One of the most important AI scientists in Big Tech wants to scrap the current approach to building human-level AI. What we need, Yann LeCun has indicated, are not large language models, but \u201cworld models.\u201d<\/p>\n
LeCun, chief AI scientist of \u201cfundamental AI research\u201d at Meta, is expected to resign from Meta soon according to<\/a> multiple reports<\/a> from credible outlets. LeCun is a 65-year-old elder statesman in the world of AI science, and he has had seemingly limitless resources at his disposal working as the big AI brain at one of the world\u2019s largest tech companies. <\/p>\n Why is he leaving a company that\u2019s been spending lavishly, poaching the most highly-skilled AI experts<\/a> from other firms, and, according to a July blog post<\/a> by CEO Mark Zuckerburg, making such astonishing leaps in-house that supposedly the development of \u201csuperintelligence is now in sight\u201d?<\/p>\n He\u2019s actually been hinting at the answer for a long time. When it comes to human-level intelligence, LeCun has become notorious lately for saying LLMs as we currently understand them are duds\u2014no longer worth pursuing, no matter how much Big Tech scales them up. He said in April of last year<\/a> that \u201can LLM is basically an off-ramp, a distraction, a dead end.\u201d (The arch AI critic Gary Marcus has ripped into LeCun<\/a> for \u201cbelligerently\u201d defending LLMs from Marcus\u2019 own critiques and then flip-flopping.)<\/p>\n A Wall Street Journal analysis of LeCun\u2019s career<\/a> published Friday points to some other possibilities about the reasons for his departure in light of this belief. This past summer, a 28-year-old named Alexandr Wang\u2014the co-creator of the LLM-based sensation ChatGPT\u2014became the head of AI at Meta, making an upstart LLM fanatic LeCun\u2019s boss. And Meta brought in another relatively young chief scientist to work above LeCun this year, Shengjia Zhao. Meta\u2019s announcement of Zhao\u2019s new role touts<\/a> a scaling \u201cbreakthrough\u201d he apparently delivered. LeCun says he has lost faith in scaling<\/a>.<\/p>\n If you\u2019re wondering how LeCun can be a chief scientist if Zhao is also a chief scientist, it\u2019s because Meta\u2019s AI operation sounds like it has an eccentric org chart,<\/a> split into multiple, separate groups.<\/a> Hundreds of people were laid off last month<\/a>, apparently in an effort to straighten all this out.<\/p>\n The Financial Times\u2019 report<\/a> on LeCun from earlier this week suggests that LeCun will now found a startup focused on \u201cworld models.\u201d\u00a0<\/p>\n Again, LeCun has not been shy about why he thinks world models have the answers AI needs. He gave a detailed speech about this<\/a> at the AI Action Summit in Paris back in February, but it got kind of overshadowed by the U.S. representative, Vice President J.D. Vance<\/a>, giving a bellicose speech about how everyone had better get out of America\u2019s way on AI.\u00a0<\/p>\n Why Is Yann LeCun fascinated by world models? <\/p>\n As spelled out in his speech\u2014LeCun, who worked on the Meta AI smart glasses, but not to a significant degree on Meta\u2019s Llama LLM<\/a>\u2014is a huge believer in wearables.<\/p>\n Amazing how the Ray-Ban Meta glasses can help the visually impaired. https:\/\/t.co\/w3ZxCFtTlE<\/a><\/p>\n \u2014 Yann LeCun (@ylecun) September 30, 2024<\/a><\/p><\/blockquote>\n We\u2019ll need to interact with future wearables as if they are people, he thinks, and LLMs simply don\u2019t understand the world like people do. With LLMs, he says, \u201cwe can\u2019t even reproduce cat intelligence or rat intelligence, let alone dog intelligence. They can do amazing feats. They understand the physical world. Any housecat can plan very highly complex actions. And they have causal models of the world.\u201d\u00a0<\/p>\n LeCun provides a thought experiment to illustrate what he thinks might prompt\u2014if you will\u2014a world model, and it\u2019s something he thinks any human can easily do that an LLM simply cannot:\u00a0<\/p>\n \u201cIf I tell you \u2018imagine a cube floating in the air in front of you. Okay now rotate this cube by 90 degrees around a vertical axis. What does it look like?\u2019 It\u2019s very easy for you to kind of have this mental model of a cube rotating.\u201d\u00a0\u00a0<\/p>\n<\/blockquote>\n With very little effort, an LLM can write a dirty limerick about a hovering, rotating cube, sure, but it can\u2019t really help you interact with one. LeCun avers that this is because of a difference between text data and data derived from processing the many parts of the world that aren\u2019t text. While LLMs are trained on an amount of text it would take 450,000 years to read, LeCun says, a four-year-old child who has been awake for 16,000 hours has processed, with their eyes or by touching, 1.4 x 10^14bytes of sensory data about the world, which he says is more than an LLM.<\/p>\n These, by the way, are just the estimates LeCun gives in his speech, and it should be noted that he has given others<\/a>. The abstraction the numbers are pointing to, however, is that LLMs are limited in ways that LeCun thinks world models would not be.\u00a0<\/p>\n What model does LeCun want to build, and how will he build it? <\/p>\n\n
\n