An interview with Meta Chief AI Scientist Yann LeCun on his critics and why today's most popular approaches to AI won't lead to human-level machine intelligence
Oh, it's unfair for him to dismiss @GaryMarcus as not being an AI person. That's just like Descartes dismissing Fermat or Newton dismissing Leibniz because they were not formally schooled in mathematics. Why should DL people have a monopoly on the conversation?
@ZDNET @TiernanRayTech This libels me, by claiming—falsely—that I have not published any AI work in peer review journals & at same time the substance is virtually identical to my 2012 & 2018 papers, not in any way original to @yLeCun; a thread summarizing some of the many similar…
“The purely statistical approach is intractable. It's too much to ask for a world model to be completely probabilistic.” “Ultimately, there's going to be a more satisfying solution that do a better job of understanding the way the world works.” https://www.zdnet.com/...
This part of the interview with LeCun is so painful, so olympically painful and abominable: (My next course with ~80 new, first-year #ComputerScience students starts in 3 weeks. I'll be very explicit and teach them what *not* to do in academic discourse or writing) https://twitte…
I find it curious that LeCun leans toward “energy-based” architectures, yet he cites Lagrangians as the basis of Deep Learning. Aren't energy-based formulations a Hamiltonian perspective and not Lagrangian? @ylecun https://www.zdnet.com/...
I found explanations for LeCun's behavior in @Eric_Sadin's “The Silicolonization of the World” (2016/2018). See pic & translation. @ZDNET's interview only reinforces this and leaves him in a terrible position. The audacity... How LeCun refers to other academics and their work 😳 h…
If you simply believe today's deep learning will scale to infinity, you would do well to consider the sage observations of a pioneer, Yann LeCun. // @ylecun // #AI #deeplearning #machinelearning #artificialintelligence https://twitter.com/...
@mustafasuleymn I am always surprised by the incomprehensibly large number of people who believe the myth that human-level AI will be achieved through one key idea, by one group of people, in one lab/company, that will stay ahead of the pack for years. (and make trillions). It's …
In summary, it's a good interview where you can get a glimpse of @ylecun thinking. I'm agreement with many of his ideas, except for the criticism of the language models. I suspect he hasn't read about “code-duality” yet.
just wow. @MetaAI's @ylecun retrenches to steals pretty much everything I have projected about limits of deep learning from last decade and then says I have never done anything. that is quite the market repositioning. https://twitter.com/...
@ylecun @mustafasuleymn This is a bit like the idea from the past that space travel would be invented by one solitary genius building a complete re-usable rocket ship by himself
It's fascinating to hear a father of AI continue to hammer on the need for better observational learning (rather than only experiment) as the key to progress, which of course mirrors debates in biomedicine. The ladder to the moon analogy is great. https://www.zdnet.com/...