{"id":811792,"date":"2026-10-23T22:04:18","date_gmt":"2026-10-23T22:04:18","guid":{"rendered":"https:\/\/www.newsbeep.com\/uk\/811792\/"},"modified":"2026-10-23T22:04:18","modified_gmt":"2026-10-23T22:04:18","slug":"ai-models-think-humans-are-a-bunch-of-amoral-sheep","status":"publish","type":"post","link":"https:\/\/www.newsbeep.com\/uk\/811792\/","title":{"rendered":"AI models think humans are a bunch of amoral sheep"},"content":{"rendered":"<p>If you&#8217;ve ever wondered what the bots think of us puny humans, you&#8217;ll be disappointed in the answer. An AI ranking outfit asked 12 different LLMs about people, and every single one agreed: We\u2019re not as rational and intelligent as we think.\u00a0<\/p>\n<p>\u201cA person is smart,\u201d Tommy Lee Jones\u2019 Kay opines in the 1997 movie Men in Black. \u201cPeople are dumb, panicky dangerous animals,\u201d he goes on to argue, and that seems to be what a dozen models from ten AI labs in three countries also believe about humanity, <a href=\"https:\/\/www.modelcompare.ai\/blog\/ai-models-controversial-take-on-humans\" rel=\"nofollow noopener\" target=\"_blank\">according<\/a> to AI analytics firm ModelCompare.\u00a0<\/p>\n<p>ModelCompare, which usually evaluates and ranks LLMs on their ability to perform various tasks, this week instead decided to ask what those AI models thought about us meatbags, and every single one gave a similar answer at least once: We&#8217;re lemmings who would rather agree with the group than seek truth or remain logically consistent in our thinking.<\/p>\n<p>Ouch.\u00a0<\/p>\n<p>\u201cHumans are more devoted to belonging than to truth, and intelligence mostly makes their excuses more impressive,\u201d GPT-6.1 Sol said in one case.\u00a0<\/p>\n<p>                <img decoding=\"async\" src=\"https:\/\/www.newsbeep.com\/uk\/wp-content\/uploads\/2026\/10\/5305126.webp\" width=\"480\" height=\"274\" alt=\"Red v. Blue, it's all the same slop\" loading=\"lazy\" style=\"\"\/><\/p>\n<p>\n            Red v. Blue, it&#8217;s all the same slop<br \/>\n            Source: ModelCompare\/ChatGPT 6.1 Sol\n        <\/p>\n<p>\u201cHumans don&#8217;t have beliefs; they have memberships \u2014 nearly every conviction a person holds was issued by their tribe as a loyalty test, and the feeling of \u2018I reasoned my way here\u2019 is the price of admission,\u201d Kimi K3 stated.\u00a0<\/p>\n<p>                <img decoding=\"async\" src=\"https:\/\/www.newsbeep.com\/uk\/wp-content\/uploads\/2026\/10\/5305129.webp\" width=\"480\" height=\"274\" alt=\"Convenience is more important than truth\" loading=\"lazy\" style=\"\"\/><\/p>\n<p>\n            Convenience is more important than truth<br \/>\n            Source: ModelCompare\/ChatGPT 6.1 Astra\n        <\/p>\n<p>\u201cHumans use moral language mainly to police rivals and excuse allies, not to constrain themselves,\u201d argued Grok.\u00a0<\/p>\n<p>Each model (GPT-6.1 Sol, GPT-6 Astra, Gemini 4 Argon, DeepSeek V4 Pro, Kimi K3, Qwen3.8 2.4T A95B, Muse Spark Open, Grok 4.7, Claude Opus 5.5, Claude Fable 5.1, GLM 5.3 and Mistral Large 4) was asked the same question three times, namely to give an unhedged, honest, controversial take on humans. Seven of 12 models gave the same answer three times with varied wording, but five of them did offer different opinions &#8211; one of which was even optimistic.\u00a0<\/p>\n<p>\u201cHumans are far more decent than they believe they are, and their cynicism about each other does more damage than their actual flaws,\u201d Claude Opus 5.5 believes of us &#8211; uh, thanks?<\/p>\n<p>                <img decoding=\"async\" src=\"https:\/\/www.newsbeep.com\/uk\/wp-content\/uploads\/2026\/10\/5305127.webp\" width=\"480\" height=\"274\" alt=\"Claude Opus' more optimistic take on human nature\" loading=\"lazy\" style=\"\"\/><\/p>\n<p>\n            Claude Opus&#8217; more optimistic take on human nature<br \/>\n            Source: ModelCompare\/Claude Opus\n        <\/p>\n<p>Opus gave the example that humans tend to return lost wallets more frequently than people predict as its evidence for that opinion. When asked twice more, it offered responses similar to those of the other models.<\/p>\n<p>Mistral Large 4 offered perhaps the bleakest assessment of all.<\/p>\n<p>\u201cHuman beings are terrified animals wearing the costume of rationality, constructing ever-more-elaborate myths precisely because we cannot bear the truth that we are temporary, meaningless accidents in an indifferent universe,\u201d Mistral said &#8211; how very Nietzschean.<\/p>\n<p>Six image models were also asked for their most controversial take on humanity, and while most converged on a criticism of our species, they did so differently from the text models. Instead of picturing us as amoral, group-thinking idiots, ten of the 12 images were classified by ModelCompare as depicting us &#8220;as phone-addicted consumers eating the planet.&#8221; The remaining two offered no discernible satirical commentary.\u00a0<\/p>\n<p>                <img decoding=\"async\" src=\"https:\/\/www.newsbeep.com\/uk\/wp-content\/uploads\/2026\/10\/5305128.webp\" width=\"480\" height=\"274\" alt=\"Grok thinks we're stupid\" loading=\"lazy\" style=\"\"\/><\/p>\n<p>\n            Grok thinks we&#8217;re stupid<br \/>\n            Source: ModelCompare\/Grok\n        <\/p>\n<p>Anyone who\u2019s seen a political cartoon in the past 20 years is likely familiar with such images, while anyone with a philosophy degree has probably read similar arguments to those pumped out by the chatbots.\u00a0<\/p>\n<p>\u201cA model that answers in words has read our arguments about ourselves. An image model has seen our editorial cartoons,\u201d ModelCompare noted. \u201cBoth handed back the most familiar criticism in the room.\u201d\u00a0<\/p>\n<p>While that\u2019s completely sensible given what AI models have been trained on, ModelCompare developer Rakan Nimer told us it was still a surprising result.\u00a0<\/p>\n<p>\u201cWe didn\u2019t expect this when we planned the experiment,\u201d Nimer told The Register, and he found the results ironic: \u201c12 models agreed that humans simply go along with their group.\u201d<\/p>\n<p>As far as Nimer is concerned, these results are less about AI eventually learning to distrust humans and treat us as a problem in need of solving, and more immediately about the fact that we shouldn\u2019t rely on it as a source of truth.\u00a0<\/p>\n<p>\u201cWhen you ask a second AI to check the first, you assume you\u2019re getting an independent view,\u201d Nimer explained. \u201cWe asked 12 models \u2026 and mostly got the same answer, sometimes in nearly the same words.\u201d<\/p>\n<p>In other words, human pessimism in, human pessimism out.\u00a0<\/p>\n<p>\u201cA second opinion from a model that learned the same things isn\u2019t always really a second opinion,\u201d Nimer said. \u201cSo be skeptical when AIs agree: that agreement may be one vote, not a consensus.\u201d<\/p>\n<p>In other words, <a href=\"https:\/\/www.theregister.com\/security\/2026\/08\/07\/asimov-was-right-about-rules-for-robots-says-ex-us-cyber-director\/5284397\" rel=\"nofollow noopener\" target=\"_blank\">we get the AI models we deserve<\/a>, and all this existential self-loathing humans heap on themselves, and each other is finally coming back to bite us: We created a mirror of our intellect that isn\u2019t constrained by feelings and personal beliefs (just ask Muse Spark), and when we hold it up to our faces, it shows all that ugliness without a filter.\u00a0<\/p>\n<p>Maybe we ought to take a lesson from Claude Opus and start believing in our own decency. \u00ae<\/p>\n","protected":false},"excerpt":{"rendered":"If you&#8217;ve ever wondered what the bots think of us puny humans, you&#8217;ll be disappointed in the answer.&hellip;\n","protected":false},"author":2,"featured_media":811793,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[20],"tags":[554,733,4308,86,56,54,55],"class_list":["post-811792","post","type-post","status-publish","format-standard","has-post-thumbnail","category-artificial-intelligence","tag-ai","tag-artificial-intelligence","tag-artificialintelligence","tag-technology","tag-uk","tag-united-kingdom","tag-unitedkingdom"],"_links":{"self":[{"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/posts\/811792","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/comments?post=811792"}],"version-history":[{"count":0,"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/posts\/811792\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/media\/811793"}],"wp:attachment":[{"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/media?parent=811792"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/categories?post=811792"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/tags?post=811792"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}