at the wrong place, should be behind "environment":

measure over (goal, environment pairs), and where


Embodied Communication Prior: Make "goal/environment pairs" to "(goal, environment) pairs" to keep it consistent with its use in the glossary.


Emotion: space before apostrophe:

due to CogPrime ’s use of


Episodic Knowledge: missing dash btw "agent's life history" -> life-history


Evolutionary Learning: missing plural 's' behind "variation"


Evolutionary Learning: spaces before ";" and apostrophe:

based portion of CogPrime ; but CogPrime ’s dynamics


Exemplar (MOSES): space before comma at "new deme , or" and missing full stop at entry's end.


First-Order Inference (FOI): derives -> derived


Forward Chainer: missing plural 's' behind "conclusion"


Frame2Atom: missing full stop at entry's end.


Freezing: missing full stop at entry's end.


Glocal Memory: entry could be a bit more explicit, in this form it's not representing the description on p303. // TODO: Suggest a better entry.


Goal-Driven Learning: strange character "—"


Grounded SchemaNode: missing full stop at entry's end.


Hebbian Learning: missing full stop at entry's end.


Importance Decay: what does the apostrophe at Atom' mean? is there missing something?


Importance Spreading: missing full stop at entry's end.


Induction: minus before the implication arrow B -> C


Inference Control: missing full stop at entry's end.


Integrative AGI: space before comma:

An AGI architecture, like CogPrime , that


Intension: missing full stop at entry's end.


Internal Simulation World: unclear use of apostrophe, letter 'q'

with various action‘ q sequences


Language Comprehension: strange character — at "Frame2Atom — Atomspace"

inconsistent use of ':', put on after pipeline at "alternatively a pipeline Link Parser"


Language Processing: space before comma at "in CogPrime , into"

missing full stop at entry's end.


Linguistic Links: strange character É at CharacterNodeÉ.


Link Parser: formatting error at "OpenCogPrime's" where the apostrophe should be.


Lobe: the use of "sometime" means that the term Lobe might be used in the future. Is that the intended meaning? Otherwise use "sometimes" for occasional use.


Long Term Importance (LTI): missing 'i' in multple at "have multple LTI values".


Map: missing full stop at entry's end.


Memory types: "types" should be uppercase; 2x formatting error after CogPrime (new line).


Mind-World Correspondence Principle: replace "as" with "and" between world and mind?

by modeling the world as mind as probabilistic


Mind OS: missing full stop at entry's end.


Moving Bubble of Attention: missing full stop at entry's end.


Natural Language Processing: add acronym (NLP)


Node: Missing closing " after "0-ary

Missing "of" btw "Some types of Nodes refer"


Novamente Cognition Engine (NCE): Kill the second half of the sentence (for redundancy reasons) "Many parts of the NCE were open-sourced to form portions of OpenCog, but some NCE code was not included in OpenCog"


OpenCog Prime: remove space; add acronym - s/b: OpenCogPrime (OCP)

missing full stop at entry's end.


Pattern Recognition: missing full stop at entry's end.


Probabilistic Logic Networks (PLN): formatting error at "OpenCogPrime's" where the apostrophe should be.


Procedure Learning: missing full stop at entry's end.


Psi: formatting error at "OpenCogPrime's" where the apostrophe should be.

Remove comma btw "motivated action, and emotion"


Psynet Model: remove space before comma at "underlying CogPrime , referred"

One 'y' too much at "psyynet model"

remove space before comma at "underlying CogPrime , but the name"


Region Connection Calculus: Is the comma correct there? "part of spatial inference, to provide relations"


Request for Services (RFS): remove space before apostrophe at "CogPrime ’s Goal-driven"; replace "a RFS" with "an RFS"


Robotic Preschool: missing full stop at entry's end.

Appendix B: Steps Toward a Formal Theory of Cognitive Structure and Dynamics

p951: replace "Section ??" to proper reference(s)

Transforming the conceptual and formal ideas of Section ?? into rigorous


p952: bogus reference

and 21, then one can consider tree edit operations as defined in [Bil05] as


p953: replace "3" with "three"

characterized by 3 three spatial dimensions and one temporal dimension, one way to


p961: margin overflow

[YGSS10] defines the generalized complexity criterion KΦ (x) = mini∈N {Φ(i, τi )|L(pi )) =


p964,5: (minor) margin overflow of expression


p965: is this Chapter 23? if so, then be explicit.

scribed briefly in 23 where we show that (in some small examples at any rate)


p965: be consistent - capitalize "About" in heading


p965: be consistent - American English

and doing so is part of our research programme going forward.


p969: these bullets make little structuring sense - remove them all from "h."

h. Hybridization of one’s integrative neural-symbolic system with a spa- tiotemporally hierarchical deep learning system is an effective way to handle representation and learning of low-level sensorimotor knowl- edge

  • DeSTIN is one example of a deep learning system of this nature that can be effective in this context

The mindspace interpretation includes the observations that

  • Linkages between the internal nodes of DeSTIN and the
  • Spatio-temporally hierarchical perception-action systems like DeS-TIN have high syntax-sematics correlation for sensory knowledge, as they embody the spatiotemporally hierarchical structure of the perceived world

Appendix C: Emergent Reflexive Mental Structures

general: all references to Figures beyond Figure C.6 need to be checked, non of the remaining figures are labeled/captioned, e.g. p998, which talks about two different figures, but the same reference:

The simplest octonionic mirrorhouse is simply the tetrahedron (see Figure C.12). More complex octonionic mirrorhouses correspond to tetrahedra with extra mirrors placed over their internal corners, as shown in Figure C.12.


p971: use numbered footnote, fill in missing chapter number

∗ These patterns go beyond the relatively well-accepted network structures reviewed in chapter ??, and constitute an


p981: bad syllable hyphen break

of the latter, without this action of will violating the notion of deterministic/s- tochastic equations of the universe.


p973: replace "paper" qith "appendix"

the standard axioms of set theory. The main work of this paper appendix is to suggest


p986: space before endash

whereas the implicit level consists of “maps” as mentioned above– collections


p989: replace "paper" qith "appendix" go beyond the scope of this paper appendix. For instance, most knowledge in Cog-


p991-2: wrong reference format

(Carruthers and Smith, 1996). But empathic simulation is a key component, and we sug-gest that, in normal humans (autistic humans may be a counterexample; see Oberman et al, 2005


p994: the numbered list in combination with the S1-S7 lead-in is rather redundant (suggest bullets instead), also consider breaking out "The next 3 stages distinguish the hominid line from that of the great apes:" as a new paragraph (shouldn't this be "next 4 stages", or just "remaining stages"?)


p995: consider making variables/expressions within paragraph text italic. e.g. (and carry forward to rest of appendix)

formed by positing three imaginary numbers i, j and k with i*i=j*j=k*k=-1,


p996: no table caption, also consider center aligning cell contents and clearing out UL cell.


p996 replace "?" with " " - 3x

Mirrorhouses are structures built up?from mutually facing mirrors which reflect each others’ reflections. The?simplest mirrorhouse possible to con- struct is made of two facing?mirrors


p1000: wrong reference format

David Finkelstein and Ernesto Rodriguez (1984)


p1003: wrong reference format

For another sort of example, consider the act of creativity, which in (Go- ertzel, 1997) is modeled in terms of a “creative subself”: a portion of the


p1004: wrong reference format - 2x

motivation. According to Freud (1962), much of our psychology consists of

ego with the neuropsychological structure that Thomas Metzinger (2004)


p1006: American English, please

engineering programme is successful, the emergence or otherwise of the struc-

Appendix D: GOLEM: Toward an AGI Meta-Architecture Enabling Both Goal Preservation and Radical Self-Improvement

Title too long - text overflows even page heading


p1011: use numbered footnote; replace "article" with "appendix" in footnote: "upon reading an earlier draft of this article appendix"


p1014: be consistent - capitalize "About" in heading


p1016: add "RL" acronym to first reference; add to glossary

Readers accustomed to reinforcement learning (RL) approaches to AI [SB98] may


p1018: fix left smart quote and add space

then the”predictably beneficial” part of the problem of "creating predictably


p1018: replace ellipses with closing endash

as new information about the world is obtained, and as the world changes ...


p1018: use numbered footnote

† AGI systems.


p1020: remove extraneous "T"

lines will be pursued for later publications. T


p1020: add correct chapter reference

marks on self-modifying CogPrime systems from Chapter ??. But the major

Appendix E: Lojban++: A Novel Linguistic Mechanism for Teaching AGI Systems

p1023: add apostrophe to "CogPrime's" - (otherwise tough to find - as opposed to the use of p1028 "powerful cognitive and pragmatic interplay between humans and CogPrime s.")

CogPrime s. Another way to think about the Lojban++ approach is that


p1027: number and label table


p1028: wrong reference format

(Powell, no date specified),


p1028: user numbered footnote

but on the order of man-weeks rather than man-years of effort.∗ Once this


p1030: rename heading from "Conclusion" to "Discussion" or something else (it is only halfway through the appendix!)


p1032 lowercase "Appendix"

so it should be assumed that the specifics described in this appendix are


p1036: inconsistent capitalization of heading - not sure is appropriate

Lojban gismu of clear use within Lojban++


p1037: replace "a list of <50" with "a list of less than 50"

Appendix F: PLN and the Brain

p1041: add space after endash

itive rather than mathematically rigorous –it would seem that we can build


Appendix G: Possible Worlds Semantics and Experiential Semantics

p1050: put in quotes and capitalize "What":

"What would it imply about the world if the items in this subsample were the only things I’d seen?"


p1050: add acronym after "science fiction" (SF)


p1055: margin overflow

first-order inferences, consider the case of intensional inheritance. IntensionalInheritance A B

Appendix H: Propositions About Environments in Which CogPrime Components are Useful

general: replace "OCP" with "CogPrime" (since that is what is used in chapter title


general: add colon ":" after each "Proposition ##" lead in, e.g. p1065

Proposition 9: In a universe of sets where intensional similarity and ex-


p1059: capitalize "are" in title

Propositions About Environments in Which CogPrime Components Are Useful


p1059: be consistent - capitalize "About" in heading

H.1 Propositions about MOSES


p1062: quote question:

"Given a particular choice regard- ing syntactic program representation, what sorts of problems match MOSES’s biases as induced by this choice?"


p1065: remove ellipses at end of Proposition 13

is a probabilistic subset of B” and “A is a pattern in B” tend to be smooth....


p1065: missing period at end of Proposition 13 note

same thing as saying that syntactic and semantic similarity are well-correlated among members of S


p1065: missing period at end of Proposition 14

Proposition 14 The set of motor control programs, for a set of standard actuators like wheels, arms and legs, displays a reasonable level of correlation between syntactic and semantic similarity

References

p1065. duplicate reference

Goe06a. Ben Goertzel. The Hidden Pattern. Brown Walker, 2006.

Goe06b. Ben Goertzel. The Hidden Pattern. Brown Walker, 2006.


p1073: GMIH08 - missing comma after "I. Goertzel"