Talk:Superintelligence

Too many lists, other types of superintelligence, definition

Lists can be good for understanding. But this article probably has too many numbered or bullet lists (currently 10, which is much more than any article I have seen). We may consider reducing it by half.

One other potential issue is that there is one section about biological superintelligence in the middle of sections about artificial superintelligence. ASI clearly deserves more content than biological superintelligence, but that might confuse readers. Perhaps we could have a section "Other forms of superintelligence", covering biological superintelligence and other speculative paths like networks and orgnanizations brain-computer interfaces.

I think the first sentence was better before (I believe this could be removed, for accuracy and simplicity: ", ranging from marginally smarter than the upper limits of human-level intelligence to vastly exceeding human cognitive capabilities", see this discussion) Alenoach (talk) 01:35, 12 September 2024 (UTC)

Biased graph

Hi, The graph shown here (File:Test scores of AI systems on various capabilities relative to human performance - Our World in Data.png) can't be generalized, contrary to what it seems to be. There are numerous sources cited on that show otherwise. Yann (talk) 18:12, 24 September 2024 (UTC)

I made this alternative version that precises the benchmark names, so that it's precise about what the results are. Is it ok to replace the old image with it? Alternatively, there could also be this one from the 2024 AI index, that gives very similar data than the one from Our World in Data, but that presents fewer benchmarks.

Otherwise, if you disagree with the concept of showing benchmark results, I would say that benchmarks are not perfect, but they offer a factual metric. As you showed, there are articles that claim that AI just doesn't understand anything and are just stochastic parrots, or that show examples of where AIs fail. But there are also many other researchers (most I would guess) that simply consider that AI understands, to some degree. And when AI starts to surpass humans on every objective test for understanding (including tests for which the data is private and unavailable from the internet), can we still contend that it doesn't understand? Alenoach (talk) 19:54, 24 September 2024 (UTC)

If you're so set on it, why not show us a better graph, more correct one? --~2026-11142-27 (talk) 14:16, 24 February 2026 (UTC)

Draft:AI-based religion

Related. I've just started it! Nextada (talk) 11:52, 5 June 2025 (UTC)

Biological Superintelligence Temins

"Several writers have suggested that human civilization, or some aspect of it (e.g., the Internet, or the economy), is coming to function like a global brain with capacities far exceeding its component agents. If this systemic superintelligence relies heavily on artificial components, however, it may qualify as an AI rather than as a biology-based superorganism." Considering such system as an AI is incorrect, since it's thinking processes rely on human brains, and computer systems are just used to connect brains to each other. By analogy, "a human with 2 peg legs, 2 mechanical hook-like arm proshtesises, 1 denture, which walks on 2 crutches" is not a robot - as his movement is directed by brain and done by muscle contraction, while inorganic parts are used so he can utilize muscle contraction in correct way. --95.167.183.86 (talk) 20:54, 30 August 2025 (UTC)

I removed the sentence, I think it is too ambiguous. Alenoach (talk) 21:57, 30 August 2025 (UTC)

Group Mind

"Undid revision 1308767361 by 95.167.183.86 (talk): this paragraph is about improving *individual* intelligence, and the concept of group mind appears to be mostly used in science-fiction rather than future studies"

Brain-computer interfaces could be used to create group minds, by sending data from one brain to another and back. And, it can be used to increase *individual* intelligence - since it is possible to be member of group mind and still be individual with it's own goals and ideas (since there are many types of group minds, and some types allow for preservation of individuality). --95.167.183.86 (talk) 14:52, 1 September 2025 (UTC)

As an encyclopedia, Wikipedia aims to synthesize what reliable sources say. Do you have a reliable source covering the concept of group mind as a plausible candidate for superintelligence? Alenoach (talk) 01:13, 3 September 2025 (UTC)

Just because something "appears to be mostly used in science-fiction rather than future studies", doesn't mean that it's impossible or won't be invented. Many things from old science fiction got invented in our time (e.g. handheld long-range communicators, aka smartphones) - and many actual serious future predictions have failed (e.g. think what Victorian-era authors thought about what 21st century would look like). As such, hard sci-fi (but not soft sci-fi) can be put on same level as actual future studies. --Ejkohojkjkohokjh (talk) 20:22, 3 October 2025 (UTC)

I agree. But when we cover science-fiction, we usually try to clearly distinguish it from statements about what will or may happen in the future. When a concept like group mind appears in a science-fiction work, it doesn't necessarily mean that the author thinks it may well happen in the real-world, it might just make for an interesting story (although as you noted, hard sci-fi may perhaps be an exception).

To include content on group mind to this article, what would be needed are reliable sources that directly supports the statement and that cover the concept of group mind in relation to superintelligence. There is also in this article the consideration that mixing science-fiction with explicit statements about the future may confuse readers, and we may not have sufficient science-fiction-related content to make a separate section on it. Alenoach (talk) 20:57, 3 October 2025 (UTC)

An issue in your edits is that most of it is unsourced and the added sources are not explicitly related to the concept of superintelligence. Thus, your paragraph risks being quickly removed (see Wikipedia:No original research). That being said, the result on rats, which is sourced, may be valuable to mention in another article, like Brain transplant. Alenoach (talk) 22:13, 3 October 2025 (UTC)

1) There is some correlation between brain size and intelligence (page itself states that "to evolve larger heads" would make humans smarter; some info is in Brain Size and Evolution of the brain pages; also see there). 2) Implanting neurons into the brain makes brain bigger ( ). 3) Cranioplasty is normally used to repair damaged skulls and replace broken segments of bone - but they can be re-purposed to make skull bigger (like, if you know how to repair a glaring hole in a bottle - then, with sufficient materials, you could make the bottle bigger by making extensions). 4) Medical Exoskeleton or steel gorget/collar can prevent now-embiggened head from breaking the neck (as exoskeletons are used for people who can't walk by themselves). All in combination, may result in extra-smart human, aka superintelligence, since bigger brain is smarter. All those are not sci-fi technologies; rather, it's 4 technologies which are already invented, but not combined yet (just like "tank" is combination of "tractor", "cannon" and "armor" - with tractors being invented in 19th century, yet a thought to rivet armor to it and cram cannons inside only came in The Great War); in other words, it's non-standart use of already existing technologies. The sources have relation to those 4 techs when taken separately, and the paragraph combines them. --Ejkohojkjkohokjh (talk) 07:55, 6 October 2025 (UTC)

The issue is that in Wikipedia, sourcing is really important. If you come up with an idea for which you can't find reliable sources, adding it to Wikipedia would be original research, which is generally avoided. Even assuming you have a source for each of the four statements, if you connect multiple sources to imply a conclusion that is not supported by any of the sources, it is original synthesis. Note also that links to other Wikipedia articles are not considered sources.

If you want to add content on biological superintelligence, it is better to start by searching for reliable sources on this topic. Then you may synthesize the content of such sources into the Wikipedia article to fill coverage gaps. If you don't find content on biological superintelligence in reliable sources, maybe you would be interested in contributing to other articles, like intelligence amplification or Brain transplant, for which it may be easier to find reliable sources. Alenoach (talk) 18:55, 6 October 2025 (UTC)

There's one source: Existential_risk_from_artificial_intelligence#Comparison_with_humans. "Scalability: human intelligence is limited by the size and structure of the brain, and by the efficiency of social communication, while AI may be able to scale by simply adding more hardware.". This means that more brains/computronium=smarter. Now, there are two methods of how more hardware can be attached to human brain. First method is connecting human brains into Group mind, so they all work like one big brain and social communication is no longer an issue. Second method is grafting additional neurons onto brain; it worked with mice many times, so it may work with humans and be upscaled to the point that grafted neurons outnumber the original neurons.. --~2026-10633-46 (talk) 20:04, 17 February 2026 (UTC)

There is a subchapter of Bostrom's Superintelligence book on brain-computer interfaces, but he says that the brain's ability to make sense of the incoming data is a bottleneck and that it's much easier to create artificial superintelligence. He says that human language is already a pretty efficient way to output information in a way that other brains can interpret, and so the potential benefits of brain interfaces are limited. Alenoach (talk) 02:46, 18 February 2026 (UTC)

You overestimate the speed and coherency of human language. Verbal machine language, such as Gibberlink, is 80% faster than human languages. If human was modified to talk in Gibberlink - via bionic vocabulator, translator connected to brain, possibly bionic ears or modified vocal chords or just special training - he could communicate much faster. --~2026-11142-27 (talk) 11:47, 19 February 2026 (UTC)

Data transmission though wires or wi-fi is orders of magnitude faster still. Additionally, direct transfer of information via wires is easier to understand: humans have to turn their idea into words, words travel from one human to other, other human interprets words back into idea - and not only each step takes time, but adds distortions into process, which results in misunderstanding. Ability to directly transfer ideas "as is" via brain interface to other members of Group mind, without converting them back-and-forth, would make it much easier to communicate accurately. --~2026-11142-27 (talk) 11:45, 19 February 2026 (UTC)

Also, on our very own page: "Another advantage of computers is modularity, that is, their size or computational capacity can be increased. A non-human (or modified human) brain could become much larger than a present-day human brain, like many supercomputers. Bostrom also raises the possibility of collective superintelligence: a large enough number of separate reasoning systems, if they communicated and coordinated well enough, could act in aggregate with far greater capabilities than any sub-agent.". This acknowledges both that size matters and that brains can be connected into single network. --~2026-10633-46 (talk) 20:18, 17 February 2026 (UTC)

Finally, page says nothing about threats of biological superintelligence to humanity. Consider that the entire origin of such creature, no matter which method is taken, is gross violation of many laws: human with modified genes (Illegal), who's selectively bred (Eugenics), high on nootropics (Drug Abuse), with 280 or so whole Raven brains surgically grafted onto his brain (Animal Cruelty, nonsensically extreme Medical malpractice on level of Team Fortress 2 Medic), Cranioplasty re-purposed to extend brains so all those grafts fits (Medical malpractice), steel collar transferring weight so head doesn't fall off (unethical, as man is being turned into cripple), with neural interface connecting him to computers (civil rights error, "is he still human or not") and other humans (ethical problems of Group mind). So, nearly all methods used to create biological superintelligence are different degrees of "unethical"; in other words, the very researchers creating such superintelligence are misaligned. And these researchers (under command of equally misaligned CEOs and oligarchs who only think about money), are the very people who train the superintelligent human - causing him to become misaligned too. --~2026-10633-46 (talk) 20:53, 17 February 2026 (UTC)

There's also question of normal humans and superintelligent beings would react to one-another - but if something, related yet different species typically view each-other as prey or concurrents. E.G. Mice and rats usually don't coexist, as rats eat mice, and mice absolutely fear rats and move out of the area; keep in mind that rats are a lot bigger and smarter than mice. If extrapolating that, then human baseliners would view hulking superintelligent transhumans with primal fear, while said superintelligent transhumans would scold or outright maul "inferior" baseliners; unless brain can be made smarter without making it bigger (e.g. Raven has very high density of neurons), smarter = bigger brain = bigger body, que huge size as of Space marine. If we take superintelligent AGIs, then this gets even more unpredictable. --~2026-11142-27 (talk) 15:11, 19 February 2026 (UTC)

Human has arithmetic mean of average number of neurons of 18670000000, chimpanzee has average of 7400000000 neurons; 2.522(972) times difference. Human has 1,201 cm³ brain volume, Chimpanzee has 400 cm³ brain volume; 3.0025 times difference. So as a rule of thumb, the lower and upper borders of what would be considered to be "equal" to you in intelligence are less than 2.5-3 times difference; anything 2.5-3 or more times stupider than you is viewed as non-sapient, and anything 2.5-3 or more times smarter than you is superintelligent. E.G transhuman with 63000000000 neurons and brain volume of 3600 cm³ would be to human, what human is to chimp. This unit of measure, called "Ape", can be useful for napkin math. --~2026-11142-27 (talk) 20:09, 11 March 2026 (UTC)

The worst part (which needs to be covered on page; "Economic Problem" or something) is that both baseliners and superintelligents/AGIs would compete for food. Superintelligent beings know a lot from many professions, are much more competent than any human; AGIs can directly connect to computers, and never tire or sleep - work 24/7 without breaks and weekends. As such, superintelligents and AGIs are much better at work than baseliners, being able to fulfill multiple baseliner jobs simultaneously. By Supply and demand, nobody would want to hire a baseliner, unless his cost is decreased; thus meaning that baseliners would have lower wages with longer work shifts. Government would try to give benefits to baseliners or force companies to hire baseliners, but that in turn would offend superintelligents and the rich. Essentially, baseliners would think of superintelligents and AGIs as them "stealing our jobs and starving us to death" - while superintelligents and AGIs would threat baseliners as "weak, worthless parasites that sap our budget". And on top of that, superintelligents and AGIs have more money and far superior technology - so if baseliners rebel, then no matter how numerous, they get blown to pieces (like Zulu were); so the only choices are to adapt or flee. --~2026-11142-27 (talk) 15:11, 19 February 2026 (UTC)

If you want to add content, you need a reliable source that clearly supports the point you are trying to make. Please also beware of original synthesis. Alenoach (talk) 19:36, 19 February 2026 (UTC)

Cleanup?

Majority of the page is written, as if Superintelligence, Recursive Self-Improvement and other related topics only apply to AI's. Parts about super-smart biological intelligences are either absent or afterthought. As such, i think page needs some re-writing. --95.167.183.86 (talk) 15:02, 1 September 2025 (UTC)

Do you have a reliable source for these topics? What is missing from the current description in the article? Elestrophe (talk) 18:10, 1 September 2025 (UTC)

In current description in the article, most information about biological superintelligences is missing. Such as how to make it, what consequences it could have, is it dangerous, how biological superintelligence would differ from artificial superintelligence in practical standpoint, etc. Most phrases are worded, as if superintelligence and related terms could only apply to AI's. It's also missing information about group mind variants. --95.167.183.86 (talk) 13:51, 2 September 2025 (UTC)

Likely AI-generated content

Hi -- I added the AI-generated tag here, specifically in reference to the 2024 edits by Orexin, who has been blocked for adding poorly reviewed AI text to Wikipedia. The text here, much of which still remains, contains a lot of strong indicators of LLM use, and needs review; issues have been pointed out above and elsewhere. Gnomingstuff (talk) 17:14, 8 September 2025 (UTC)

Here are the edits. The edits honestly aren't bad except for the formatting, and unlike with many of Orexin's other edits, the references really existed and were relevant. I reviewed it at the time, and it seems good in terms of accuracy. But yes, pretty sure it's AI-generated, and it really overused lists. What should we do about it? Alenoach (talk) 21:47, 8 September 2025 (UTC)

If the text is an accurate description of the sources, the main problems are the "Advantages," "Challenges," etc. sections, which are not well sourced and full of weasel words, and various pieces of synthesis/OR like These examples highlight the potential for catastrophic outcomes even when an ASI is not explicitly designed to be harmful, underscoring the critical importance of precise goal specification and alignment.. Especially questionable in this case because the whole AGI extinction risks thing is... not somewhere where there is agreement on what is "critically important." Gnomingstuff (talk) 17:17, 12 September 2025 (UTC)

Hi @Aidy P. I saw that you made a commit trying to remove AI-generated content. For your information, here are the commits that motivated the "AI-generated" template. I have verified the content, it looks decently sourced and accurate. But there are remaining issues with tone and overuse of lists. Alenoach (talk) 19:56, 18 November 2025 (UTC)

Thanks :) Aidy P (talk) 22:14, 18 November 2025 (UTC)

“a stark example”, “another illustrative scenario”, “These examples highlight the potential for catastrophic outcomes even when an ASI is not explicitly designed to be harmful, underscoring the critical importance of precise goal specification and alignment.”, “emphasizing the need for extreme caution in ASI development”. The excessive use of lists were the items are one sentence long, the “superficial analysis” verbs and the overuse of present participle clauses at the end of sentences are MASSIVE AI-generated telltales. 100% at least partially AI-generated. I say a good portion of the article needs to be entirely rewritten. Aappalajaarpunga (talk) 16:06, 1 October 2025 (UTC)

This stark example of a case underscores the unintelligence of AIs that delve our site, and therefore highlights the critical caution of extreme importance of using human-generated texts in Wikipedia development. --~2026-11142-27 (talk) 17:44, 11 March 2026 (UTC)