Review Article

ISSN: 2574 -1241 DOI: 10.26717/BJSTR.2025.60.009461

Biologically Inspired Cognitive Architectures and

Generative Pre-Trained Transformers are Friends
or Foes?
Emanuel Diamant*
1
Independent Research Engineer, Israel
*Corresponding author: Emanuel Diamant, Independent Research Engineer, Israel

ARTICLE INFO ABSTRACT

Received: January 24, 2025 BICA*AI is a well-established long-lasting R&D enterprise aimed at creating computational architectures intend-
ed to emulate Human-level Artificial Intelligence. Recently and quite unexpectedly in its field has appeared an-
Published: February 06, 2025 other contender – a GPT-based AI tool designed to mimic man-computer conversation in a user-friendly natural
human language. As its designers claim, the device exhibits signs of General AI. After an exciting and joyful
Citation: Emanuel Diamant. Biologi- reception, it became clear that the new competitor does not fulfill its expected promises – it returns wrong
and misleading responses, deceptions, and disinformation. The issue raised a wave of public objections and a
cally Inspired Cognitive Architectures request to stop and prevent further device deployment. On the other hand, the device designers claim that the
and Generative Pre-Trained Trans- imperfections are temporary, and very soon the product will rich its avowed qualities. No, this will never happen!
formers are Friends or Foes?. Biomed The purpose of this paper is to explain that the current approach to GPT-based AI tools design is initially flawed,
wrong, and unsuitable because it ignores the basic definitions of Intelligence and Information pro-cessing. The
J Sci & Tech Res 60(3)-2025. BJSTR. paper joins the general awareness that unrestricted and free dissemi-nation of wrongly designed GPT-based AI
MS.ID.009461. tools poses a threat to human society, similar to the threat of careless biological weapon research.
Keywords: BICA*AI; GPT-Based Tools; General AI

Introduction ous vision tasks (image classification, object detection, image gener-
ation, and video processing), audio-related applications (speech rec-
BICA*AI – which stays for Biologically-Inspired Cognitive Archi-
ognition, speech synthesis, speech enhancement, music generation),
tectures for Artificial Intelligence – is a well-established long-lasting
and various multimodal scenarios (visual question answering, visual
(since 2005) R&D enterprise aimed at creating computational archi-
commonsense reasoning, caption generation, speech-to-text transla-
tectures of human intelligence, [1]. “The ultimate goal of research
tion, and text-to-image generation), [4].
in cognitive architectures is to model the human mind, eventually
enabling us to build human-level artificial intelligence. To this end, The prime GPT model, developed by OpenAI, was consequently
cognitive architectures attempt to provide evidence of what partic- enhanced by more and more sophisticated versions, such as GPT-2 in
ular mechanisms succeed in producing intelligent behavior and thus 2019 and GPT-3 in 2020. In December 2022, OpenAI launched a free
contribute to cognitive science”, [2]. Generative AI or GPT – which preview of the ChatGPT model, a new AI tool based on the GPT-3.5
stays for Generative Pre-trained Transformer – is a recently emerged model version. On March 14, 2023, OpenAI released GPT-4, both as
AI tool and application, designed to mimic a computer conversation an API and as a feature of ChatGPT, [5]. According to OpenAI, the pre-
with a user in natural language and simulate the way a human would view received over a million signups within the first five days. As
behave as a conversational partner. [3]. GPT models are trained on an anonymous source revealed, (cited by Reuters in December 2022),
massive amounts of internet text data and are capable of generating OpenAI is projecting $200 mil-lion revenue in 2023 and $1 billion
responses that closely resembled human writing. Originally designed revenue in 2024, [5]. Inspired by OpenAI’s success, other GPT devel-
for purposes of NLP tasks, transformers have been adapted for vari- opers have launched their own products: Amazon, Google, Microsoft,

Copyright@ : Emanuel Diamant | Biomed J Sci & Tech Res | BJSTR.MS.ID.009461. 52662
Volume 60- Issue 3 DOI: 10.26717/BJSTR.2025.60.009461

Baidu, GitHub, Meta, Apple, IBM, Nvidia, and others have created their Wozniak, have signed this letter asking all engineers to stop immedi-
own GPT-based products and put them to market examination, [6]. ate AI development outside of GPT-4 for 6 months due to fear of “los-
From this short exposition of BICA versus GPT-4 achievements, it is ing control” [13]. In addition, the Future of Life Institute states that
perfectly clear that GPT-based products have a far more privileged po- “development moratoriums should be public and verifiable” and “if
sition in the eyes of Human-level AI tools developers and users. How- the decision to suspend AI development cannot be taken immediate-
ever, upon a closer inspection, it turns out that the position of GPT- ly, governments should intervene with AI development authorities.”
based AI tools is not so extremely great as it looks initially. Despite the [13].
generally recognized success in creating artificial devices that mimic
In response to this call, on May 17, 2023, a Subcommittee of the
the human ability to communicate through oral or written conversa-
U.S. Senate Judiciary Committee held a hearing titled “AI Oversight:
tion, GPT-based devices have proven to be a potential danger to their
Rules for Artificial Intelligence”[14]. Similar measures of administra-
users, [7-9]. According to many publications, GPT is a large language
tive control and administrative order have been adopted by the par-
model that uses deep learning to create human-like text.
liamentary commissions of the European Union, Great Britain, Japan,
Or, more certainly, it is a computing system designed to generate and other countries (A more detailed report on the actions of many
sequences of words, code, or other data, starting from an initial input countries in this direction can be found in [15]). At the same time, it
called the prompt, [10] When determining the sequence of words, remains as a problem that these measures of administrative control
it simply predicts the next word based on statistical correlations in ignore the scientific and technical aspects of the problem of creating
the training data and prompts, [9]. Since the system sees only formal, Human level Artificial Intelligence systems. Parliamentary commis-
statistical patterns between words, it does not understand the mean- sions do not deal with these issues, rightly claiming that this is not
ing of words and does not understand their semantics. Therefore, it within their competence. At the same time, the organizations involved
is also called the “stochastic parrot”, randomly stitching together in the creation of these systems confidently declare that they have
words without reference or meaning, [8]. All other negative features already reached the level of General Intelligence. This is exactly
of the system subsequently follow from this - the model inherits bi- what Microsoft claims in its recent publication: “it could reasonably
ases and errors in the training data, the model is very sensitive to the be viewed as an early (yet still incomplete) version of an Artificial
design and formulation of prompts. [11], her answers are inexplicable General Intelligence (AGI) system”, [16] Sorry, but this is not true.
to users, her results are not verified or validated, and users cannot The uncontrolled and unsupervised use of systems whose creators
determine the validity of the model’s output, or whether the model claim that they have already reached the level of General Intelligence
is simply a “fiction”, [9]. On the other hand, the model produces very can only be incomparably more dangerous to society. Because the
coherent, natural-sounding, and human-like responses that users find general public is not competent in these matters. And scientists and
compelling and readily trust them, even if they are inaccurate, [9] At researchers competent in these matters claim that we are still very
the same time, the system is exceptionally fast, very user-friendly, far from the level of General Intelligence. And the main reason for this
responds to natural language prompts and requires little or no user is that today we do not know at all what Intelligence is, and what the
training, references or values, [8]. Text generation at speed, scale, and brain’s cognitive abilities are supposed to be.
ease of use makes the GPT model exceptionally well suit-ed to wide-
1. Christopher Koch: “AGI is ill-defined because we don’t
spread misinformation and deliberate misuse of the system as a tool
know how to define intelligence. Because we don’t understand it…”,
for fraud, as a “weapon of mass deception” [9].
[17].
The main risk of weaponization is related to the potential ability
2. Luciano Floridi: GPT language model “has nothing to do
of GPT-3 to dramatically increase influential production, which is like-
with intelligence, conscious-ness, semantics, relevance, and hu-
ly to be human, [11]. From the above, it becomes clear that GPT sys-
man experience and mindfulness more generally”[10]. I deliberately
tems can pose a serious danger to society and humanity. Realizing this
provide here the quotes from Christopher Koch and Luciano Floridi –
and seeking concrete action to eliminate the dangers and harms that
Christof Koch is a chief scientist of the Mindscape Program at Seattle’s
threaten society, several leading AI researchers, as well as a number
Allen Institute. He has a background in both AI and neuroscience, he
of people who work in companies participating in the AI race, signed
is the author of three books on consciousness as well as hundreds
an open letter calling for a six-month pause in the training of AI sys-
of articles on the subject, including features for IEEE Spectrum and
tems, more powerful than GPT-4, [12]. The letter, published on March
Scientific American, [17]. Luciano Floridi is a professor of philosophy
30, 2023, states that “In recent months, AI labs have been stuck in an
and ethics of information at the University of Oxford, best known for
out-of-control race to develop and deploy ever more powerful digi-
his work on the philosophy of information, and information ethics.
tal minds” [13]. “AI systems with human-competitive intelligence can
According to Scopus, he was the most-cited living philosopher in the
pose a serious danger to society and humanity, as extensive research
world in 2020, [18].
has shown” [12]. Over 1300 people, including Elon Musk and Steve

Copyright@ : Emanuel Diamant | Biomed J Sci & Tech Res | BJSTR.MS.ID.009461. 52663
Volume 60- Issue 3 DOI: 10.26717/BJSTR.2025.60.009461

Therefore, Again system’s technical parameters. The semantic aspects of the message
are of a paramount importance and thus must be met. In accordance
Christopher Koch: “AGI is ill-defined because we don’t know
with the soul and spirit of these requirements, I have developed my
how to define intelligence. Because we don’t understand it…”, [17].
own definition of information. (Interested readers can look into the
Luciano Floridi: GPT language model “has nothing to do with references [25,28,29]).
intelligence, consciousness, semantics, relevance, and human
My Definition of Information Sounds Today Like This
experience and mindfulness more generally”, [10]. Their opinions
sharply diverge from that of other prominent researchers in the field “Information is a linguistic description of structures observ-
of AI, the so-called “founding fathers” of AI – Yann LeCun [19], Geof- able in a given data set.” In a data set, the data elements are not
frey Hinton [20], Rodney Brooks [21], and others, who are concerned distributed randomly, but due to the similarity of their physical pa-
mostly about the social dangers associated with uncontrolled use of rameters, are naturally grouped into some kind of clusters or cliques.
GPT-based AI tools. In their most recent publications, not even a word I propose to call these clusters primary or physical data structures. In
about the technical and scientific aspects of the issue is provided. I the eyes of an external observer, these primary data structures are
deliberately provide here the quotes from Christopher Koch and Lu- arranged into larger and more complex agglomerations, which I pro-
ciano Floridi, because their views unconditionally support my own pose to call secondary data structures. These secondary structures
views on the subject (and of the BICA*AI R&D activities). During the reflect the observer’s view of the grouping of primary data structures,
past years, I have repeatedly expressed my viewpoints on the critical- and therefore they could be called meaningful or semantic data
ly important issues of Intelligence and Information, usually omitted in structures. While the formation of primary (physical) data struc-
today’s AI research studies. Interested readers could find some rele- tures is determined by the objective (natural, physical) proper-
vant papers in the Reference list – [22-27]. But for consistency of our ties of the data, the subsequent formation of secondary (semantic)
discussion, I will provide here a short excerpt of the ideas that were data structures is a subjective process governed by the conventions
exposed once in these publications. and habits of the observer (or a mutual agreement of an observers’
group). As said, the description of the structures observed in the
What is Intelligence? data set should be called “In-formation”. In this regard, it is neces-
There is a widely shared opinion that human intelligence cannot sary to distinguish between two types of information – physical in-
be defined as a single trait or as General Intelligence. Theories of Mul- formation and semantic information. Both are language descriptions;
tiple Intelligences are steadily gaining mainstream attention. Howev- however, physical information can be described using a variety of
er, most frequently, Intelligence is perceived as an umbrella term that languages (recall that mathematics is also a language), and semantic
em-braces (integrates) a multitude of human cognitive abilities (to information can be described only using the observer’s natural lan-
sense, to perceive, to interpret the surrounding environment; to rec- guage. (See [25] for more details). Information processing is carried
ognize, to categorize, to generate plans, to solve problems; to predict out in a hierarchical fashion, where the semantic information of a
future situations, to make decisions and select among alternatives; to lower level is transferred to the next higher level, where it becomes
learn, to memorize, and so on), which altogether produce the effect part of a structure of higher complexity. This agglomeration is carried
of intelligence, [26]. Another commonly shared belief is that human out according to subjective rules fixed in a proto-typical (refer-
cognitive abilities are all a product of human brain activity. Brain ential) structure called the observer’s memory, which is stored in
– as it is generally agreed and accepted – is busy with information the neuron’s body.
pro-cessing. So, we can accept – Intelligence is a product of the brain’s
An important consequence of the above definition of information
information-processing activity. More generally – Intelligence is the
is the understanding that information descriptions always mate-
ability to process information, [27]. In such a form, the definition is
rialize as a set of words, a fragment of text, a narrative. In this
applicable to all domains of natural living beings and to artificial hu-
regard, an important note should be made – in biological systems,
man-made designs as well.
these text sequences are written with nucleotide letters and ami-
And Now: what the Hell is Information? no acid signs. This turns the information into a physical entity, into a
“thing”, with its weight, length and other physical properties. For the
Although the term “Information” is widespread and extensively purposes of our discussion, this is an extremely important remark.
used today, no one actually knows what it means and what it actual- So: The brain is processing information. Neurons are the functional
ly stands for. The concept of “information” was at first introduced by units that perform the du-ty. Despite their discrete structure, neurons
Shannon in 1948. Then other scientists joined the mission – Kolm- are not separate functional units—successful information processing
ogorov, Fisher, Chaitin. However, none of them was ready to define requires close cooperation between work partners. For this reason,
what is “information”. They were busy with the “measure of informa- neurons are connected in a network in which they communicate with
tion”. That was enough to improve the performance and reliability each other, transmitting, exchanging, transferring - in a word – jointly
of technical communication systems. In modern sciences, the needs processing information.
of communication cannot be reduced only to the optimization of the

Copyright@ : Emanuel Diamant | Biomed J Sci & Tech Res | BJSTR.MS.ID.009461. 52664
Volume 60- Issue 3 DOI: 10.26717/BJSTR.2025.60.009461

GPT-based AI Tools in the Light of Information or disapprove generative AI tools’ merits or demerits. I am trying to
Processing stay on the technical side of the story and to analyze GPT AI tools’
technical realities (and myths). In this regard, it is perfectly clear that
As it was just explained above, Intelligence is a product of Infor- the reliance of GPT AI tools on Deep learning (Ma-chine learning)
mation processing, where Information is a linguistic description of training techniques dismisses any claim of its association with Nat-
observable data structures. The bulk of information processing in the ural Language Processing–NLP assumes semantic information pro-
brain is Semantic information processing. In this regard, the reliance cessing, while Deep learning is busy with endless enormous data pro-
of GPT-based implements on Large Language models (LLMs) seems cessing. The “black box” nature of such processing makes its results
natural and justified. (Because Semantic information is a linguistic erroneous and unpredictable. That is the source of most GPT AI tools’
description, a text, a narrative). But a closer look at the LLM’s par- faults and failures. And these flaws cannot be repaired at further stag-
adigms reveals that linguistic components in the training set of the es of processing. They are fundamental, ground-based, built-in. They
LLM models appear as a set of intentionally broken small word chunks are forever. DARPA’s 4-years (2015 – 2019) attempt to create explain-
called tokens, or model parameters. The size of model parameters is able Deep learning tools (explainable AI (XAI) tools) has failed forev-
continuously increasing – GPT-4 is a ~600 billion parameter model er, [30] In this regard, it will be interesting to recall that Ali Rahimi,
(Some people suggest it’s a trillion). Earlier models (GPT-3/3.5) had the 2017 NIPS Award winner, in his award speech, declared that “the
about 185 billion. In such a case you cannot speak about a language current practice in machine learning is akin to “alchemy”, [29]. If
model, it looks more like a statistical data model. As such (and so it Machine learning is alchemy, then what can be said about the concept
is stated in the relevant literature) – “These statistical models need of Intelligence? (Intelligence, as you now know, is a product of infor-
to be trained with large amounts of data to produce relevant results”, mation processing, semantic, not physical, information processing).
[10]. Deep Learning (Machine Learning) mechanisms used for train-
ing the LLMs are also a valuable argument in this regard – Machine However, the designers of GPT-based appliances stubbornly claim
learning principles are applicable only to data patterns mining and that “Intelligence is a multifaceted and elusive concept that has long
discovery. The description of data patterns (structures) is called phys- challenged psychologists, philosophers, and computer scientists.
ical information. Intelligence, as you remember, is busy with semantic There is no generally agreed-upon definition of intelligence, but one
information processing! Therefore, we can conclude that ML neural aspect that is broadly accepted is that intelligence is not limited to a
networks are appropriate only for physical information running, and specific domain or task but rather encompasses a broad range of cog-
not for semantic information processing, which is required for proper nitive skills and abilities” [16]. In the continuation of this, just given,
Intelligence handling. quote, they claim that such an approach to Intelligence is coherent
with the grandfathers’ intentions, expressed in the “Proposal for the
The “black box” nature of Large Language Models, as well as all Dartmouth summer research project on artificial intelligence, August
other designs relying on the Artificial Neural Network paradigm, can- 31, 1955.” (Signed by John McCar-thy, Marvin L Minsky, Nathaniel
not be suddenly overturned, enhanced, and become an Explainable Rochester, and Claude E Shannon). This argument does not seem
Neural Network model. Data-driven physical information descrip- credible. Especially, if we recall that grandfathers’ views of the prob-
tions cannot in a moment be transformed into semantic information lem were very blurred and preliminary. They have failed to assess the
descriptions, that is, into a linguistic description, a piece of text, a lan- complexity of the task (Intelligence definition) and the mutual contra-
guage sample. (That is what GPT-based AI devices claim to be able to dictions between its basic constituents. Recall the story of how Min-
do). Another popular and false claim is that GPT-based AI tools are sky proposed to hire a student to solve the problem of vision during
close to be considered as General AI accomplishments. Again, that is the student’s summer vacations. (The problem even now remains
a popular and widespread misunderstanding. Se-mantic information unsolved) [26]. Recall how Shannon rejected the slightest notion of
processing is done by following the rules saved as a reference proto- semantic information: “These semantic aspects of communication are
type (a memory-saved reference) of the current-level semantic struc- irrelevant to the engineering problem… It is important to emphasize,
ture. These rules, these memories are the observer’s private property. at the start, that we are not concerned with the meaning or the truth
That is, semantic information processing is always subjective. That is, of messages; semantics lies outside the scope of mathematical infor-
the idea of General Intelligence is wrong and cannot be implement- mation theory”, [31].
ed in any natural or artificial construct. Multiple Intelligence (just
mentioned above) being a composition of several different subjective Despite the proud claims of GPT designers that they were lucky
semantic information processing structures can be seen as a valid fea- “to give a preliminary assessment of GPT-4’s intelligence, which is an
ture of the future Human-Level Artificial Intelligence. arduous task given the lack of formal definition for this concept, es-
pecially for artificial systems”, [16]. Any sign of understanding that
Some Concluding Remarks Intelligence is a product of information processing, and that Informa-
The hype around GPT AI tools does not fade. But the purpose of tion, in its turn, is a composition of two, physical and semantic in-
this paper is not to take part in the ongoing discussion, not to glorify formation entities, does not appear in the GPT designers’ statement.

Copyright@ : Emanuel Diamant | Biomed J Sci & Tech Res | BJSTR.MS.ID.009461. 52665
Volume 60- Issue 3 DOI: 10.26717/BJSTR.2025.60.009461

And that is another source of GPT tools failures. Finally, I hope that 15. (2023) Reuters, Which countries are trying to regulate artificial intelli-
the answer to the question posed in the title of this paper is perfectly gence?

clear – GPT-based AI tools do not present any threat or danger to the 16. Sébastien Bubeck, Varun Chandrasekaran, Ronen Eldan, Johannes Gehrke,
AI tools design community. The hype will fade, and BICA*AI will con- Eric Horvitz, et al. (2023) Sparks of Artificial General Intelligence: Early
experiments with GPT-4.
tinue to pursue its goals – to design tools that will allow it to emulate
17. Glenn Zorpette (2023) GPT-4, AGI, and the Hunt for Superintelligence Neu-
Human-level Artificial Intelligence.
ro expert Christof Koch weighs AI progress against its potential threats.

References 18. Luciano Floridi. Wikipedia, the free encyclopedia.

1. Biologically inspired cognitive architectures, From Wikipedia, free ency- 19. Melissa Heikkilä, Will Douglas Heaven (2022) Yann LeCun has a bold new
clopedia. vision for the future of AI. Yann LeCun’s big bet for building intelligent
machines, MIT Technology Review.
2. Iuliia Kotseruba, John K Tsotsos (2020) 40 years of cognitive architec-
tures: core cognitive abilities and practical applications. Artificial Intelli- 20. why is ‘godfather of AI’ Geoffrey Hinton worried? These are the 4 dangers
gence Review 53: 17-94. he fears about the tech. Euronews.

3. ChatGPT. Wikipedia. 21. Glenn Zorpette (2023) Just Calm Down About GPT-4 Already and stop con-
fusing performance with competence. says Rodney Brooks.
4. Tianyang Lin, Yuxin Wang, Xiangyang Liu, Xipeng Qiu (2022) A Survey of
Transformers. Arxiv 3: 111-132 22. Emanuel Diamant (2002) Biologically Inspired Image Information Con-
tent.
5. OpenAI. Wikipedia.
23. Emanuel Diamant (2009) Machine Learning: When and Where the Horses
6. Chatbot. Wikipedia. Went Astray?.
7. Chloe Xiang (2023) ‘Just Another Hype Cycle’: Elon Musk Reportedly 24. Emanuel Diamant (2008) Unveiling the mystery of visual information pro-
Building ‘Based AI’ Because ChatGPT Is Too Woke. cessing in human brain, Brain Research 1225: 171-178.
8. Emily M Bender, Timnit Gebru, Angelina McMillan-Major, Shmargaret 25. Emanuel Diamant (2015) The brain is processing information, not data:
Shmitchell (2021) On the Dangers of Stochastic Parrots: Can Language Does anybody knows about that? MDPI “Information”.
Models Be Too Big? pp. 610-623.
26. Emanuel Diamant (2015) Advances in Artificial Intelligence: Are You Sure,
9. Alejo Jos´e G Sison, Marco Tulio Daza, Roberto Gozalo-Brizuela, Eduardo C we are on the Right Track? Transactions on Networks and Communica-
Garrido-Merch´an (2023) ChatGPT: More than a Weapon of Mass Decep- tions 5(4).
tion, Ethical challenges and responses from the Human-Centered Artificial
Intelligence (HCAI) perspective. 27. Emanuel Diamant (2018) Designing Artificial Cognitive Architectures:
Brain Inspired or Biologically Inspired?. BICA 2018 Conference, Prague.
10. L Floridi, M Chiriatti (2020) GPT 3: Its Nature, Scope, Limits, and Conse-
quences, Minds and Machines. Springer 30: 681-694. 28. Emanuel Diamant (2016) Shannon’s definition of information is obsolete
and inadequate. It is time to embrace Kolmogorov’s insights on the matter.
11. K McGuffie, Alex Newhouse (2020) The radicalization risks of GPT-3 and Conference Paper.
advanced neural language models.
29. Emanuel Diamant (2018) Artificial Neural Networks: a bio-inspired revo-
12. Chloe Xiang (2023) The Open Letter to Stop ‘Dangerous’ AI Race Is a Huge lution or a long-lasting misconception, and self-delusion. A rejected Con-
Mess. ference Submission.
13. (2023) Open letter. 30. David Gunning, Es Vorm, Yunyan Wang, Matt Turek (2021) DARPA’s Ex-
14. (2023) Transcript: Senate Judiciary Subcommittee Hearing on Oversight plainable AI (XAI) program: A retrospective, Authorea.
of AI. 31. Emanuel Diamant (2016) Computational Intelligence: are you crazy?
Since when has intelligence become computational. Conference paper.

ISSN: 2574-1241 Assets of Publishing with us

DOI: 10.26717/BJSTR.2025.60.009461
• Global archiving of articles
Emanuel Diamant. Biomed J Sci & Tech Res
• Immediate, unrestricted online access
This work is licensed under Creative • Rigorous Peer Review Process
Commons Attribution 4.0 License
• Authors Retain Copyrights
Submission Link: https://biomedres.us/submit-manuscript.php • Unique DOI for all articles

https://biomedres.us/

Copyright@ : Emanuel Diamant | Biomed J Sci & Tech Res | BJSTR.MS.ID.009461. 52666