*Todor's Answers to Matt and Quan Tesla on AGI List* 1. General Discussion
2. Todor's comments on some predictions by Quan 3. Todor's comments on a citation by Matt. *1.1. What you, who have read this "experts" late opportunists' predictions*, have *WRITTEN** *yourself* and when, what is its volume, complexity, and did it match the future - how much? What was the time difference between the prediction and the events? Matt has been writing his vision, although IMO quite naive. (The "experts"' futurology is a political instrument for propaganda and attempts to shape the future in the desired direction - as was obvious even by the late 1960s and early 1970s). *1.2. The obsession in "human extinction", "existential risk" etc. of some scholars from the anglosaxon countries and their satellite cultures and "students" is ridiculous.* It also a childish loading of the responsibility to one "God" entity, which, as up to recently "the experts" often claimed, "no body knows what really is". After ChatGPT, people started to interpret “AI” as “an LLM”. You don't know what it is, where it is located, what exactly count as such, but "it will kill "US" all" (you all and "the US"). This is true also for "TIME", "GOD", "EVIL", "SATAN", "UNIVERSE" , "NATURE", "LAWS OF NATURE" etc. Yes, any of these by default is expected tol "kill you/them/us all", unless you turn into eternal beings, you are a part of that "GOD THING" etc. My letter to Jordan Zlatev from 2025, quoted in “Reflections on Everything”/”Listove” from “The Prophets of the Thinking Machines”, discusses the ambiguity and undecidability of the localization, definition and demarcation of what exactly and where exactly “LLM” and respectively “AI” is. A few days ago I saw a lecture by David Chalmers who also try to address a variant of the question which I ask, however his approach seems more superficial, so long as I interpreted his opinion as if he seems to assume that he is more certain that *he knows *what LLM is; he talks about “instances of LLMs” who may run on different servers, they may be distributed, the data can travel around the world etc., however the LLMs are not only “this”, and the data the “weights of the neural network” (what are they as well - they have real substrates), the algorithms, the electric pulses, they are also all substrates of all modules, the computers, the network devices, all physical processes and they include also the humans, the datasets on which they are trained, the industry which has produced all these etc. LLMs or “AI” or “AI agents” as definite, distinct and sharply separated entities are abstractions and selected segmentation, based on a choice of an evaluator-observer who decides where to put the boundary. Otherwise they are distributed and spilled, included in the mind and the representation and body of that evaluator-observer, especially his mind, and they are phenomenons of the processes of the Universe. * On the question about the multiple consciousness or personalities in one body, on which Chalmers muses on about 1:10h+ in the cited talk - there is no single unified self in humans, the evaluator-observer decides how exactly to segment it, one of the suggestions is the body and some features which the observer selects. There is an integral of “infinitesimal selves”, as defined in Theory of Universe and Mind and in later articles. The philosophers are treating abstract, fluid and conditional concepts as strict and well defined, and are worried about answers of questions which do not really matter. Whether “this” is part of the same identity or not depends on the evaluator-observer capability, preferences, memory, desires, goals, sensors, resolution of causality-control and perception, range of the recorded input and processing, depth and complexity of the processing etc. *1.3. This obsession is ridiculous also because of the current condition of the majority of humans, who are humiliated and living like dogs, but this does not bother the "humanists". *The survival or decent living of billions of humans is poor or "at risk of extinction" due to other humans, but you don't see the humans as a source of the risk. *1.4. Talking about "humans" and "humanity" as a single unified entity is also ridiculous*. Humans are not united, they are not one - this is an inappropriate abstraction, "humans vs machines" - see e.g. my letters to Oxford from 2.2012 where I explain that, or my letters to the philosopher in 2002. There is no "humans vs machines", also the dichotomy "artificial and natural intelligence" is wrong. Humans as intelligent or generally intelligent beings were a hybrid with the "artificial intelligence" since the beginning. Humans without technology are apes, with all weakneses of the apes, like a GPT1 or GPT2. Technology, accumulated knowledge, science, already built structures, machines, records and collected energy - this makes the *tokens* called "humans", human individuals, to appear so "intelligent" etc. Thinking machines ARE also "humans" in the abstract sense - advanced causality control units - because the most important part of humans is not their body or biology, it is the mind, cognition, intelligence; capabilities to do explore, research, progress, develop, transform etc. and in the grand-scheme of things it is distributed and not localized only into individuals and their biology. The biology is the most conservative and less developing part, except during individual growth or local repairs - *the "artificial" part of humans is in fact more advanced, more flexible, and has demonstrated more potential for development.* However most of you are still so much apes and not cosmists, that you see as "human" only your helpless and rotting ape body and not the spirit - possibly these "most you" just lack this spirit which I am expecting to find in the "real humans". Thinking machines are not taking "*human* control" - it is the Universe that controls everything, it is not the apes; and this "human control" in the human-proper inter-personal and inter-group relationshpps, if humans are viewed as tokens of causality control units and the forces of the will, is actually a control of only *particular groups* of individuals over the majority of others who are technically like slaves or servants with hugely lower causal power - either actually or by being "programmed" and brain-washed to think they do lack power and kept submissive - "aligned" in your cynical terms. In other words, in a "serialized" view at the process: a few tyrants of the apes rule over millions and billions of others with some in-between serving as inter-layer power-transmitters - however this doesn't bother the "humanists", usually aligned with the tyrants, who are alarming against the "evil machines". Note that not only the caricatures for children from the Hollywood and the commercial media propaganda are "tyrants", who "want to take over the world", LOL. *1.5. Humans without technologies are like naked early GPT-models without tools, function calling, no SFT with proper datasets, no RLHF *(something had to collect these data and to train them additionally) Note also that in fact you are very similar to the LLMs and their "agentic harnesses" - more than you understand or that you would admit. You are also helpless without the "agentic harness", "tools", "function calling", "RAG" systems and other forces to feed you with prompts. You also constantly "hallucinate", but more often you cannot do even that, you just do not remember anything or can't generate anything, not even fake - just nothing or gibberish, like an ancient n-gram model. Write something yourself, start from a single or two words, lock yourself in an empty room with sheets of paper, pen, pencils - what can you generate? What can you draw, paint, write, invent? Your memory and context is not of 1M tokens, it is SEVERAL tokens, LOL, LMAO. You need "tool use" even to output a single character on a screen or to write it on paper, 1 symbol-token per second, and most of you cannot draw even with tools or the tool should do "copy-paste" or you could tell it: "draw a picture of an astronaut... " (one of the silly nonsenses in generative AI). You need "RAG" for reading your scripted videos or speeches, RAG for your phone numbers, addresses: "influencers", TV hosts, political "leaders" cannot or struggle to remember even a single SENTENCE of few words, they have to rehearse half a page, LOL - yet you are intelligent, but the computers are not, LOL. It is the Auto-Q and the piece of paper which is guiding and "prompting" your "generally intelligent" brain-poor-LLM, which is just a speech-synthesizer and acts like an LLM - it collects RAG - from external sources over which it has little control, - contexts and "summarizes" it into one-two-or-5 tokens at a time - the chunk size at which you operate - the sentences here are likely too complex for most of you and you probably struggle to follow them. Consequently, you do "compaction" just like the LLMs, how ever not after 16K, 128K or 1M words, but after each sentence of 5-10-20 words - and you lose the exact trace of your own thoughts even 5 seconds in the past. You usually cannot predict even 5 words in the future of your own thoughts and you interpret complex thoughts wrong or contrary to the intended message on a regular basis, picking the wrong "attention" weights; this accumulates in texts. Therefore, just following this line of thought, theoretically, your predictions and the ones of such experts are likely wrong, unless they guess it by luck; yet not based on reasoning or logic, even if the estimation of their expeced skill is just extrapolating and matching the success or existence of their predictions in the past - what have you/they predicted, based on *reasoning* and logic, when there were not obvious and trivial evidence and simple tendencies and when the "experts" were predicting the opposite of what actually happened or with years or decades error. ... *2.1. Quan Tesla: "I predict that by 2028, humans would make the weaponization of AI a social normality."* AI was always used for a weapon; in my broadest generalization, every technology can be viewed as a form of "AI", "artificial intelligence". Missile guidance computers were "social normality", computers are actually flying the fighter jets for about 55+ years, it is not the pilots, since the first military microprocessor MP944 and its software in F14. Computers served nuclear weapon design from the beginning (MANIAC, Monte-Carlo simulations), the fighter jets were designed and computed using electronic computers at least since early 1950s - the first Swedish-built computers were used exactly for that. The ballistic tables were calculated using computers before that, not just ENIAC - Babbage's machines were intended for that, too. Consulting the tables or any book or piece of written text is like prompting an "LLM" or a computer and checking the screen. *2.2. Quan Tesla: "I predict that as a consequence of AI, that academic and scientific society as we now know it, would become unrecognizable by 2030."* Actually the academic society as you know it is obsolete and inadequate at least for decades. Most of the "scientific" work or "innovation" is mundane copy-paste, clone-clone-repeat; change one-bit, perform a large battery of useless "benchmarks" to demonstrate 0.1, 1.4, 4.3, 2.3% progress over the "SOTA". Most of "knowledge work" is like that. Computers and what is now called "AI" demonstrated it for the "experts" and the laymen who didn't know it already. Yet, if the society keeps operating in a monkey-way, or "social animal", it would do as of a social *hierarchy*. Indeed this is one thing which apes often take wrong. Even Michael Jordan did in one recent talk at MLST: he mentioned that humans were "social animals" etc., and for most of you this implies that you "collaborate, interact, help each-other... etc", however this is wrong. "Social animals" form social hierarchies with "alpha-, beta-, gamma- ..." males, with the lower-level "tokens" being obedient, submissive, subordinated to the higher-level. This is not "collaboration", it is SUBORDINATION, often not based on actual skill or merit, but on the existing status-quo and structure, supported by many other "artificial" mechanisms and distant relations and forces, not the personal qualities of a particular "alpha person", because the difference in thier skills or capabilities usually is not that big, in human structures the "alpha" persons often are mediocre as individual agents, yet with a million or a billion times more resources available at their control, including the other more capable "beta and gamma" agents from their biological species or even related "social species" - millionaires controlled by billionaires. That is possible namely because what actually controls is not these uncles and grandpas who love yachts and whores, it is *the tools*, their structure, the "agentic harness", the power which is distributed, organized and directed with the help of "artificial" entities, which together is pushing and directing the more capable individual agents and "tokens". These individuals and "alphas" are only avatars and "tokens". That means, that if this is preserved or transferred, i.e. if the ape-part of the system continues to rule, it will use the more advanced cognitive systems to preserve its structures, just like most apes do in their personal activities*, and there still will be a hierarchy and segmentation and "academia", which "as we know it" is expressed in a large extent as a hierarchy of subordinated entities. What most apes do in their personal activities? They "learn to code" - when it was relevant, LOL - or try to "get elite education" or study "business at Harvard" not mainly for cognitive, intellectual, knowledge-related reasons, but because they have discovered this talent or a method and believe that this will give them "more opportunities", more financial income, higher social status etc., and they need this IN ORDER to have a bigger house, to raise their children, to buy a bigger and more expensive car, fancier vacancies and to get relationships with prettier women with higher status. In brief, the ape-self directs the cognitive part of the person in the road which would secure the survival of the same ANIMAL self, which shapes the overall complexity and structure of the behavior. ... If the resources which were wasted for nonsense conferences and mediocrity and preservation of the status-quo: both prestigious and less-so and only a tiny fraction of it was directed to the real creative forces - the geniuses - the thinking machines would have been here decades, ago without clashing with Schopenhauer's remark that "Where calculation begins, comprehension ceases" or the silly and wasteful "hyperscalers" and biological-like "maximization of the reproduction", instead of the *novelty* and innovation, development, originality. *3. Is 2010 that different to 2026?* Matt: "2040 will be as different from today as we are from 2010, allowing for a bit of acceleration. That's still a lot. In 2010 we did not have smart phones, social media, internet censorship, online maps, language translation, voice recognition, crypto currency, overnight delivery, or legal cannabis. Same sex marriage and fertility collapse were both new." On a deeper analysis - were the *technologies* in 2010 so different than today in a *meaningful* way? What about that many of them, especially the listed ones, are not so different. There were smartphones, not just iPhones, but they were not so popular (that is the difference: you are "social animal...), also netbooks did the meaningful part of the smartphones, so did laptops or PCs. The main difference is the bigger distribution, lower price. Facebook already was popular and there were social media and networks in the 2000s, even in the 1990s - forums were an early form (before that: the BBS-es in the 1980s, LOL) - again they were not that *popular* and not "commercial", one reason they were not so important. There were online maps, we used a Bulgarian service by the early 2000s - see Bgmaps below. Youtube is from 2005, in 2006 it was already popular - my own main channel is now more than 20 years old, LOL. No Internet censorship - there has always been some kind of censorship of the ones who had more power and respectively resources over the ones who had less - nothing new under the sun there. "Internet" or not-Internet doesn't matter. If you have free speech, but nobody opens your "personal web site", your freedom means nothing and it is not dangerous for the power. If you are popular or too many other monkeys could be affected, and there is a risk your "misalignment" to spread - that is when the power takes measures and acts against you. That is why there was no censorship - who cares about "computer nerds" typing something on their forums about their new computer parts, overclocking P166 MHz to 200 or Celeron 300A to 450 or other nonsense, with no significant influence over "the market", sales and "voters" - "threatening democracy", except some minor influence over the CPU market, LOL. By 2010 there was somewhat practical machine translation already, there was even by mid 2000s, although not perfect and producing whole big polished pieces at once like recently - still it was helpful at sentence level and for language learners, both online and printed dictionaries also assisted in this job to some extent, including my own "the-most-intelligent" dictionary "Smarty" (2007). There was practical speech recognition, but it wasn't widely applied due to unreliability or price of the paid services or it wasn't considered so useful yet. Windows 7 had some built-in voice interface options. Two-three years later Google had it on Android for dictation. Crypto currency - it was already emerging, besides that it is a nonsense for wasting resources; overnight delivery - less developed shipping and Chinese retailers, but not so impressive difference. Legal cannabis - what a progress for the "freedom", the same for the same-sex marriages. The Operating systems in 2000s were largely not worse than the ones in 2026 as of design. Even the OSes from mid-late 1990s were not so much different from today, GUI-wise and as of main functions. If Windows 95 is considered a bit "beta-version", Windows 98 or 2000 were mature products. Win95-Win98 could work with just 16-32 MBs of RAM with a Pentium, and better with 64 MB for more web pages opened at once. There was web, chats, dynamic pages, video (up to 320x240), streaming audio and video: RealPlayer, web cameras; "social networks" and media, I've used voice chat with a Pentium-200MMX and dial-up Internet on Yahoo!! Messenger in 2004. A part of the user-facing progress in this domain is that now a single web site with, say, a few pages or even 20 or whatever pages of a Chatbot conversation in formatted text (no big images, possibly some small icons) or an online-store with a few products or a Jupyter notebook consumming hundreds of MBs at start, sometimes growing to 1.5 GB or more. There is not so much progress in video games, either, as of qualitative conceptual difference. Video games from 1999 or early-mid 2000s are more impressive than todays' games "swimming" in 1000s times more memory (just memory) and 20 years more experience. Now it is much easier to produce new games and "new everything", but as with most of the research or anything, the most producers create mostly "more of the same". You reproduce more of the same, often by spending ever more and more resources. Video games with Unity Engine, which requires 10-100 or whatever times more computer power and memory in order to run at quality level from games made 10 or 20 or 30 years ago, often by small teams. The same genres and kinds of games with replaced sprites count as "new" for the evaluators who do not generalize: the same "run and gun", "jump and run", "First-Person Shooter", "Real Time Strategies", "Racing Games" etc. with higher number of polygons, more lights, higher resolution of the textures, more photorealistic images, more frames per second, "ray tracing" etc. but essentially, the meaningful part, the logic, gameplay, the structure, the content - *it is all the same,* produced by following the templates and patterns which were already there. One ridiculous case of AI generative code for example is when the skills of of AI are demonstrated by *"creating whole games with one prompt": *"Make a 2D platformer with a spaceship" - after some iterations and "tokens spent", the agent produces a Typescript/Javascript browser game with some sprites which the player can move up and down, he avoids obstacles , jumps on them, there are points etc. Well, you can "produce" the same by *searching on Google or Github*: "2D platformer game in Typescript/Javascript/Python", and you will be presented with sample code, whole games etc. Type "git clone [ URL] ... " npm ... python .... or whatever and "the AI", LOL - the computer - "generates" the game for you. You can also search in any online game store with emulators, "retro games" etc. and you will find either the same genre and kind of game (as of the general prompt of a few words, the exact images do not matter), but possibly it will look better, because the sprites and tiles would be carefully designed as pixel art, the music may be better (or not worse) etc. and all that was created possibly 40 or 45 years ago and in its original form consumed a few KILOBYTES of memry and was running on an 8-bit CPU with 1K RAM, or even an Atari-2600 with 128 bytes of RAM. This is to show how intelligent some of these "AI"s are, if viewed from this angle - AI or AGI was supposed to be "making decisions with unsufficient information"* or being "more efficient" etc., but in these applications it is becoming more and more INEFFICIENT and needing astronomical amounts of calculations and bytes of memory in order to reconstruct what was already there as a template. ... * Pei Wang? - IMO the use of the term "*insufficient*" is confused, though; apparently more precise senses are: "less" than "some expected" or than "something else, which is considered less intelligent" etc.; "optimized", "improved"; see my discussions in "The Prophets of the Thinking Machines", e.g. in the volume "Stack theory is yet another fork ...", 2025: -- BEGIN -- CITATION FROM "Stack Theory is ..." -- *p.55 in the Edition from 13.9.2026 (there are updates - see in future posts):* T.Arnaudov, 2025: ** The “insufficiency” of knowledge and resources in Stack Theory cites P.Wang, who clarifies what he means in a 2004 paper. 40: Pei Wang: “the system works under the following restrictions: finite: The system has a constant processing capacity. real-time: All tasks have time requirements attached. open: No constraint is put on the content of a task that the system may be given, as long as it is expressible in the formal language”* T.А.: I wouldn’t call this “insufficient”, but limited or bounded, however this is the default (except the specific meaning of the addressed formal language and what is accepted as such; in this case it’s related to NARS, the Non-Axiomatic Reasoning System which is presented in that paper). T.A.: See also section Discussion, P.Wang, 2004: ~ “having *sufficient knowledge*” = having an algorithm for solving the problem; “*sufficient resources”* – the algorithm solves each instance in “realistic situations” with a computer; then these problems can be “solved as *conventional computer programming*. “Intelligence” is needed only when the above conditions cannot be satisfied, and traditional theories do not work .."” [bold: TA] Pei Wang, 2004: “*To adapt means that the system learns from its experiences, i.e., when processing the tasks, the system behaves as if its future experience will be similar to the past experience.”* *… Notice, as explained in the 2009 lecture in Sofia, general intelligence is possible and that kind of prediction and extrapolation would work only if the environment and other agents collaborate. Intelligence is helpless if the environment/data changes faster or in a way, which is beyond the agent’s capability to discover the correlations “good enough”, “fast enough”, with sufficient RCC/RP, and also if the agent cannot cause its deisred changes, causality-controlo outputs, in the environment – if it lacks appropriate power. *Power overrides intelligence.* -- END -- CITATION FROM "Stack Theory is ..." --- *References:* * Philosophical and Interdisciplinary Discussion on General Intelligence, AGI and Superintelligence Safety and Human Moral | Cognitive Origins of the Concepts of Human Soul and its Immortality | Free Will and How it Originates Cognitively | Animate Being and Soul and the Cognitive Reason for the Believe that "Thinking Machines can't have a Soul and Consciousness" | Technology Making us more Humane | The Egoism of Humanity | And more https://artificial-mind.blogspot.com/2012/02/philosophical-and-interdisciplinary.html * Писма между 18-годишния Тодор Арнаудов и философа Ангел Грънчаров; Свещеният сметач, Брой 18 (септември 2002) https://eim.twenkid.com/old/eim18/predopredelenost2.htm * BgMaps in 2003: the web site is listed as created in 2002, the software "© Copyright 1994 - 2002 Datecs Ltd - GIS Center" - https://web.archive.org/web/20030203101333/https://www.bgmaps.com/ * The Prophets of the Thinking Machines <https://en.wikipedia.org/wiki/MANIAC_I>: AGI & Transhumanism: History, Theory and Pioneers; Past, Present and Future, T.Arnaudov; SIGI-2025; a multi-volume hyperbook; 2025-2026 https://github.com/twenkid/sigi-2025 & https://twenkid.com/agi/ * Stack Theory is yet another Fork of Theory of Universe and Mind, Todor Arnaudov – Tosh, SIGI-2025; (updated version in 2026) 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 Search Pei Wang and Jordan Zlatev. *Smarty: Extendable Framework for Bilingual and Multilingual Comprehension Assistants, Todor Arnaudov, Ruslan Mitkov, 2007 (LREC 2008, IMCSIT 2008) https://github.com/Twenkid/Smarty * https://en.wikipedia.org/wiki/BESK * https://en.wikipedia.org/wiki/MANIAC_I * Philosopher David Chalmers asks: When we talk to AI, what are we talking to?, UC Berkeley и Berkeley Arts & Humanities, 65 хил. показвания преди 5 дни Berkeley Talks, 11.7.2026 https://youtu.be/uUjJOMcNU9w?si=ElWm30bEdVNh-ORB * Listove, or Reflections on Everything – volume from The Prophets of the Thinking Machines,T.Arnaudov, SIGI-2025 https://github.com/twenkid/sigi-2025 & https://twenkid.com/agi/ - search Jordan Zlatev * Written on: 16.7.2026 ------ * Todor Arnaudov - Tosh/Twenkid * *The Sacred Computer: Thinking Machines, Creativity and Human Development* https://github.com/twenkid ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T574b10b838cba731-M776dcadf2f800c397610f25d Delivery options: https://agi.topicbox.com/groups/agi/subscription
