Method and Apparatus for Computing the 20 percent Chance of AI Leading to Human Extinction

 

---

TU Wien 

https://www.facebook.com/tuwien/posts/pfbid02kFy38zP49EE6Nxr4vwjQEp13TZzE1uzBqNYb7kqkWg2RSVxfbjELGb4q5ZFqZBuKl?__cft__[0]=AZhPLpDj-4JE8IVznKQYQ-Uxawx9rOmM2sMWgipWoZavKdIsvMJVHxj2_Z05lYsgtRLfDVqqSnipwrQzsSPAUwEgc3liQmoCOUU4iJeEzudWBY68c8M5MpLqzONPqTd0GKwftf8qRVU&__tn__=%2CO%2CP-R

///////

Todor Arnaudov

The Grandfather should publish the method and apparatus used for computing the exact probability of 20%. Can he send the logistics and the training run? Maybe it was trained on a dataset with histories of civilizations, creating AI (1233055 images, collected by Fei Bei Mee from MIT via transgalatic connection by SETI with the aliens in 2452 Galaxies; ithe "Astro turk" service and student Klingons helped with the data labeling, under the supervision of captain Kirk; the model architecture was designed and implemented by Ilya Alexeev, a recent PhD student, and run for a week on his gamer's PC with a single 5090 GPU); Read more in the preprint at Arxiv 2602.12443.

///////

Todor Arnaudov

Dan Ivan Thanks for your comment and to the admins at TU Wien for tolerating critical and satirical comments. Speaking "more seriously", IMO any probability estimation is not serious or "scientific" and researchers talking with such figures are ridiculous (Bostrom as well). The problem is also ill defined. Even in gambling there is rigor, empirical data, attempts etc. These are random guesses based on no data, made by masters of statistical methods and proponents of empirical methods, working with "big data" from nature. In addition there is a Credit assignment problem. Logically "it could happen", there is a chance of a virus, a "dictator" pressing the "red button", asteroids, aliens, changes in the laws of nature etc. Another irony is that the art of operation with logic and little data to compute discrete and deterministic results belongs in theory from an "enemy camp" to his "connectionist" Bible (actually "Neural networks are also symbolic", Arnaudov, 2019 )

///////

Maximilian Theodor Kircher 

Todor Arnaudov they are usually very transparent about the fact that these are nothing but their personal gut estimates.

It's a subjective indicator; they communicate subjective estimates, not testable hypotheses. Nothing dishonest or unprofessional about that, unless it is misrepresented as the latter.

///////

Todor Arnaudov

But why are they invited? Why don't you invite in some top university "Extra sensory masters", talkers with the outer galaxies - how big is the chance an Alien civilization contacting us in the next 3 years, 5 years, 10 years? How should we prepare, depending on different probabilities? The extra sensory masters may also have gut estimates. Hinton and others are taken and "sold" as authorities and given the right to speak, to fear monger or to be promoted, because their image is of scientists, their work is reproducible, based on "empirical data" etc., and not due to their skills in gambling or gut feelings. Yet this event includes the exact percentage as a headline in the title for clickbait, acting like bookmakers and emphasizing the "Extinction", not "Discussing the dangers" etc., which is "scientific". Note also that Hinton "et al." "Godfathers" were surprised and scared by the speed of development of the direction of "their" own research, i.e. they had *wrong* predictions by many decades, e.g. Bengio, i.e. they didn't really understand their own work so well. Ones who did predict it correctly are "cranks". Now the professors who didn't have a clue about their own work are authorities about futurology, where they were provenly completely off, among many other fake experts who are now monetizing trivialities, obvious and already banal and visible trends, on social media and social events.

///////

Dan Ivan

Todor Arnaudov 20% aka 1 in 5 is not that difficult to predict, that is one thing. But the situation becomes ironic that those who created/fabricated/developed now crying "Oh, we should be careful, this is definitely dangerous if developed at a higher level... we were not aware how fast and unpredictable developed itself after initial start-up!"... Go figure...

----

--- END OF QUOTE ---

///////

The Grandfather and the other aligners should read about the "Credit Assignment Problem", addressed in other posts on AGI List, including recently, and elsewhere, e.g. Minsky, 1960. 

See the Minsky's 1950s and 1960 cited work for the original list of references:

As summarized in: https://github.com/Twenkid/SIGI-2025/blob/main/AGI-The_Prophets_Of_The_Thinking_Machines-Arnaudov_2025.md

* "The Prophets of the Thinking Machines: Artificial General Intelligence & Transhumanism: History, Theory and Pioneers; Past, Present and Future", Todor Arnaudov, SIGI-2025, p.478-479 (in the edition from Jan 2026):

[Translated by Google Translate: with minor formatting]

...

Prophets and pioneers. Scientists and Schools: The Prophets of Thinking Machines 

(...)

Marvin Minsky – another participant in the Dartmouth "workshop" and a leading figure in classical AI [393], offering insights and generalizations during its early decades.

* Steps towards Artificial Intelligence, M.Minsky, 10.1960

https://web.archive.org/web/20231108233913/https:/web.media.mit.edu/~minsky/papers/steps.html

https://courses.csail.mit.edu/6.803/pdf/steps.pdf

"An interesting overview of the state of AI at the time, identifying

directions, concepts, methods, and generalizations that remain valid to this day —

such as reducing problem-solving to search (verifying proposed solutions),

gradient methods (hill-climbing), and local optimization (applicable when

the task can be represented as a sufficiently smooth function of coordinates).

The connection between gradient and heuristic approaches; the "predictive"

school of thought — mentioned towards the end in the context of Ray Solomonoff’s work;

and the importance of hierarchical modeling. Planning: the idea that any

form of planning is essentially predicting the future; systems with a "goal";

methods used in early game-playing programs like Samuel’s checkers, etc.

Classification (pattern recognition, model-schemas)

based on heuristically extracted key features of the objects under study — 

features that must remain constant despite various forms of distortion

(invariance). The challenge of discovering new, useful recognition features

and combining them to form a recognition system. Decomposing complex

objects into sub-objects and describing the intricate relationships between

their parts. Recognition methods from that era. Mention is made of

learning Bayesian networks and maximum likelihood estimate.

Parsing images composed of nested geometric

figures: a specialized language/code for describing their types and

relationships (inside, to the right, to the left, below, …); nesting and

recursion — describing arbitrarily complex objects from a small number of

component parts; note the "ability to parse" — to "devote full attention"

to selected parts of the image and apply all available resources

(to attend fully — compare with *transformers* ) — when the current

description is insufficient to achieve the immediate goal. Etc.

Here, Minsky points to the "credit assignment problem" [394]

— a challenge that continues to be studied in machine

learning as of 2024; see the 2024 overview of approaches to it

in the footnote.

* [393]  Also known as GOFAI – Good Old-Fashioned AI.

* [394]   A Survey of Temporal Credit Assignment in Deep Reinforcement Learning, E. Pignatelli et al.,

July 2024 https://arxiv.org/pdf/2312.01072 Minsky's page on the MIT server was

accessible until 2023–2024, but as of December 20, 2024, it no longer loads.

Original in Bulgarian:

"

Пророци и пионери. Учени и школи: Пророците на мислещите машини

• Marvin Minsky – Марвин Мински

Друг от участиците в Дартмутската „научна работилница“ и водеща личност в класическия ИИ [393] с прозрения и обобщения в първите десетилетия. * Steps towards Artificial Intelligence, M.Minsky, 10.1960 https://web.archive.org/web/20231108233913/https://web.media.mit.edu/~minsk y/papers/steps.html https://courses.csail.mit.edu/6.803/pdf/steps.pdf

Интересен обзор на състоянието на ИИ тогава, с разпознати направления, понятия, методи и обобщения, валидни като цяло и до днес, като свеждане на решението до търсене – проверка на предложените решения, градиентни методи (hill-climbing) и локална оптимизация, когато задачата може да се представи като достатъчно гладка функция на координати. Връзката между градиентни и евристични подходи; както и „предсказващата“ школа, спомената в края с работата на Рей Соломонов; важността на йерархичното моделиране. Планиране: всяко планиране също е предвиждане на бъдещето, всяка система с „цел“, методите в ранните програми за игри като „шашки“-те на Самуел и т.н. Класификацията (разпознаване на образци, модели-схеми, шевици) чрез евристично извлечени важни особености от разглежданите обекти, които трябва да бъдат неизменни под различни форми на изкривявания („invariant”, invariance). Задачата за откриване на нови полезни за разпознаването особености и за съчетаване на множество от тях за да се образува система за разпознаване. Разделяне на сложните обекти на подобекти и описване на сложни отношения между частите им. Методи за разпознаване от тогава. Споменават се обучаващи се мрежи на Бейс и 393 Познат и като GOFAI – Good Old Fashioned AI – Добрият стар (старомоден) ИИ определяне на максимална правдоподобност (maximum likelihood estimate). Отчленяване на изображения, съставени от вложени геометрични фигури: език/код с тясно предназначение за описание на вида и за взаимоотношенията им (във, отдясно, отляво, под, …); вложеност и рекурсивност – описание на произволно сложни обекти от малък брой съставни части; забележи – „способност да отчлени – да „отдели пълно внимание“ на избрани част от картината и да приложи всички налични средства“ (to attend fully – сравни преобразители (transformers)), когато текущото описание не е достатъчно за постигане на настоящата цел. И пр. Тук Мински посочва задачата за „приписване на значимост“ 394 – credit assignment problem, която продължава да се изследва в машинното обучение и през 2024 г. – виж обзор на подходите към нея от 2024 г. в бележката под линия. Виж също препратките от работата към авторите от времето, напр.:"

394 A Survey of Temporal Credit Assignment in Deep Reinforcement Learning, E.Pignatelli et al., 7.2024 https://arxiv.org/pdf/2312.01072 Страницата на Мински на сървъра на MIT работеше до 2023-2024 г., но към 20.12.2024 вече не се отваря.

(...)

[393] Познат и като GOFAI – Good Old Fashioned AI – Добрият стар (старомоден) ИИ

[394] A Survey of Temporal Credit Assignment in Deep Reinforcement Learning, E.Pignatelli et al., 7.2024 https://arxiv.org/pdf/2312.01072 Страницата на Мински на сървъра на MIT работеше до 2023-2024 г., но към 20.12.2024 вече не се отваря.

... 

Read More

Twenkid - The Child of AGI - is Challenging the Grandfather of AI Yann Lecun

Yann Lecun cites a post, which is acknowledging that his ideas were correct etc.

A  part of the concluding punch lines:

* It is "dime a dozen", but people decades older than me who *literally* repeated and ripped-off my suggestions, [strategy, plans, principles, theories, directions, conclusions, thoughts ...] and observations decades later, got prized with billions to *waste* and I am not even mentioned. They did it even in my own country, where one Bulgarian-Canadian became an "architect" of an institute in Sofia, with statements which were a *20 years late rip-off* of the above-cited essay, which were sold as  "innovative" and ground-breaking :))), "for the first time in Eastern Europe" etc.

* As of "Dime a dozen"--> yes, or even "Five a dozen" -->
The current  "supercomputer" of my lab is called "PETAK I", where  "Pet" means "5": from: 1. "Pentium" (historically the CPU and brand on which TUM was created), 2. The CPUs of all nodes: Core i5 (all old ones, 11-14 years old models, LOL); 3. Five nodes of the cluster (the initial full configuration) 4. A parody CPU-name from a science fiction work from 2004 from that theory ("Pentium 5"; "Петият Петак") and 5. In Bulgarian it also means "5 cents"... LMAO

"""Tunisia.AI

Ala Eddine Ayadi

 ·

Администратор

Експерт в групата на тема Изкуствен интелект и машинно обучение

 ·20 април в 21:34 ·

Yann LeCun may have been right about something important: next-token and next-pixel prediction are probably not the most efficient path to real world understanding.

For years, the industry has been scaling generative models under the assumption that bigger models, more data, and more compute would eventually produce deeper intelligence. LeCun has been arguing the opposite: predicting every word or every pixel forces models to spend huge amounts of compute on surface details instead of learning the underlying structure of reality.

That’s the core idea behind JEPA (Joint-Embedding Predictive Architecture): instead of reconstructing the world pixel by pixel, learn a compact latent representation and predict what happens next inside that space.

The problem is that these models have historically been unstable. They suffer from “representation collapse,” where the latent space becomes too simple to carry useful information unless you add complex training tricks, auxiliary losses, or frozen components.

A new paper, LeWorldModel (LeWM), shows a much cleaner approach. It trains end-to-end from raw pixels using only two losses: a next-embedding prediction loss and a Gaussian regularizer on the latent space. This drastically simplifies the training setup compared to prior approaches.

The efficiency gains are striking. The model has around 15 million parameters, trains on a single GPU in a few hours, and can plan up to 48× faster than larger foundation-model-based world models, while staying competitive on several 2D and 3D control tasks. Its latent space also appears to capture meaningful physical structure and can detect physically implausible events in controlled environments.

This doesn’t mean generative AI is a dead end. LLMs remain extremely powerful. But it does reinforce a key technical point: for world modeling and physical reasoning, predictive latent-space approaches may be far more compute-efficient than brute-force generation.

The real shift might be this: not models that generate everything, but models that understand enough of the world to predict what actually matters."""

https://www.facebook.com/yann.lecun/posts/pfbid0oEkmbuzwdvNC6JWRDoRzaDAyLNtzPZzvvman5ob89Z1v5AYvuQdTrQEjFcGJ3958l?__cft__[0]=AZaoNFuahXF3_DoSg2wL8ID4WPPsHBRWvjU51k3Sa742m0aQX2r7nl0VWyDXlyXx5eBBbxc-OWgq18zIcllnPjdIm_HNuV1Mga9hvE-Nga9s6_vMV0SjQVFusWFQhxpNCbrhQX0HjKTjh5WwzRfqeSFqWjE6X0vy_kidQqxMcHABvF2r6roXC6OCa1EcctDCHZNSND1P9hmQT8Cl2AcxeawX&__tn__=%2CO%2CP-R 

Todor Arnaudov

From the books "The First modern AI Strategy ..."... and "Stack Theory is Yet another fork of Theory of Universe and Mind" published at SIGI-2025

This idea, together with the prediction and next-token prediction (but in multi-scale, multi-precision hierarchy of resolutions of causality-control and perception), was published and explained nearly 25 years ago in Theory of Universe and Mind and presented during the world's first university courses in AGI in 2010 and 2011. Y.Bengio also rediscovered it 2017-2018 (Consciousness prior) and his example is almost literary repetition of an introductory definition from a treatise published about 14 years earlier. The author was a teenager, LOL.

Yann LeCun:

@Todor Arnaudov as I pointed out on another platform, ideas are a dime a dozen. The hard part, for something like this, is to implement it and to make it work.

The whole idea of hierarchical representations and learning by prediction is very old.

But learning hierarchies of representations didn't really work until convolutional nets were shown to do it in the late 1980s and more forcefully in the early 2010s (this took a while).

===

Todor Arnaudov:

Hi, first thanks for your answer as I didn't expect this honor. I don't disagree that there were earlier "prophets", I recently published a hyperbook with a related name (nearly 5000 pages in total), where one of the intros in one of the sections with collectons of related, prior and later work is a citation from the Holy Bible:

"There is nothing new under the Sun"

Some of the prior work doesn't get enough credit and is unknown, even the "fellow AI  historian" Schmidhuber doesn't mention them, e.g. the Soviet lab of Bongard and his colleagues etc. (E.g. once I caught Chollet literary restating insights from the 1967 book "Проблема узнавания" - perhaps he didn't know; he also rediscovers definitions for general intelligence of mine, published in 2001 (he couldn't know about it) - see the link at the end and the reviews of the LLMs).

 The Bible is called "The Prophets of the Thinking Machines: Artificial General Intelligence & Transhumanism: History, Theory and  Pioneers; Past, Present and Future", SIGI-2025 - and yes, almost nobody will bother to even open it. :))

BTW, e.g. IMO your PhD student Marc’Aurelio Ranzato deserves more credit for his pioneering work in DL and his insights (which perhaps [are] ~ also yours) -- his work is credited in my historical collections here: https://twenkid.com/agi/Lazar_The_Prophets_of_the_Thinking_Machines_20-8-2025.pdf ~p.21.

I do agree that I had to push to implementations immediately (not your type of NNs though) and perhaps my claims would be accepted after I implement them all by myself (Or if I or somebody else had - 20 years ago with no collaborators or any funding, no mechanical Turks to labe a gazillion of data and computing iterations, compared to 20 years later and all the collected resources in all senses of the word: i.e. IMO the difficulty of the implementation is supposed to decrease and be "discounted" with time like in RL; an idea 25 or 50 years ago may end up more "valuable" than an implementation in the present - see generative AI and the final citation below)

* I know about your dismissive opinion about "ideas", e.g. your comments to Schmidhuber's recent challenge, that you also could find ideas in your unpublished notes or something etc. and I've listened to your answers to him since 2022, "The path towards autonomous AI..." - I remember you defended yourself with referring to Optimal Control etc.

However many works are proposals, theoretical etc. but still get recognized, while other prior ones - don't and are even "humiliated". Also the core novelty there in my reading of the paper was also matching the mentioned TUM (and too general, it was not an implementation too); in general it looked like another cognitive architecture, which were popular in the cognitive science and the AGI community decades earlier, perhaps I have to reread it.

* I understand that if you dismiss even the German, who is at a comparable status as yours or, say he has more ground to be believed that he is, then you (and almost anyone) wouldn't recognize the claimed "priority" or even just the "contribution" of some obscure "self-proclaimed" "crank" or the mentioned theory, no matter the evidence (maybe you wouldn't even bother to check any evidence or count it as "theory" or anything).

BTW, your recent work about the brain/humans as "not general ..." also matches and is closely related to my prior work/accounts, beginning in early 2000s, however with different interpretation of the observations. The limitations don't deny the concept of general intelligence and the possibility of general principles and modules (prediction-compression etc.) I may address the correspondences in a paper.

*  Stack Theory is yet another Fork of Theory of Universe and Mind, SIGI-2025

https://www.researchgate.net/publication/398934575_Stack_Theory_is_yet_another_Fork_of_Theory_of_Universe_and_Mind_-_Appendix_Volume_to_The_Prophets_of_the_Thinking_Machines_Artificial_General_Intelligence_and_Transhumanism_History_Theory_and_Pioneers 

* The first modern AI strategy was published by an 18-year old in 2003 and repeated and implemented by the whole world 15-20 years later: Bulgarian Prophecies: How would I invest one million for the greatest benefit for the development of my country? https://twenkid.com/agi/Purvata_Strategiya_UIR_AGI_2003_Arnaudov_SIGI-2025_31-3-2025.pdf (Bongard, 1967 vs Chollet,2024 p.169-170)

* BTW, cheers from Kyuchuk Paris - that's the district in the city of Plovdiv, where TUM was created. 🙂

* This is the world's first modern "AI strategy", 2003, repeated and implemented by "the whole world" 15-20 years later: https://twenkid.com/agi/proekt.htm 

* It is "dime a dozen", but people decades older than me who *literaly* repeated and ripped-off my suggestions and observations decades later, got prized with billions to *waste* and I am not even mentioned. They did it even in my own country, where one Bulgarian-Canadian became an "architect" of an institute in Sofia, with statements which were a *20 years late rip-off* of the above-cited essay, which were sold as  "innovative" and ground-breaking :))), "for the first time in Eastern Europe" etc.

* As of "Dime a dozen"--> yes, or even "Five a dozen" -->
The current  "supercomputer" of my lab is called "PETAK I", where "Pet" means "5": from: 1. "Pentium" (historically the CPU and brand on which TUM was created), 2. The CPUs of all nodes: Core i5 (all old ones, 11-14 years old models, LOL); 3. Five nodes of the cluster (the initial full configuration) 4. A parody CPU-name from a science fiction work from 2004 from that theory ("Pentium 5") and 5. In Bulgarian it also means "5 cents"... LMAO

 Also as I predicted in 2013 (counterintuitive to all "experts" up to just a few years ago, I namely wrote this article *because* of clueless "experts" predicted the opposite; they were later cited thousands of times for their *WRONG* world-model and wrong predictions):

"Creative Intelligence will be First Surpassed and Blown Away by the Thinking Machines, not the "low-skill" workers whose jobs require agile and quick physical motion and interactions with human-sized and human-shaped environment"

https://artificial-mind.blogspot.com/2013/10/creative-intelligence-will-be-first.html

" (...) For the intellectual jobs - it's much easier to pick a computer, run the appropriate software or connect it to the service,

and get it thinking - you already have decent cameras, microphones and many sensors even in smartphones. (...) The bottom line is that the "white collars" are more endangered in current-time economy. Perhaps that kind of economy could hardly survive the AGI revolution. I guess it may turn upside down for a while - the low-skill workers could get higher pay, because intellectual activities will be done in 1 ms for free... 😉  We, the smart guys (the smart asses, see "Super Smartasses" the graphical series ) wouldn't be needed by anyone... Not that we are needed now. :))"

 * The prediction of the generative AI (however it could have been created by the late 2000s-early 2010s - it came *too late*, not too quick as Hinton and Bengio "complain"; not with gradient-descent of course):  

 -- Creativity is Imitation at the Level of Algorithms - An outline sketch of a possible path of development of the Artificial Intelligence "Emil" 

https://www.researchgate.net/publication/395129890_Creativity_is_Imitation_at_the_Level_of_Algorithms_-_An_outline_sketch_of_a_possible_path_of_development_of_the_Artificial_Intelligence_Emil

 * Petak I: https://github.com/Twenkid/SIGI-2025/blob/main/petaki.md

Read More

Lazar - the volume from The Prophets of the Thinking Machines with a new version with corrections up to 29.4.2026

Read the latest corrected edition of the volume "Lazar", about 130 pages, consisting of an enormous survey and summary of many fields of AI and AGI, historical and recent.

Other updates may be published soon with more narrative introductions, summaries and conclusions.

Download from https://twenkid.com/agi and the Github page of the virtual conference Thinking Machines 2025/Self-Improving General Intelligence 2025 (SIGI-2025):

https://github.com/Twenkid/SIGI-2025


Direct link: https://twenkid.com/agi/download.php?file=Lazar_The_Prophets_of_the_Thinking_Machines_20-8-2025.pdf

Perhaps I will upload it to Academia.edu and ResearchGate perhaps after a few more editing and additions.


A brief contents listing some of the domains:

#lotsofpapers  ... #lazar

Lots of Papers: In AI, ML, CV, ANN, DL, …  throughout history, classical 1950s, 1960s, 1970s, 1980s, 1990s, 2000s, early 2010s to 2020s. Computer Vision, Reinforcement Learning, Program Synthesis. Lifelong Learning, Human-Computer Interaction, Mixed Initiative Interfaces (Agentic Systems); Evolutionary programming, Genetic Algorithms, Self-improving agents; Speech Synthesis, Speech Recognition, Audio Generation etc.  Groundbreaking or important researchers or related to the flow and context of the reviewed topics; and a few Bulgarian researchers who participated in some of the works.

*  Lifelong Learning, Continual Learning, Reinforcement Learning (historical Q-Learning, modern Deep Q-Learning: Atari DeepMind …), RL for LLMs, policy optimizations (DPO, PPO, OREO) … Conditional Random Fields (CRF); Chain-of-Thought prompting …

* Survey of other techniques and research in computer vision,  preceding the explosion of the application of convolutional neural networks after 2012: DBN, RBM, MRF, SIFT etc.

* Survey on Early Deep Learning architectures and Normalizations; seminal papers and PhD theses by pioneers in DL from the schools of LeCun, Hinton, Bengio: M.Ranzato, V.Mnih, A.Krizhevsky, I.Sutskever, A.Mohamed …

*  Survey of Object Recognition and Classification before Deep Learning

* Selected Computer Vision works from 1960s to 2020s

* Exploration and introduction of concepts and techniques in computer vision, machine learning, neural networks

*  Mixed-Initiative Interaction (Agentic Systems)

*  Audio: Speech Synthesis, Audio Generation, Speech Recognition; from 1980s to 2010s

*  Neural Machine Translation, Language Models, LLMs, Text Generation, Text and Language Learning and Representation, Word-Embedding

* Alternative approaches for sequence and next words prediction, instead of neural networks: stochastic memorizer, sequence memorizer* Transformers - the seminal paper from 2017

*  Transformer architectures for images and for reducing the quadratic complexity

*  Neural Program Synthesis

*  Transformer architectures for images and for reducing the quadratic complexity

*  Evolutionary Algorithms | Genetic Algorithms | Genetic Programming
*  Genetic Programming, Genetic Algorithms, Evolutionary Programming: Part II

* Summary and selection of important concepts in Evolutionary Algorithms | Genetic Algorithms | Genetic Programming etc.

*  Vision Transformers – ViT

*  Multimodal Learning, Dialog Learning, Continual Learning

*  Diffusion Models

*  Self-Improving General Intelligence, Recursive Self Improvement

 ...

Read More

Common Confusion About Agency, Will, Soul etc. of Humans and Biological and Artificial Agents - Todor's comment on The Asymptotic Illusion: On the Ontological Limitations of Artificial Agency

A discussion about the paper on Academia.edu:
* The Asymptotic Illusion: On the Ontological Limitations of Artificial Agency, Mohamad Al-Zawahreh, 25.12.2025
https://www.academia.edu/s/06c854dde5

Todor Arnaudov
10 days ago

Hi, this is a common confusion - building a straw-man of Turing machines or "computation" as defined in this abstract sense, while the counterpart argument refers to the full "stack" definition of the systems, down to the lowest level virtual universes (See Theory of Universe and Mind, T.A. 2001-2004+, e.g. "Stack Theory is yet another fork of Theory of Universe and Mind", 2025). Neither computers, nor LLMs, AI (in whatever form) or whatever are "just" algorithms, "computation" etc. - they are immersed in the Universe and their implementation includes those "non-computable" parts as well. On the other hand, all these advanced thinking, theories, logic, reasoning about "quantum" etc. would have been impossible without all the science and technology, all "mechanical" processing etc., i.e. the mighty "conscious" being wouldn't have been capable to even formulate these problems and their [personal] "superiority", they would have been just silly apes. There are a lot of discussions about that in TOUM, including the latest additions in the hyperbook "The Prophets of the Thinking Machines: AGI & Transhumanism: History, theory and pioneers; past, present and future". One point which I addressed in a letter to the cognitive semioticists Jordan Zlatev: the LLMs have no consciousness, but "just" etc. [see in appendix "Reflections on Everything" (Listove) from "The Prophets"]. However I haven't read anything but my own thoughts regarding the following problem, which is valid for humans and their mind as well: WHAT[an LLM is] and WHERE *exactly* the LLM is located, why and who (evaluator-observer) sets the boundary exactly there (at what resolution of causality-control and perception) and what *exactly* it is when fully defined in order to function. It is not "just" the weights of the NN (what the h* is "NN" in the physical universe - it includes the whole hardware, all chips, all connected computers - your PC or smartphone, the routers, the Internet servers, the main GPU servers; it included the original data, the data in the memory (other representation); it includes also the *HUMANS* who interact, their biology, their states; it is a result of a chain of operations and activities of all related humans and other technologies and ultimately it requires the whole universe to be what it is. However, this is refuted and "vulgarized" to "they are algorithmic, they are "just 1s and 0s", "switches" (the earlier nonsense about computers, still defended e.g. by F.Faggin). No, computers are not "just 1s and 0s", and they emerged and operate in the same Universe as "humans"; besides humans are created and built also *WITH* the participation of all computers and technologies, i.e. humans are *PARTS of the machines* as well. "We are merged" without the need of "physical"/"short-molecular-distance"* way of interaction. See also an article about that in "The Prophets ..." in volumes: "Science Fiction. Futurology. Cybernetics. Transhumanism" and cited also as an appendix in the small monograph "Universe and Mind 6", which is related to this discussion. * Another school with this confusions (or choices) is "Relevance Realization", which is explaining definitions of Theory of Universe and Mind, published more than 20 years earlier, however they put a stamp "this is not computable", "the opposite of computation", while the original [work] claims it is computable, but hierarchical, multi-scale, multi-domain, multi-precision, preemptive, has "interrupts" etc., they are built by the same fabric of the Universe computer and they interact with it and fetch "sensory" data from the common "pool". The real computers are like that, they are not the primitive and abstract strictly serial, one scale, "fixed symbols" Turing machine outputting AAABBBBBAAAAC-HALT. --- References: * Stack Theory is yet another Fork of Theory of Universe and Mind, T.A. 2025, SIGI-2025, https://artificial-mind.blogspot.com/2025/09/stack-theory-is-yet-another-fork-of.html

* Reflections on Everything/Listove (Листове, Листове по всичко) - a volume from The Prophets of The Thinking Machines: Artificial General Intelligence and Transhumanism: History, Theory and Pioneers; Past, Present and Future, T.Arnaudov, 2025 https://github.com/twenid/sigi-2025 See the cited excerpt from Todor's letter to Jordan Zlatev on what and where the LLMs are (it will be published as an article/paper as well), after p.340 "* Sentience or consciousness of another “entity” is in the eyes of the evaluator * Thoughts from a letter by Todor Arnaudov to the cognitive semioticist Jordan Zlatev , 16.8.2025" * "We are merged..." - see my article, answering Tim Tyler's video, published both in Reflections on Everything and in the volume "Science Fiction. Futurology. Cybernetics. Transhumanism" from The Prophets ... *Коментар на Тодор Арнаудов към видеото на Тим Тайлър „Да се слеем [с машините]“ от 10.2023 Tim Tyler: Let's merge! (...), p. ~175 in Listove The text in English, posted on Youtube: Todor: IMO humans and machines - actually the technologies, the recognizable systems, entities in the Universe (and the ones which are not recognizable for now) are merged anyway. The human individuals as bodies, entities are one "view", "rendering", a way of sampling of the actual intrinsic representation of the underlying properties and processes. The same goes for any piece of hardware, computer, robot, any object. They are what they are under a particular sampling of the data, in another sampling they are physically, causally, energic-based etc. connected and part of the causality, influence, events network and interrelated. In one philosophical school humans are defined as the set of "social relations". Also human individuals without technology and starting from scratch with no previous culture and language, which is also technology, are not very much more capable than apes in the first generations, or in the first hundreds of thousands of years. Technology, the environment and previous recorded knowledge and the social, scientific and technological "software" that gets loaded into our minds, all from the whole universe allow human beings to be so clever etc. So we are part of the human-machine (technology) system anyway.

---

The distributed representation of humans and everything in a field-like form, the causal id/causal tags* theory etc. is discussed also in other volumes, such as "Universe and Mind 6" and "Is Mortal Computation Required for the Creation of Universal Thinking Machnes" at SIGI-2025. ** Note that "causal IDs" in my explanation are not referring to Michael T. Bennet's "causal identity" concept, but the match of similarly titled concepts in his fork (with unrecognized prior work) is funny. The causal IDs in my speculations are about hypothetical information and interaction entities or signals which allow or make the wholes, beings, systems to "feel" as a whole. I derived them from the insight that all systems in all scales and multi-scales encounter inevitable lags between their constituent parts, and these lags are varied. One way of a given causality-control unit - the building blocks in TOUM - to know that a signal is from a given whole, could be by special interactions and information, these "causal IDs", which could be updated and kept active for particular reasons, e.g. a chain of particular chemical, physical or whatever interactions within given .... etc. and fade. E.g. on the page on Github about TOUM in late 2023; most of "Universe and Mind 6" was written in the Spring of 2023 and in 2023-early 2024, but it was published as a completed monograph in 2025. https://github.com/Twenkid/Theory-of-Universe-and-Mind

  

Read More