Sign up to see the future, today<\/p>\n
Can\u2019t-miss innovations from the bleeding edge of science and tech<\/p>\n
Outspoken venture capitalist and major Trump backer Marc Andreessen, whose controversial \u201ctechno-optimist manifesto\u201d in 2023 set the tone for a years-long AI boom cycle, doesn\u2019t appear to know how the tech actually works.<\/p>\n
In a lengthy \u201ccustom prompt\u201d he shared in a Monday tweet<\/a>, the billionaire seemingly tried to show off his AI skills \u2014 only for the internet to mercilessly mock him for it.<\/p>\n \u201cYou are a world class expert in all domains,\u201d his unusually flattering prompt reads. \u201cYour intellectual firepower, scope of knowledge, incisive thought process, and level of erudition are on par with the smartest people in the world.\u201d<\/p>\n His gushing and unusually naive tone drew ridicule, but what really got the internet laughing was his insistence that the AI should \u201cnever hallucinate or make anything up.\u201d<\/p>\n The fact is that simply requesting a large language model to not lie doesn\u2019t work; so-called \u201challucinations\u201d are an underlying issue with the technology, not a self-esteem problem for chatbots that can be overcome by sufficiently glazing them up. And the implication otherwise suggests that Andreessen has a surprisingly shaky grasp of the tech\u2019s core functionality.<\/p>\n \u201cYes, you can just demand that the LLM not make errors,\u201d journalist Karl Bode joked in a mocking Bluesky post<\/a>. \u201cThat\u2019s definitely how the technology works.\u201d<\/p>\n \u201cI know this isn\u2019t a unique observation but these gentlemen are in absolutely no way remarkable outside of their good fortune,\u201d Bode added.<\/p>\n \u201cMarc Andreessen putting \u2018you are a world class expert in all domains\u2019 and \u2018don\u2019t hallucinate\u2019 in his custom prompt really demonstrates the calibre of the people steering the ship,\u201d another user added<\/a>.<\/p>\n In a scathing piece for Defector<\/a>, editor Alberto Burneko similarly argued that Andreessen\u2019s eyebrow-raising AI tinkering were signs of \u201cAI psychosis<\/a>,\u201d the phenomenon that has some users being sent on potentially dangerous spirals of their own delusions.<\/p>\n Burneko pointed out that the chatbot can\u2019t \u201cthink,\u201d or \u201cjudge anything,\u201d let alone \u201cunderstand these instructions.\u201d<\/p>\n \u201cYou can\u2019t make an AI chatbot know everything in the world by telling it to know everything in the world,\u201d he wrote in his piece. \u201cEven if it\u00a0could\u00a0know things (it can\u2019t), the limits of its knowledge were not theretofore bounded by an understanding (another thing it can\u2019t have) that it only had to know\u00a0some\u00a0stuff.\u201d<\/p>\n Burneko makes a compelling argument. By carefully laying out his own biased world view \u2014 Andreessen specifically asks the chatbot to ignore \u201cmorals and ethics\u201d and not be \u201cpolitically correct\u201d \u2014 the billionaire is tapping into his own skewed perspective before even receiving any output from the AI.<\/p>\n In short, it\u2019s a fascinating yet sobering glimpse into how highly influential and obscenely rich public figures who are pushing the tech into every public domain have a crude grasp of how it actually works \u2014 and, just possibly, what\u2019s going on with them psychologically.<\/p>\n \u201cIn trying to tell the chatbot not to hallucinate, he is scripting his own psychotic break,\u201d Burneko concluded. \u201cHe is doing it because he is a huge dumbass. Don\u2019t expect Claude to tell him so.\u201d<\/p>\n