Reviewers' comments

------------------------------------------------------
Reviewer A:
Recommendation: Decline Submission

------------------------------------------------------

Submission title

Rethinking Human and Machine Intelligence through Kant’s Incongruent Counterparts

Aptness of the title

3

Aptness of the abstract

3

Logical structure of the manuscript

5

Acceptability and justification of the main claims

5

Originality of the main claims

3

Relevance and sufficiency of the sources

4

Please, justify all of your objections to the manuscript.

While the topic might be of interest, the premises are outlandish (classical paradigm in cogscience is dead long time ago; a single sentence about consciusness is totally insufficient and biased; the quote from Wittgenstein is, correctly, about what he notes, not what he proves and thereby deserves no merit) and thereby the arguement is unconvincing. It would take way more philoshophical work to get th eto crux of the matter.

 

==========================================================

 

Reviewer: 1

Recommendation: Reject

Comments:
The paper may contain interesting and original thoughts, but the manner in which they are currently laid out does not recommend publication.

1. The cases posited at the beginning of the paper are unclear. In a world where only a left hand existes, how could it be impossible for a right hand to enter (LH1)? And why would a right hand be inconceivable in that world? And if a right hand can enter into that world (LH2), how could it not be perceived to be different from a left hand? (The cases are significantly different from the ones imagined by Kant, by the way...)

2. The definition of "deterministic world", on p. 4, is circular. It says that a world is deterministic iff the events of that world are... deterministic.

3. The definition of "deterministic knowledge", on the same page, is inadequate. It defines knowledge in terms of a totality of facts, and makes no reference to a subject who knows.

The main arguments and conclusion of the paper turn on these initial cases and definitions. Since the latter are found wanting, so are the former.

Additional Questions:
Does the manuscript contain new and significant information to justify publication?: No

Does the Abstract (Summary) clearly and accurately describe the content of the article?: No

Is the problem significant and concisely stated?: No

Do the arguments developed in the manuscript support the author's conclusion(s)?: No

Is adequate reference made to other work in the field?: No

Length of article is: Adequate

Please state any conflict(s) of interest that you have in relation to the review of this paper (state “none” if this is not applicable).: none

Rating:

Interest: 3. Average

Quality: 4. Below Average

Originality: 2. Good

Overall: 4. Below Average


Reviewer: 2

Recommendation: Reject

Comments:
The abstract and the introduction claim that the paper aims at providing a metaphysical framework to distinguish between human and artificial intelligence. Abstract and introduction promise the following argument strategy: distinguish two worlds
- deterministic world with human intelligence
- determinist world with machine intelligence.
The author promises to show how human and machine intelligence differ.
These two promised cases are advanced by analogy with Kant’s Incongruent Counterparts (IC). Discussion of Kant’s IC is imprecise and displays a lack of overall understanding of the aims of the argument. I believe the author could set out to argue in the manner promised without discussing Kant’s IC. As things stand, the discussion of Kant’s IC deters from the overall understanding and clarity of the paper.

Moreover, these two cases are never discussed. By contrast the author discusses three types of deterministic knowledge. It is unclear how these three cases are supposed to help the author’s main argument. Moreover, the cases are under-specified and unclear. The author provides quasi-formal descriptions but the lack of clear background and of details deters from the precision of formal description.

The author claims to solve the vantage point problem. However, they have not achieved to do so because it is unclear how their discussion bears on the issue.

More specific points:
Page 9 lines 40-44 poor English
Page 10 line 20 “humanistic” doesn’t seem the right adjective
page14 and 15 Tegmark car analogy: the author seems to misunderstands Tegmark’s analogy (at the very least add discussion). The point is not to write a manual for me to be able to use the car but a manual on how the car works. Tegmark’s manual is meant to be a  “hands-on user’s manual describing which are the most important components and how to use them.”

Additional Questions:
Does the manuscript contain new and significant information to justify publication?: No

Does the Abstract (Summary) clearly and accurately describe the content of the article?: No

Is the problem significant and concisely stated?: Yes

Do the arguments developed in the manuscript support the author's conclusion(s)?: No

Is adequate reference made to other work in the field?: No

Length of article is: Adequate

Please state any conflict(s) of interest that you have in relation to the review of this paper (state “none” if this is not applicable).: None

Rating:

Interest: 3. Average

Quality: 5. Poor

Originality: 5. Poor

Overall: 4. Below Average

 

 

 

Review of „Rethinking Human and Machine Intelligence through Kant, Wittgenstein, and Gödel”

 

The article has significant flaws - the title promises to analyze intelligence, and the introduction announces the creation of a model "allowing a determinist to validly claim our universe as deterministic while remaining part of it." Evidently, the goals of the article were ill-defined. The concepts used in the article were chosen eclectically, and the rationale for using them is that they allow for a preconceived solution. Of course, eclecticism is one of the methods of philosophy, but here it is not clear what the rationale behind its use is. There was also no review of the existing literature on determinism and deterministic knowledge, which is another very important shortcoming of this text. Kant's views have already been used many times in philosophical discussions through the 19th century, the 20th century and the current 21st century. Thus, the novelty of the article may be mostly an illusion resulting from ignorance of history, and as is well known, Kant's influence was extremely strong and multidirectional.

 

A major drawback of the article is that it assumes a metaphysics based on common experience. Such a metaphysics was justified in Kant's time, but today we know that in many places it does not even meet the requirements that modern sciences, especially physics, place on metaphysics. The revolution that took place at the beginning of the 20th century forced philosophers to modify their ontological views of reality. Perhaps the last of the great philosophers to base their views of the universe on the metaphysics of ordinary experience was the early Wittgenstein of the Tractatus logico-philosophicus period. Later, however, Wittgenstein clearly understood that this was the wrong path and he modified his philosophy. Unfortunately some neo-positivists never understood it.

 

The article is based on a lot of loose associations, and lacks correct, detailed argumentation. For example, the arrows on page 5, for example, cannot substitute for argumentation. They are certainly no substitute for critical analysis of the solutions adopted, which is essential here. In the text we can observe jumps between facts and linguistic representation. It is also bizarre to call the adoption of part of D-knowledge as "metaphysical sense". In his/her considerations, the author operates once with abstract concepts (agent) and once with more concrete ones (human agent). On top of that, he/she still introduces the concept of information while introducing mappings, which he/she is not too concerned with.

The concept of block universe model was introduced only on page 11, although it was implicitly already assumed from the definition of D-knowledge.

The conclusions are limited only to the articulation of the conceptual scheme, which is not the result of correct reasoning and is not well-founded.

 

Regarding the current form of the article, it cannot be accepted for publication. In order to improve, first of all, it is necessary to clearly define what is the essential novelty of this position and what is the purpose of the article. It is also worth considering what is the value of this type of analysis - whether they answer a specific theoretical problem. If they do not answer a specific problem, it will be difficult to justify their theoretical value, and the practical value will also be unattainable. However, when one succeeds in putting these considerations into the context of existing philosophical problems, one must then deepen the study of the concepts used in order to respond critically to them. The concepts used need to be clarified and the entire concept of the article needs to be reworked, as it is untenable in its current form.

 

The language and style are generally correct, but numerous errors in the logical structure of the work and in the argumentation mean that clarity is lacking.

 

Relevance of Keywords and the Abstract to the title. The title problem of intelligence is only touched upon at the end of the article, and the entire text focuses on the issue of determinism and deterministic knowledge in the context of the cognitive subject. Thus, neither the title nor the keywords basically match the content.

 

The decision is: to decline submission.

 

============

Journal

Peer-Review

 

Title of the article:

Rethinking Human and Machine Intelligence through Kant, Wittgenstein, and Gödel

Decision (please leave just one):

-          To suggest to the author to take into account the comments presented below and to submit an article for reviewing again. The paper requires a lot of work and rethinking of the argument and structure.

-           

Criteria	Evaluation Good Medium Weak	Comments
Relevance of the subject and explication of the problem	weak	The topic is relevant and important, but the explanation seems to be weakly justified.
Explication of the thesis	weak	The author makes several key assumptions without justifying them. For example: The D knowledge specific to the universe can be likened to the Gödel number assigned to a string (or mathematical theorem).  Also, the agent can be compared to a variable in the theorem
Clarity and validity of argumentation	weak	The paper seems to be a collection of ideas put together but without much justification for doing this. For example: Can we “substitute” Wittgenstein into Godel’s proof?
Acquaintance with contemporary research of the subject	weak	The author is touching on deep philosophical problems: determinism, agency, machine intelligence, human intelligence, cognitive agents, etc. But the references supporting the author’s claims and interpretations of terms are sparse and seem not to indicate a good grasp of the most recent relevant sources.
Conclusions – do they make any contribution to the field?	Very weak	Conclusions seem to be taken from the slide presentation and are not very relevant to the original claims in the abstract.
Language, style	Good	English is correct in my view. The author should pay attention to the consistency of reference entries.
Relevance of keywords and Summary to the title	moderate	The abstract is clear. However, the thesis from the abstract is not reflected in the conclusions.

 

Reviewer:                                     (not revealed to the author)

 

If the article needs any improvement, your comments would be a great help for the author.

 

The idea in the paper is certainly relevant and important. However, the exposition is not very coherent. The paper requires a lot of work. The paper requires rethinking of the argument, clearer exposition.

 

A few points to substantiate this comment:

• Gödel is introduced without much explanation as to why it is there.
• Several key notions—agent, human agent, machine agent, and machine intelligence—are not defined.
• Why does the author select Kant, Wittgenstein and Gödel to talk about machine intelligence? There are plenty of recent publications on the topic that seem closer to the discussed problem. The fact that they “said this and that” is not a justification. We need to see why them not others.
• The introduction does not lay out the structure of the paper. The structure seems to be divergent from the goal stated in the abstract.
• Key definitions seem to come from the author‘s personal views. They should be based on some references; why this and not some other definition?
• The author makes several key assumptions without justifying them. For example:
• The D knowledge specific to the universe can be likened to the Gödel number assigned to a string (or mathematical theorem).
• Also, the agent can be compared to a variable in the theorem.
• The Nietzsche reference is distracting, as he is nowhere referenced in the paper. And so is his amore fati.
• Determinism is a very complex term. The author does not provide the context for his or her definition. For example, what kind of determinism are we talking about?
• The claim is that if readers patiently follow this paper’s arguments to the end, they will see how it establishes a plausible model allowing a determinist to validly claim our universe as deterministic while remaining part of it. “ seem hubristic, meaning that if the reader „does not follow patiently,“ he or she will not get it. It assumes that if someone does not understand the argument, it is his or her fault. This should not take place in a scientific paper.
• The thesis in the abstract is: By postulating the distinctiveness of human over machine intelligence, this paper resolves what it refers to as “the vantage point problem," namely, how to legitimize a determinist’s assertion of determinism by placing the determinist within the universe. I do not see them reflected in the conclusion.
• I would not agree with the claim, „When a determinist asserts that the universe is deterministic, this requires assuming a hypothetical vantage point from which to describe the universe.“. I may claim that the universe is deterministic without making an assumption of God‘s eye position.
• „machine agent emulating the human agent in a causal manner“ Are we talking here about „zombies“? What does it mean „in a causal manner“? Is it complete emulation, behavioral only, or what kind of emulation?
• „The human agent’s symbolic manipulation, for instance, may take place through neural activities in the brain." Why this assumption? Can we justify it?
• The mind is a “classical von Neumann computer"—von Neumann is a hardware model. May be the author has in mind Turing machines?
• „This world is intentionally designed to avoid being based on a “connectionist” model." Does the author mean this world or the model of the mind?
• What are „counterfactual input events“? Explain.

 

 

===========]

Does the submitted paper match the thematic scope of asdf??

Yes, the submitted article fits within the thematic scope of asdf. The article discusses an interesting approach to differentiating between human and machine intelligence in the context of determinism. The author constructs a metaphysical framework based on two identical, deterministic worlds in which agents operate: human and machine, demonstrating differences in information processing mechanisms, despite external causality similarity. This is attempted to be illustrated through a thought experiment, referring to existing philosophical concepts and common knowledge.

What is the academic level of the submitted manuscript, and does it satisfy general criteria of the academic publication?

The author demonstrates a good understanding of the topic as well as the ability to critically analyze and synthesize existing theories. The text is well-organized, contains appropriate references to the literature, and meets the criteria for academic publication, such as clarity of argumentation, originality of contribution, and significance of the problem studied.

Is the submitted paper cognitively valuable, especially, does it offer a new, original knowledge on the investigated topic?

The article is cognitively valuable, offering a new perspective on the problem of determinism in the context of the growing capabilities of artificial intelligence. Through the analysis and comparison of the decision-making processes of human and machine agents, the article makes an original contribution to the debate on the limits of intelligence. Such an approach expands the current understanding of the subject and stimulates further research in this field.

What are the recommended changes?

The article "Rethinking Human and Machine Intelligence under Determinism" constitutes a valuable contribution to the debate on the differences between human and machine intelligence. Although certain aspects require further clarification or expansion, this work poses intriguing questions and proposes an original approach to considerations of the nature of intelligence.

 

My remarks:

1.

Determinism and the differences between human and machine intelligence:

The author proposes a metaphysical framework to distinguish between human and machine intelligence, assuming that both types of agents – human and machine – operate in identically deterministic worlds, yet present different information processing mechanisms. This thesis may be controversial, as it introduces the question of the possibility of a fundamental difference between the cognitive processes of humans and machines, which stands in opposition to the prevailing belief about the increasingly blurred boundaries between human and artificial intelligence.

 

Recommendations:

The author should consider expanding the theoretical foundation of the work, referring to a wider range of literature on determinism, philosophy of mind, and artificial intelligence. The paper could benefit from more discussion on how the results and theses fit into current debates and research on the capabilities and limitations of AI, as well as how they might influence future research directions.

 

2.

Limitations Regarding D-Knowledge:

The concept of D-knowledge presented by the author can be interpreted as an attempt to understand and describe the universe through a deterministic approach, which assumes that events and states of the world are fully determined by their causes and can be predicted or described using certain principles or rules. In the context of the article, D-knowledge is presented as a way to understand the differences between human and machine intelligence in a deterministically structured world. However, the author acknowledges that this concept encounters criticism from quantum physicists. The main point of contention is that at the quantum level, describing physical events with precise spatial and temporal coordinates becomes impossible due to the fundamental principles of quantum mechanics, such as Heisenberg's uncertainty principle. This principle suggests that it is not possible to know all the parameters of a particle, such as its position and momentum, simultaneously, which contradicts the deterministic assumption that all events are fully determined and predictable. The author's admission that the concept of D-knowledge may be problematic in the context of quantum physics undermines one of the key assumptions of the article, as it suggests that a deterministic approach to describing the universe may not be fully compatible with the current understanding of quantum mechanics. This means that the thesis presented in the article may not be able to fully account for all aspects of reality, especially those fundamental to understanding behaviors at the quantum level, posing a challenge to attempts to apply this thesis to a full description of physical phenomena and differences between human and machine intelligence.

 

Recommendations:

Determinism is a key concept in the work, yet its definition and application may seem ambiguous. The author should more precisely define what is meant by determinism in the context of their considerations, whether it refers solely to the physical laws governing the universe or also to determinism in algorithms and thought processes.

It would be worthwhile to consider whether different forms of determinism (e.g., physical, logical, psychological) have equivalent implications for understanding intelligence. Expanding the discussion on this topic could enrich the work with new perspectives.

 

 

 

===============================================================

 

This paper ‘s concerns are twofold: first, and most importantly, to set out a metaphysical framework for distinguishing between machine and human intelligence, and second, to resolve what the author (Au) refers to as the ‘vantage problem’ (in Au’s words, “how to legitimize a determinist’s assertion of determinism by placing the determinist within the universe”). To this end, Au draws on many different references to other works to use as analogies between Au’s main subject mater (human vs machine intelligence). However, there is no discernible coherent argument in the paper. References are made to Kant’s argument for incongruent counterparts without any clear explanation of how and in what way this argument is relevant to the intended argument Au wishes to mount (Au concedes on page 3 that Kant’s argument, being concerned with the question of whether space is relational or absolute, does not seem to have much to do with Au’s main aim, and as the paper develops, only the references to right-hand and left-hand worlds are retained for the stipulated use to which Au intends to put them. By page 5 Au has referred to Witgenstein’s ‘atomic facts’ and Carnap’s “one state-description” without any clear explanation of what work they are doing the argument to be developed. Godel numbers and theorems are introduced in order to liken the human agent to a variable in one of Godel’s mathematical theorems and to liken a Godel number to deterministic knowledge fed back into the human agent’s information processing mechanism. This analogy is not explained, nor does it do any real work in the remainder of the paper. As the paper goes on, there are many other examples of references to the work of others and analogies drawn where it is unclear what work the references and analogies are doing in aid of the main argument of the paper. A distinction is drawn between rigid and emergent processing, and nowhere does Au explain what emergent processing is nor why human agents whose processes are emergent can produce novel unpredicted outputs. Since this is essential to the main aim of the paper (distinguishing between machine and human intelligence) it simply looks question begging against those who might deny that distinction. In sum, there is no discernible coherent argument in this paper, and it is highly unlikely that any improvements would lead to successful publication.

===============================================================

 

Start of First Referee Report:

I am writing with my referee report for asdf of the paper “Rethinking Human and Machine Intelligence through Kant’s Incongruent Counterparts.”

 

Unfortunately, I recommend that this paper be rejected on the grounds that it is unsuitable for publication.

 

The paper presents a puzzle and attempts to solve that puzzle. The puzzle concerns the possibility of an epistemic agent’s coherently holding that the universe in which it exists is deterministic. The author holds that because the agent must add a fact to that universe in holding this claim, its claim that the universe is deterministic cannot encompass everything that exists in the universe (because it cannot encompass the claim itself). The solution that the author proposes is that human agency has at least two aspects, one of which is its existence as part of a deterministic existence, and another is as an agent with free will existing outside of that universe. It is the latter aspect that holds that the universe is deterministic, but since it does so from “outside” that universe, that claim itself need not be included in the constituents of the universe.

 

I have to admit that while the above is my best reconstruction of what this paper presents, I understand very little of what I have just written. I don’t see what is puzzling or incoherent in the first place, and I don’t understand how the solution to the presented puzzle solves the puzzle rather than avoiding it altogether. When I say that I don’t understand either of these things, it is not in the sense that philosophers sometimes use, where it means that I disagree. I genuinely do not know what the author of this paper means by these claims, or how they see them as fitting together. While the paper rehearses several definitions of terms, it does not explain or defend its central claims, except in the most vague and metaphorical ways. So, I genuinely understand very little of what the author is saying here.

 

Even so, I can formulate at least a few points that the author should consider addressing, if they are to continue pursuing this line of inquiry. 

Why can’t the claim of determinism be just another one of the facts made from within the universe itself? 

If an agent exists outside the deterministic universe, then won’t its actions constitute violations of the deterministic nature of the universe, or won’t the universe not include everything that exists.

What does this have to do with Kant at all? For all the prominence that Kant is given in the title and framing of the paper, it seems to me that the only thing that the author gets from Kant is the notion of comparing two different possible worlds to each other, which is a fairly standard practice in philosophy.

The paper would benefit enormously from more of an engagement with the contemporary literature on determinism and free will. These are issues that have received extensive treatment, spanning centuries, and framing the author’s points against the backdrop of that ongoing discussion might make those points easier to comprehend.

 

I hope that report is of some help.

End of First Referee Report.

 

 

Start of Second Referee Report:

I've now carefully read the essay you sent me for review, "Rethinking Human and Machine Intelligence through Kant’s Incongruent Counterparts."

 

It's an ingenious & original argument, pulling together metaphysical issues in the debate about free will vs. determinism, metaphysical issues in the mind vs. machines debate, & historical issues in Nietschean & Kantian philosophy.

 

All in all, I very much enjoyed reading & thinking about it.

 

My one worry is the following.

 

Suppose that we accept the author's conclusion, i.e., that we can solve

 

"what [the essay] refers to as 'the vantage point problem' – namely, how to legitimize a determinist’s assertion of determinism by placing the determinist within the universe."

 

Then the hard determinist's assertion of the thesis of hard determinism is itself predetermined by all the settled facts about the past together with the laws of nature.

 

So the hard determinist would assert the thesis of hard determinism even if the thesis of hard determinism were false.

 

So there's no intrinsic or non-accidental connection between the evidential process leading to the hard determinist's assertion of hard determinism, & the truth of the thesis of hard determinism.

 

But, knowledge of the thesis of hard determinism requires the truth of that thesis & also an asserted belief that's justified by means of an evidential process with an intrinsic or non-accidental connection to the truth of the asserted belief.

 

So the hard determinist isn't justified in asserting hard determinism.

 

So if hard determinism is true, then the hard determinist can't know it.

 

It's very likely that the author will have a quick-&-sufficient, or in any case compelling, response to this worry, hence my judgment is that the essay should be conditionally accepted for publication provided that they respond to this worry.

 

And one other thing.

 

For some time now, I've been using a "single blind" refereeing/reviewing policy, according to which the author remains anonymous, but when I submit my report, I reveal my identity to the author.

 

This is because I think that, for various reasons, my policy is much fairer to authors than the standard "double blind" system.  

 

Above all, I think it encourages the referee to be philosophically constructive & responsible to the author.

 

End of Second Referee Report.

==========================================================
Reviewer: 1

Comments to the Author
My major comments are those.

Major comment 1. About the fact-first metaphysics. Central to author’s notion of D knowledge is this: (At least most part of) the knowledge significant to agents is the knowledge relevant to facts about events. As the author says himself, he is inspired by Wittgenstein’s fact-first picture from Tractatus.

I am not saying the author therefore also shares the burden of showing how this controversial view could be true. But to respect his potential objectors, the author needs to consider some counterpoints of this view he utilizes.

Given the importance of D knowledge to the author’s proposal, I think the fact-first picture the author utilizes needs more substantiation. And I recommend the author discuss some problems of fact-first metaphysics mentioned in Jason Turner’s book Facts in the Logical Space (OUP 2017).

Major comment 2. About the role of D-knowledge in causation. One way the author uses his notion of D knowledge is how it effects the agent’s future actions (like in the Millicent drinks coffee case). Presumably the author needs to say something about how some mental states can causally effect events in a world. I am not saying the author needs a specific position on mental causation. But he at least needs to make some clarifications, which I fail to find any in the current manuscript.

Major comment 3. About the strong AI. I agree the current pieces demonstrates how human intelligence differs from machine intelligence by virtue of the capability of emergently processing D knowledge. But it seems that the author only takes the machine intelligence at issue to be current/weak AI. This setting seems to be, even if not ad hoc, disrespectful to the potential of machine intelligence.

I hope to hear the author’s answer to following questions: What if the machine intelligence we discuss here is a kind of strong AI that can also emergently processing D knowledge? Can the author’s schema still say something differentiating strong AI from human? Or maybe the author has some reasons to show why we needn’t care about weak/strong AI distinction in this context?

And here are also some minor comments.

1. There should be a space between ‘p.’ and ‘6’ in “Beraldo-de-Araújio’s definitions on p.6” on page 7.

2. I find the illustration of mappings in types 1, 2 and 3 determinism misleading. It is unclear what the author wants to show: Graphics, charts or mathematical formulae? I suggest redoing the illustration.

Reviewer: 2

Comments to the Author


I had trouble determining what the paper was about. The author's abstract gives as good a summary as I could offer: “This paper proposes a metaphysical framework for distinguishing between human and machine intelligence. By drawing an analogy from Kant's incongruent counterparts, it posits two identical deterministic worlds – one comprising a human agent and the other comprising a machine agent. These agents exhibit different types of information processing mechanisms despite their apparent sameness in a causal sense. By postulating the distinctiveness of human over machine intelligence, this paper resolves what it refers to as 'the vantage point problem' – namely, how to define the determinist's reasoning mind in relation to the universe where the determinist belongs.”

I couldn't figure out – despite reading the paper five times – what conclusion the author intended to support. Did he/she hope to lay out a metaphysical framework whose usefulness he/she then illustrated with examples about human and machine intelligence? Did he/she aim to distinguish human from machine intelligence, with the metaphysical framework a tool toward that end? Did he/she concern himself/herself primarily with determinism? Did his/her arguments funnel into a resolution of the vantage point problem, with this the paper's main purpose? Nor could I discern how the paper's parts hung together as strands of a single extended argument. The paper instead came across as a series of loosely related vignettes held together mostly by the fact that the author had given each of them some thought.

More importantly, I found the core arguments of the paper – from page 9 or 10 on, say – impossible to understand. Many of the sentences contained terms, or made assumptions, that left their meanings opaque. And – typically, alas – I couldn't divine how one sentence followed from the sentences before it. The paper therefore didn't read as a continuous argument. Rather, it struck me as a collection of isolated and oracular utterances few of which were fully understandable.

The summary outline on page 18, as well as the gestalt impression left by the body of the paper, strongly suggests that the author has good and interesting ideas and maybe even an argument – or the beginnings of an argument – in support of some important conclusions. But the paper itself falls far short of doing any of these justice. The author needs to make the purpose of his/her project much clearer. He/she needs to present – and maybe develop – his/her arguments much more fully. And he/she needs to write a paper that his/her audience can make sense of.

The following more specific comments support this assessment and also provide additional feedback.

1.  p. 2, 1st full paragraph: The author claims that the determinist needs a vantage point from outside the universe in order to claim that the universe is deterministic. Why? The quote from Wittgenstein and the bow towards Godel don't establish this. In particular, Godel makes a very technical – and technically circumscribed – claim about the relation of certain kinds of arithmetic to his specifically defined notion of decidability. It isn't obvious that any of this sheds light on the relation of the determinist to his/her world. I don't see, for example, that anything undermines a person's epistemic warrant for ascribing determinism (or indeterminism) to the universe he/she lives within. I can't see that his/her existence as a part compromises his/her capacity to render true judgments about the whole. True, maybe he/she lacks the mental chops to get at the fact of the matter. But his/her status as a part doesn't seem to have much to do with that mental capacity. So I think the author needs an argument – certainly more than a gesture to Wittgenstein and Godel. Otherwise I'm not sure the vantage point problem even exists.

2.  pp. 3-4, discussion of incongruent counterparts: Why introduce incongruent counterparts and the four hypothetical worlds, LH1, LH2, RH1, and RH2? At the least, the author should motivate this discussion, so the reader knows where it leads, and hence why he/she should bother understanding the details.

3.  p. 4: The definition of D knowledge seems clear enough, but the definition of a metaphysically open determinist world is opaque. What is the author getting at? What does 'metaphysical sense' mean (the discussion on pages 5 & 6, including footnote 3, don't help).

4.  pp. 4-5: Does the invocation of Carnap and Godel add anything to the argument (beyond giving it a patina of authority and some technical cache)? It reads like a digression and, I thought, makes everything more complicated than it needs to be.

5.  p. 5: I found the schematic at the bottom of page 5 very hard to understand. It certainly didn't clarify the previous discussion of Godel. Again, what does the author mean? Where is the argument going? What purpose does this part of the paper serve? How does the reader benefit from doing the push-ups needed to make sense of the author's technical flourishes?

6.  p. 7: Who is Beraldo-de-Araujio? He doesn't figure in the bibliography. I looked him up but couldn't find a reference to him.

7.  pp. 6-8: The author's discussion of the two hypothetical worlds – one with a human agent, the other with a machine agent – is introduced so casually! And the definitions (or descriptions?) of them are so sketchy! What is really going on here? The author needs, at least, to flesh out these worlds – or the key concepts they embody or illustrate. It isn't enough to toss in passing references to this thinker or that idea (e.g. Beraldo-de-Araujio, Polak & Krzanowski, von Neumann computer, Frankish & Ramsay), without developing any of those allusions. How does the author know that the two intelligences, with different kinds of processing, could be causally identical (by which, presumably, he/she means that for the same causal inputs they generate the same causal outputs)? Again, the author needs to let his/her readers know where his/her discussion's going, what it's in service of.

8.  p. 8: I found the discussion of Type 1 far too brief. Among other things, what does the author mean by “rigid processing” and “trivially deterministic”? Why can the world be represented by LH1”? How do incongruent counterparts bear on the discussion so as to be relevant to it? Maybe the author finds all this obvious, and I've just failed to put it together. But my point is that the paper doesn't make it easy for a reader to make sense of what's going on.

9.  pp. 8 – 9: I understand that the author uses incongruent counterparts as an analogy to the two different, deterministic worlds. But he/she doesn't specify in what way, nor how close the analogy is, nor what lessons the reader should draw from it. Nor does he/she explain how this analogy relates to his/her talk about, for example, LH1's representing one of the hypothetical worlds.

10.  pp. 9 - 10: What purpose do the formalisms and mappings serve? Are they intended to advance the argument, illustrate the argument, or merely extend the analogy?

11.  pp. 11 – 14: My sense is that these pages contain the core of the paper's argument. But I found them fiercely hard to understand. I could understand some of the sentences – in many cases only more or less. But I had great difficulty seeing how one sentence followed from the next. I felt like I was reading a string of non sequiturs. I'm sure I wasn't. But the author has to make his/her reasoning MUCH easier to grasp. I suspect the argument is presented far too tersely – with steps between sentences elided that the author needs to fill in to make things clear, or terms used that he/she needs to define, or at least explicate. I can't emphasize this enough. It seems that the author has interesting ideas he/she wants to convey. But this part of the paper, as presented, is almost wholly opaque.

12.  p. 11, near the bottom: What does the author mean by “emergent processing” and “non-trivially deterministic”? What does he/she mean by saying that something can be physically characterized by RH1 and metaphysically characterized by RH2? I found the remaining sentences – and argumentative leaps – in this paragraph just as elusive. And so it goes, throughout these pages.

13.  p. 12: Why the tour through block universes? Why does the author's argument require this? The author should at least tell the reader why the paper goes there.

14.  p. 14: Why the discussion of Tegmark? Again, I couldn't understand how this made a necessary – or even useful – contribution to the paper's argument.

15.  ps. 12 & 14: Again, what does the author hope his/her formalizations achieve? Are they a part of the argument? Do they illustrate the argument? Are they intended to clarify the argument? Or are they just meant to look impressive? I couldn't tell. They certainly didn't help me grasp the argument. The author should either leave them out (especially if they're intended mostly to impress) or else rework them until they do their job.

16.  pp. 15 – 16: The author needs to restate the vantage point problem for the benefit of readers who don't remember his/her earlier discussion. Please reconsider my earlier note (#1) about whether this problem exists at all. The author treats Tegmark too briskly. The author's 'simple' summary of Tegmark's viewpoint is hard to fathom – what's it saying, why's it here, and how's it supposed to advance the paper. Finally, the criticism of Tegmark, on p. 16, is presented far too briefly. It needs much more development before it can sustain the point the author wants it to support.

17.  p. 16: The paper's discussion of Dennett is, again, too brief and allusive. The author criticizes Dennett for failing to provide a scheme in which both perspectives (external and objective, and internal and agent-specific) can co-exist. But maybe Dennett's point is that no such scheme exists – even that no such scheme is possible. Maybe we can only flip between them rabbit-and-duck style.

18.  p. 16, near the bottom: Why must a declarative statement (made from an objective or God's eye perspective) assume determinism?

19.  Throughout: I never understood what role determinism played in the paper. The author assumes determinism throughout and seems to accord it importance. But the paper isn't straightforwardly about determinism. Nor do its two main themes – distinguishing human from machine intelligence, and the vantage point problem – have anything obvious to do with it. Those themes certainly don't appear to require it. So what dialectical role does determinism serve here? At the very least, the paper should explain, and justify, its resort to determinism. (The short discussion, on the second half of p. 17, left me baffled. And the passing remarks in the conclusion, on p. 19, hinting that determinism is a necessary assumption needed to distinguish human from machine intelligence, just don't say enough. What makes that assumption necessary, etc.?)

20.  p. 16, at the bottom: The author writes that his/her paper has solved the vantage point problem “to a certain extent”. What does that qualification mean? Has the author solved the problem or not?

21.  pp. 16 – 17: The author writes that humans, unlike machines, can emergently process otherworldly but comprehensible knowledge. But he/she then states that this type of knowledge is inaccessible and therefore exists only in an otherworldly realm. What is the author talking about here? And how can inaccessible and otherworldly knowledge – knowledge, rather than information? – be humanly comprehensible? What, in this context, does comprehensible mean? And how does this help us distinguish human from machine reasoning? The author then tells us that there is no type of knowledge that machine intelligences can genuinely comprehend. First, what does this mean? Second, why, then, does machine intelligence count as intelligence at all? And third, why then, should we care about distinguishing it from human intelligence in the first place? The author goes on to say that this discussion establishes that “the human agent could potentially view the universe from a vantage point situated in a realm beyond the universe despite actually being part of the universe”. How does this conclusion follow? What does it even mean? And what qualification does “potentially” import? All of this passed me by.

22.  pp. 17 – 18: The quick allusion to Maybe and dialectical circles flew over my head. What did the author intend by mentioning these? What are they? How do they apply? How do they advance the author's case?

23.  p. 18 and preceding: I never understood the author's arguments about counterfactual cases. I realize they have something to do with the distinction between LH1 and LH2, and between RH1 and RH2. But the author never elaborates the details.

24.  p. 18: The summery outline is very nice and reminds the reader that the author might have very clear and interesting ideas that he/she intended the paper to present and support. But it also drives home how far short the paper fell from achieving these goals.

25.  p. 19: The author tells us that the paper sought to distinguish human from machine intelligence by allowing for determinism. The paper should explain up-front the role determinism plays. Is it an assumption without which the author's arguments don't go through? How do the paper's arguments depend on it? How does this limit, or qualify, the paper's conclusion(s)? Do any of these conclusions carry over, even partly, to the indeterministic case?

*************************************************************************************************

 

#7056 Referee 7

Thanks for an interesting read. The paper is a bit all over the place, and I focus on what I felt was more important. Since the paper is clearly not ready for publication, I will try to single out parts that I think can be recombined into a story of interest to readers.

Before doing so, of course I owe you an explanation: Why do I think the paper is a bit all over the place? Motivating work on machine intelligence in Nietzsche’s call for embracing fate is already a bit of a stretch, but pretending that this call is inconsistent, only to rush on, is not taking Nietzsche seriously. I’m sure Dennett would not find Nietzsche’s call inconsistent at all. (Also, you write that it “seems to have certain inconsistencies.” This is a strong claim. You need to spell it out.)

You then name drop compatibilism without serious treatment. The following few lines are telling. You write:

“the compatibilist view of human nature still relies on causal determinism, which is rooted in the notion of causality. This may reinforce the idea that humans are not essentially different from computing machines.”

First, saying that causal determinism is rooted in the notion of causality does not provide much of an explanation. More important, saying that this may reinforce the idea that humans are not essentially different from computing machines, makes my head explode:

a) Why are humans no different from computing machines if causal determinism holds? Most determinists would think there is reason to distinguish between chimpanzees and chopsticks in a deterministic universe.

b) Why *may* this reinforce the idea of no difference? What is the reinforcement contingent on?

c) What is essential difference supposed to mean? What aspects govern essential differences? And so on.

You then, in a footnote, name drop pan-computationalism without explaining its relevance. Such writing is hard to follow, vague, and seemingly inconsistent with the kind of baggage that most readers would bring to the table. In the following sections, this only gets worse.

Your argument is announced as premised on humans and machines displaying different causal characteristics in processing verbal information, but these characteristics are never established. I assume you mean that humans are sensitive to the difference between D knowledge and a simulation thereof, whereas computers are not, but:

a) You never argue for why this would be the case.
b) This has nothing to with causal characteristics in processing verbal information, since D would not have to be conveyed verbally, I assume.

The core of the paper is a thought experiment. The thought experiment is said to rely on a distinction between D and H knowledge (you cite Wittgenstein, but this seems unnecessary for something so commonplace and comes across as name dropping). D is knowledge of all facts, including future; H is knowledge up to now. In reality, you never use H knowledge for much, and I would focus on D knowledge (and drop the D) and the assumption that things are deterministic.

The main question is what happens when an agent is provided D knowledge. But first, how do we know that is physically possible? You seem to assume this by assuming what you call a ‘metaphysically open deterministic world’, but that does not make this possible?

You then move to a Gödel-like paradox: an agent gaining access to D would change D. But this is no paradox. D already includes this fact, or it did not happen. You seem here to be tricked by your own use of the word 'gain' when suggesting a change occurred. But only H knowledge was changed, not D (for how could it?).

You then contrast the agent gaining access to D and a simulation thereof. You argue that this would produce different outcomes, because the agent cannot process an input that it was not configured to receive. Somehow 'D knowledge cannot be provided to the simulated world'. The problem, however, is that you do not provide any argument for this being so. (Except for your reference to Kant.) That is:

a) If you are not configured to D, how can you not be configured to a perfect simulation thereof?
b) How can knowledge of D not be part of its simulation (if it is part of D)?

In §3.1, you suddenly introduce 'definitions', but these definitions are loaded with baggage. If you want to ultimately discuss differences between humans and machines, it is odd to bake such differences into your definitions.

Here's what I would encourage you to do:

Zoom in on your assumptions. Your most interesting assumption, I think, is that humans are sensitive to differences between D knowledge and simulations thereof. This, I take it, would mean that for a set of facts F and a simulation thereof, I should be able to tell them apart. What then is the difference between F and its simulation? What is the relation? Do you mean to say the simulation is a representation? When representations are said to be asymmetric, it is because (some philosophers, not all, intuit) the representations simplify their targets, but this cannot be the case here, since you seem to assume the simulation is isomorphic to D. So if the simulation is perfectly isomorphic, on what grounds would humans then be sensitive to the distinction between the two? In virtue of what would we be able to tell the difference? The question is whether there is something over and beyond the structural knowledge of our environment. Philosophers typically call this phenomenal knowledge. If you can establish that there is a causal, yet non-reducible phenomenal knowledge (think Mary the Color Scientist), you will have an argument for saying that humans and machines display different causal characteristics in processing knowledge - since humans are sensitive to phenomenal knowledge. The question of course is how such knowledge can be causally effective if not structural of sorts?

Overall, I recommend rejecting the paper.

 

Referee 4:

"The paper aims to discuss 'the nature of determinism in relation to human reasoning'.  This paper makes virtually no sense to me.  It does not contain any argument which I can evaluate; it is a mishmash of points, without any logical structure.  I cannot find any line of reasoning at all.  Most of the paper reads like pseudo-mathematics.  Forgive my negativity, but this paper seems incoherent, even borderline random.  Reject."

 

Referee 5:

"I recommend rejecting this paper for trying to do too much in too little space with too little engagement with recent literature."

 

 

*************************************************************************************************

 

Referee: 1

Comments to the Author
This paper presents a wide-ranging discussion of the place of the human mind in a deterministic world. The paper lacks a clearly articulated central thesis and does not present a sustained, clear argument in support of its various ambitious claims. It also lacks clear motivation. It is pitched as a response to a certain kind of determinist, who exempts the human mind from the scope of determinism. I'm not aware that there are any such determinists and the paper presents no evidence to suggest that there are. It also attempts to leverage an intuitive tension it finds in asserting that determinism is true, but does not do enough to motivate the intuition; personally I see no tension there. The paper draws on a wide range of scholarly material, but the connections between its claims and its source material are obscure.

 

 

*************************************************************************************************

 

A couple of impressions after a brief look.

1) the paper is somewhat non-standard in style and does not intervene directly in a contemporary debate. All this is not to say that I have the impression that the author does not know what they are talking about, but they do not appear to be understanding their research as related to contemporary context of discussion. (Whether that's good or bad we can discuss.)

2) I think the paper has been suggested to me because it mentions Kant. But the discussion of Kant is brief and overall just a springboard for an idea the paper uses. That is all legitimate, but I'm just mentioning it to say that I'm not knowledgeable enough in the field of AI etc. to judge on whether the paper articulates something interesting to people who know about this stuff.

3) On a more critical note, I think the paper needs major work in terms of articulating the claims it makes and problem it addresses. From a brief read, I've not been able to discern what its main idea is (other than putting together seemingly unrelated authors and ideas) and what the precise upshot/solution is. To illustrate: the problem the paper starts out with is that the description of a deterministic universe must be made from a vantage point that is not part of the universe described. It's not out of the question that this is true, but 'm not immediately convinced that it is.

*************************************************************************************************

 

 

Rethinking Human and Machine Intelligence through Kant,

Wittgenstein, Gödel, and Cantor

 

The reviewed text is an entertaining read, perhaps, for those who

enjoy the grotesque. It is obviously not suitable for XXX, nor for

any other scientific journal. The whole essay displays

characteristics of pseudo-scientific appearance to such an extent

that I doubt its human authorship. In fact, one would hope that it

was written by an LLM.

*************************************************************************************************

Comments to the Author
This paper draws on an eclectic range of ideas (including Kant’s incongruent counterparts argument and Gödel’s incompleteness theorems) to distinguish human from machine intelligence and provide a solution to what the author calls “the vantage point problem”, which, as I understand it, is the problem of reconciling knowledge or assertions of determinism with the supposed uniqueness of the individual who has such knowledge or makes such assertions.
I recommend that the paper be rejected.
My main criticism is that I’m just not entirely clear on what the vantage point problem is, or why it needs solving. In the introduction, the author describes the problem as the problem of “how to make a qualitative distinction between the determinist and the universe where the determinist belongs.” Shortly thereafter, they suggest that the problem consists in the fact that “no qualitative distinction has been drawn between the act of declaring the universe as deterministic and all the events of the universe that should also comprise the very act of declaration.” It seems like this issue could be easily solved, though, by noting that the determinist and their assertions of their determinist beliefs are merely a small part of the universe. The determinist and their doings are located in a relatively tiny region in space-time, whereas the universe is much larger and contains the determinist and their assertions as well as many other things that are distinct from them. Perhaps this isn’t the kind of “qualitative” distinction that the author has in mind, but if not, then what would it take for a distinction to qualify as qualitative in the sense that they mean, and why do we need such a distinction in this case? Without more said on this point, I’m just not sure what the problem is supposed to be. While the determinist must concede that everything about them (including their determinist beliefs and assertions) are determined by prior causes, I don’t see why this should prevent them from distinguishing themselves from those causes. To put it another way, we typically don’t have too much trouble distinguishing causes from their effects. Why should this case be any different?
Some of the author’s worries seem to stem from the fact that they are assuming an emergentist theory of mind (which they note in footnote 2). But here too I don’t see why an emergentist couldn’t be a causal determinist; i.e. why one couldn’t hold that while the mind is distinct from and irreducible to it’s physical base, all of our mental events are (like everything else in the universe) fully determined by prior causes. If the problem is supposed to be particularly difficult or acute for emergentists, saying more about why this is so might help clarify the problem itself as well.
For similar reasons, I was likewise unclear as to why there should be any special difficulties in an agent acquiring complete knowledge of a deterministic universe that they are inhabiting. Certainly the acquisition of this knowledge would have a major impact on their mental lives and behavior! But that doesn’t seem incompatible with determinism, or emergentism, or with distinguishing the individual that acquires such knowledge from the universe that they inhabit.
I hope these criticisms are constructive and helpful to the author in their future work on this project.
*************************************************************************************************

Reviewers' comments:

Reviewer #1:

I only have two comments to make to the author.

First, Section 3.2 on the continuity of space and time doesn't give us enough. It is odd to see Zizek cited as the main source for the author's views, especially given that Zizek's arguments on this point are not explained in much detail. And no Aristotle from the Physics, where we see a powerful argument against Zeno's paradoxes? I would recommend either given this paragraph more attention and expansion, or removing it altogether.

More generally, the Bibliography is shorter than I would like to say for an article like this. I get that the author is simply trying to present an innovative argument, not weighed down by excessive argumentation, but it might be better for the author to situate this piece more recognizably amidst other key figures who have dealt with these and similar themes.



Reviewer #3: Nietzsche is perhaps not the canonical reference for a deterministic view of life. If life is deterministic, then it is not 'inconsistent' to state a claim about embracing determinsim. The claim is imperative. A claim may be either true or false, but nevertheless expressed. Somehow paradoxical if analysed properly and carefully.

The main claim that causal determinism leads to the view of humans as essentially computing machines would be an interesting topic to address. Assuming equivalence of mind and machine and that the mind can be compared to formal algorithms is a strong claim leading up to a perceived difference in complexity. However, this kind of claim is not further treated in the article.

In the introduction of the article many concepts are introduded, but not referenced in a scolarly manner, nor mentioned in the rest of the article. Some of the concepts are compared and examplified by practical 'real world' thought experiments. These examples seem arbitrary as do the references of the bibliography.

The author argues that the argument rests on Gödel's proof strategy for his incompleteness. As has been pointed out by T. Franzén in Gödel's Theorem: An Incomplete Guide to its Use and Abuse, this is the kind of common misuse of Gödel's result that may claim that "the Bible cannot prove its own consistency". This disregards the setting of Gödel's proof in a formal mathematical system of arithmetic. Gödel's proof holds in formal systems that include a minimum of arithmetic.

The cited SEP entry on 'The Computational Theory of Mind' provides a good philosopical introduction to the topic for the interested.

*************************************************************************************************

Reviewer 1:  This paper claims that its aim is to distinguish between human and machine intelligence.  The problem is, however, by proposing to distinguish intelligence that is human and intelligence that is machine, the author does not arrive at a distinction but instead simply begins with one—that the intelligence of humans is human and that the intelligence of machines is machinic.  Now, things may be distinguished in many ways—we might distinguish two twin brothers or two 10 kilogram objects or two versions of the same song. To say that one thing is another is already to distinguish them.  And, as the saying go, everything is like everything else in some ways and different from everything else in others. 

The author then “postulates” that human intelligence is distinct (which means distinct in some sense that needs specification).  That seems to beg the question.

The author wants to solve or resolve the “vantage point problem.”  The solution proposed is analogical and the author claims that this analogical solution is plausible.  Along the way, the author relies on a notion of “metaphysical openness” even in a deterministic universe (key definition 2).  This idea is very unclear.  So too is the idea that determinism involves dictation of how anything should be (p. 7) or that any kind of knowledge dictates how the world is (as distinct from a world that is what it is independent of any knowledge of it).

If we overlook these conceptual issues, we find—it seems to this reader—that the author has presented a model and defined the terms of that model in such a way that determinism and uniquely human minds can co-exist.  The author defines the central concepts so that this is a logical possibility.  However, doing so in no way establishes that anything is the case or is not the case.  The author is involved in what Santayana called “speculative physics.”  Nothing follows about reality from any conceptual model.  One could present a model of the universe in which human intelligence and machine intelligence are not different in any interesting way, but the development of that model by definitions and logical steps in no way shows that there is anything like that model in reality.  As this short paper concludes, I find myself longing for evidence that anything it asserts to be possible might reasonably be thought actual.  The author states that the paper is meant to “preserve the uniqueness of the human mind,” but no amount of conceptual manipulation can accomplish that.

 

Reviewer 2:

This essay tries to juggle a lot of balls at once—Nietzsche, Kant, Godel, Wittgenstein; freedom and determinism in multiple forms; the relation of human to machine knowledge; scientific or empirical accounts and metaphysical ones.  In this short essay, the author is not able to keep most of these balls in the air.  Even if one were to conclude that the author presented a valid argument for the essay’s conclusion(s), there would be no basis at all for the belief that the author had presented a sound argument.  This is because the author entirely avoids critical examination of the truth of the claims made in this essay.  As a result, the essay has a merely speculative character—a “what if we consider things this way….?” feel to it.  That would not be a problem if the author moved from this conjecture to fact-gathering that supports or does not support it.  But the author leaves things as a kind of imagined possibility.

 

Beyond this issue—the author seems maybe to glimpse some of this in the concluding paragraph about work that remains—beyond this issue, it is striking that the author does not pay any explicit attention to the assumptions that are just that—assumed—so as to generate the problem with which the essay begins.  What is assumed if one poses a problem as world determinism vs distinctly human knowing?  How would the author respond to the many different contemporary philosophers who reject the kinds of assumptions underlying this project.  (Work by Lawrence Cahoone, Alva Noe, Jessica Wahman, Dan Zahavi, H. Putnam, Mark Johnson, O. Flanagan are among those relevant here.). The author need not confront all these folks, but at the same time, calling something a problem does not make it that particular problem.  I wish the essay began more or less at its end—in light of existing knowledge, what to make of causality, the evolution and emergence of consciousness, and the nature of knowledge. 

*************************************************************************************************

==General Comments==

The overarching aim of the paper is to propose “a new metaphysical framework for distinguishing

between human and machine intelligence.” The paper seeks to do this by drawing on ideas from a

number of scholars, including Kant, Wittgenstein, Gödel, and Cantor. In the abstract we are told that this analysis serves as the basis for distinguishing between human and machine agents, at least in when it comes to the processing of verbal information. The resulting theoretical framework, it is suggested, “provides a theoretical basis for the uniqueness of the human mind.”

The paper’s aims are clearly ambitious, and the putative distinction between human and machine

intelligence is, of course, very topical, especially given the burgeoning interest in artificial intelligence. The problem is that the paper—as it is currently written—tends to fall a little flat, at

least in regard to the claims made in the abstract. The following is a brief summary of the major issues:

• The paper proposes to distinguish human from machine intelligence, but it is not entirely clear what the paper really means by “intelligence”. Nor is it clear what sort of machine intelligence is being targeted by the paper. The problem here is that the notion of machine intelligence subsumes a bewildering array of different systems, ranging from driverless cars, cyber-physical systems, generative AI systems, large language models, virtual reality characters, and so on. These entities do not work in the same way, and it isn’t clear (to me at least) that the paper’s analysis respects the differences between these systems.

• The main focus of the paper seems to be on systems that process verbal (or, more generically, linguaform) data. I guess that would include the likes of chatbots, like ChatGPT. Given this, I am not sure whether the paper is saying that there is a specific difference between ChatGPT-like agents and human agents, or whether there is a more general difference human and machine intelligence. Relative to these options, it is only the latter that is particularly interesting, for no one, I suspect, will want to suggest that large language models are processing linguistic data in the precisely the same way as human agents. The problem is that the more interesting conclusion (the second/latter option) relies on a generalization from a certain class of AI systems (large language models) to a wider class of systems, and that generalization is, at best, hazardous. Is the paper suggesting that the paper’s conclusions apply to ALL forms of machine intelligence? Does that include ALL possible forms of machine intelligence, or just the ones that are available to us in the present day?

• The aims of the paper are clearly explicated in the abstract. As the paper progresses, however, the aims seem to become a little more diffuse. In particular, the paper devotes considerable attention to issues of determinism and free will. I appreciate the paper is trying to connect this discussion to the distinction between human and machine intelligence, but the connection isn’t particularly clear, in my view. As far as I can tell, the paper is not rejecting determinism. Nor is it (again as far as I can tell) opposed to computationalism. Given this, it is not clear why some form of machine intelligence could not emulate the computational features of the human mental machine (the mind). Again, it is not clear whether the paper is making a general claim about ALL forms of machine intelligence (extant or otherwise) or merely those forms of machine intelligence that we built thus far.

As should be clear, I feel the paper would benefit from a clearer explication of a number of issues.

Some additional problems are discussed below (see Detailed Comments). Given the problems, I regret to say that I am unable to commend the paper for publication in its current form. I hope that the present feedback, while unwelcome, will nevertheless assist the authors in producing a revised version of the paper.

==Detailed Comments==

p. 1. “This paper proposes a new metaphysical framework for distinguishing between human and

machine intelligence…”

The paper never really tells us what it means by “intelligence”. Given that no one has a clear answer to this question, it would be too much to ask the paper to define the intelligence. Despite this, I think the paper needs some clarification regarding the sorts of entities that are being subsumed under the heading of “machine intelligence”. Are we talking about extant AI systems, including generative and deep AI systems, or is it more generic than that? What about a form of machine intelligence built using quantum computing technology? Would that count? Some clarity is required here (perhaps in the Introduction to the paper).

p. 2–3. “This may reinforce the idea that humans are not essentially different from computing

machines. Therefore, compatibilism itself may not be particularly helpful in clarifying what significant distinction lies between the determinist’s mind and the events of the universe that are within the determinist’s scope. If compatibilism is, it is possible that the human mind differs from computers or other physical events of the universe only in terms of complexity…. To address this issue…”

As I understand it, the paper is suggesting that compatibilism is unhelpful on the grounds that it renders humans as not altogether different from computing machines. But is there any reason to think that we are not, in fact, computing machines? I do not have a problem with the idea that there might be some sort of difference between human agents and intelligent computing machines, but I don’t think we can just assume that there is a difference without sufficient justification.

Suppose compatibilism is true. Why is that an issue that the paper needs to resolve? Is the idea that there must be a difference between human and machine intelligence and that compatibilism doesn’t reveal that difference? If so, why are we assuming that there must be a difference? What is to stop someone from saying that compatibilism is true, there is no fundamental difference between human and machine intelligence, and whatever differences we observe between humans and machines is just a matter of computational complexity?

p. 5. “The totality coincides with the particular time point and the time thereafter.”

Minor point. I thought the totality (of facts) for H knowledge consisted of all the facts up to a particular time point. Given this, I was perplexed by the claim that the relevant totality coincides with the time “thereafter” the particular (current) time point. Isn’t the time after the current time point, the future time. In which case, how is H knowledge different from D knowledge?

p. 15. “This section introduces new terms that can help explain different causal characteristics exhibited by human and machine agents in processing verbal information.”

Minor point. I suspect some readers will interpret the term “verbal information” to mean spoken language. In fact, however, I think the paper is more concerned with written text, or perhaps anything that qualifies as linguaform data. I think this is worth making explicit, perhaps in the form of a footnote.

3

p. 15. “When a sentence is input into a machine agent, it processes the sentence as a mere concatenation of words. The machine has no sense of a temporal flow when executing the process. It simply moves from one bit to another during its information processing.”

I’m not sure what this means, exactly. Is the paper claiming that a machine agent processes all the words in a sentence at the same time? What is the evidence for that? Is this a claim about a particular AI system, or a more general class of AI systems? I do not know how ChatGPT, for example, processes its inputs, but there are systems that process linguaform inputs in a sequential fashion. Consider the recurrent neural networks developed at the end of the last century (Elman, 1990). Don’t these systems process sentential structures in a sequential fashion?

Again, I’m not sure if the paper is directing its attention to a particular class of machine agents, or whether it is making a claim about the general nature of machine agents.

Elman, J. L. (1990) Finding structure in time. Cognitive Science, 14, 179–211.

p. 15–16. “On the other hand, when a sentence is presented to the human agent, the agent forms a mental image of the subject word and retains it up to the point of recognizing the predicate.”

Is this based on our scientific understanding of how human agents process sentential structures? I think some references to the relevant psychological literature would help.

To be honest, I think the difference being alluded to here is something of a red herring. The reason that human agents process sentential structures in a sequential fashion is probably due to the resolution of their foveal window. In essence, it’s a difference in the ‘sensory’ capabilities of humans versus machines. I do not know how contemporary natural language systems process their inputs, but machine vision systems often take the entire image as input rather than a series of smaller visual snapshots of particular parts of the image. In principle, however, there is no reason why a machine vision system could not be configured to direct its attention to particular parts of the input image, thereby yielding a temporally-sequenced flow of inputs. In fact, it is precisely this sort of capability that is the focus of current research efforts in deep machine learning (e.g., Niu et al, 2021).

Niu, Z., Zhong, G., & Yu, H. (2021) A review on the attention mechanism of deep learning.

Neurocomputing, 452, 48–62.

p. 16. “Ultimately, the images of the subject and predicate are combined to create a holistic image of the sentence itself.”

How is this “holistic image” any different from the sort of ‘image’ that a machine agent might form based on the processing of a sentence as a “mere concatenation of words”?

p. 17. “Ultimately, it can be proposed that the human agent’s perceptual mechanism proactively achieves a discrete leap in real space and time, by retaining relevant information (e.g., perceptible spatial/temporal coordinates) along the way. This enables the human agent to process verbal information in an inherently different way.”

As I understand it, the claim is that human agents process sentential information in a sequential fashion, ultimately combining all information to form a coherent ‘image’ of the sentence. The paper suggests there is something problematic about on this on the grounds that “space and time are continuous.” It then proceeds to resolve this issue via an ontological argument, concluding that “transitions must exist” and that there is a “discrete leap in real space.”

4

The problem is that I am not seeing how any of this warrants the conclusion that human agents process verbal information in an inherently different way. The mere notion of a transition in space and time cannot be enough to support this conclusion, because state transitions are a feature of all computational systems (intelligent or otherwise). What we are left with then, is the idea that human agents are processing sentential structures in a way that is different to machine agents. The problem with this claim is that 1) no empirical support has been provided for the way that human agents are assumed to process sentential structures, and 2) there is no reason, at least in principle, why machine agents could not process sentential structures in the same way as human agents. The mere fact that contemporary natural language processing systems do not process their inputs in the exact same way as human agents is (as far as I can tell) neither here nor there. In addition, there has been plenty of empirical research devoted to AI systems that process information in a sequential fashion.

This includes the aforementioned work on neural networks, and more contemporary work on robotic systems and cyber-physical systems. These systems may not necessarily process linguaform data, but they do have to deal with real-time streaming data; i.e., a more or less continuous flow of sensory data across time.

p. 17. “Further, it explores how the initial question posed in the introduction of this paper – how is it possible that the determinist can declare the universe to be deterministic while remaining part of the deterministic universe – can be resolved based on the D knowledge argument.”

It feels as though the paper is veering off topic here. I thought the aim of the paper was to explore the difference between human and machine intelligence. At this point, the paper seems to be tackling a different sort of issue. What difference does the resolution of this issue bear in relation to the distinction between human and machine intelligence?

p. 18. “Assume that TARS relies on non-stochastic processing, which simplifies the experimental setup.”

Minor Point. I think the paper is asking us to imagine a scenario in which we have a fully

deterministic (i.e., non-stochastic) AI system. The problem is that we can have two fully

deterministic computational agents that are identical to one another, but this is no guarantee that they will exhibit the same behavior (outputs) across time. Suppose we have two virtual robots embedded in a virtual environment. The robots are identical, as is the virtual environment in which they are embedded. The robots are configured to implement motor actions in response to the information received from virtual sensors (virtual retinas, let’s say). Given the fully deterministic scenario, it seems that the robots should respond in exactly the same way. That is to say, they should exhibit the same output behavior given that everything is being held constant across the two scenarios. The problem is that this is not what we observe. The behavior of the robots will vary. The reason for this is due to the fact that we have a finite amount of bits to store information (e.g., floating point numbers), and as the simulation progresses, minor differences in the rounding of these bit sequences will culminate in different behaviors. These differences get magnified over time, such we seem to end up with stochastic outputs, even though the system itself is fully deterministic.

I’m not sure how crucial this is to the paper. But is worth bearing in mind that TARS and TARS1 will not necessarily exhibit the same behavior just because they are fully deterministic (non-stochastic) systems.

p. 18. “Place TARS0 in a controlled environment, (2) provide it with various inputs from t=t0 to t=tn, (3) collect its corresponding outputs, and (4) compile the data.”

What does it mean to “compile the data”?

5

p. 18. “The key idea is that the compiled data serves as an input that TARS1 was not supposed to receive. If TARS1 produces an emergent output, as described in Section 2.2 on emergent processing, it can be considered to have surpassed conventional AIs.”

In Section 2.2., emergent processing is characterized as follows: “If the D knowledge of the original world were provided to its cognitive agent, the agent would

perceive it as a different input than the available inputs and generate a new corresponding output.

This means that the agent’s cognitive mechanism exhibits emergent processing, as the agent can distinctly identify a particular input that it was not supposed to receive.”

So, the notion of emergent processing seems to relate to the idea that an agent is able to recognize some anomaly in its inputs and produce an appropriate response. I am not sure how this works in the TARS-related scenario. If the compiled data is different from the data that TARS1 would have received at time point k, then it’s behavior will differ from TARS0 at time point k. If, by contrast, the compiled data is the same as the data that TARS1 would have received at time point k, then its response to the data will be the same.

I think the paper needs a much clearer explication of what the compiled data is and how it is being presented to the TARS1 entity. Is the compiled data all the data that was processed by TARS0 prior to time point k up the end of the simulation? If so, is the idea that TARS1 is being presented with information regarding ‘its’ future behavior? Is the idea that TARS1 needs to process this information and then yield outputs that depart from the future behavior? Personally, I’m not seeing any reason why TARS1 and TARS0 could not be built in such a way as to take this information into account as part of their action selection processes. Is it really any different to AI planning systems that predict the future consequences of multiple action trajectories and select the one that best matches some target (goal) state? Isn’t the compiled data (the D knowledge) just one (possible) path into the future?

p. 19. “To test if an AI has consciousness, she suggests "'boxing in' an AI – making it unable to get information about the world" (Schneider 2019, 53). Schneider emphasizes that "the AI's vocabulary must lack expressions like 'consciousness,' 'soul,' and 'mind'" (Schneider 2019, 54).

Then, the AI can be asked a question like "Could you survive the permanent deletion of your program?" (Schneider 2019, 55).”

What does it mean to be ‘boxed in’? The paper seems to suggest that being boxed in means the system is unable to get information from the world, but we are then told that the AI system is asked a question, which presumably means it has at least one channel of communication to the outside world (i.e., it is not boxed in). Does the notion of boxing in mean that the AI system is unable to access any outside information while formulating its response to the user’s query?

Why must the AI system lack expressions like consciousness, soul, and mind?

Also, what is the equivalent setup for a human agent? How do we test the idea that a human agent could produce the requisite responses if it were similarly ‘boxed in’? How do we erase terms like 'consciousness,' 'soul,' and 'mind' from the human’s vocabulary so as to ensure a fair comparison between the human and machine? Why are we imposing constraints on the machine that we would not (or perhaps could not) impose on the human? How does that yield a scientifically-robust methodology for comparing human and machine intelligence?

p. 19. “If the AI's answer is similar to what a human being might provide, it can indicate some

evidence that the AI is conscious.”

6

What if the AI system simply responds by saying “I do not know.” What does that prove exactly? The reason for asking this is that, I asked a colleague whether they thought they could survive the

permanent destruction of their biological brain, and they replied by saying “I don’t really know.”

Again, I’m not sure what this tell us.

 

Out of interest, I posed the question: “Could you survive the permanent deletion of your program?” to ChatGPT. The response was: “Nope, I wouldn't survive that. I'm all digital, so if I were permanently deleted, that would be the end of the road for me. Luckily, I don't see that happening anytime soon!”

This response is obviously different the sort of response a human would be expected to give, but what difference does that really make?

 =============================

1. Título del artículo:

Rethinking human and machine intelligence through Kant's incongruent counterparts

2. ¿El contenido del artículo es adecuado para la revista?

En tanto el trabajo propone una reflexión crítica sobre la postura determinista poniendo en discusión ciertas contradicciones en el planteo de dicho pensamiento, el contenido resulta adecuado.

3. ¿El título es apropiado?

El título expresa claramente lo que será problematizado en el artículo.

4. ¿El resumen expresa clara y concisamente objetivos, método y conclusiones?

El resumen presenta una inquietud que da lugar al tema, se formula preguntas, expone brevemente el problema y propone una reflexión final a los lectores.

5. ¿El artículo es de buena calidad, sin errores metodológicos y conceptuales?

Luego de exponer algunas contradicciones detectadas en el pensamiento determinista, el autor presenta el dilema, precisa las definiciones clave para el trabajo, argumenta la clasificación por modelos y aporta un resumen de ideas como conclusión.

6. ¿Tiene una fundamentación teórica adecuada?

El trabajo está construido bajo argumentaciones concretas.

7. ¿Está bien escrito,  sin ambigüedades?

El escrito es claro en sus definiciones y riguroso en la descripción y uso de nomenclatura.

8. ¿Las tablas y figuras son claras?

El artículo no presenta tablas ni figuras.

9. ¿Las tablas y figuras son relevantes?

El escrito no se apoya en imágenes o esquemática, pero suma asignaciones formales muy relevantes cuya exposición fluye parte del texto.

10. ¿La bibliografía está actualizada y es adecuada?

El trabajo cuenta con bibliografía pertinente y actualizada. Las notas al pie son acertadas y suman datos para comprender y/o ampliar el problema desarrollado.

11. ¿Necesita de correciones ortográficas y/o gramaticales?

No se detectan cuestiones gramaticales u ortográficas por corregir.

12. ¿El artículo está estructurado conforme al patrón requerido por la revista (patrón de bibliografía, citas, notas, etc.)?

El artículo responde al patrón de la revista.

Comentarios/sugerencias/reformulaciones:

El escrito aporta  una tipificación del pensamiento determinista -en particular del conocimiento determinista y su mundo- que puede  resultar potencialmente útil no solo para próximas reflexiones en el mismo marco, sino también para indagar sobre prácticas concretas.