Post
"... [Yann] LeCun has been saying ... what the industry doesn't want to hear: Large Language Models are a dead end on the path to human-level intelligence. They are, at their core, text-prediction engines, extraordinarily good at retrieving, recombining, and generating language, but fundamentally incapable of understanding the world they talk about. They lack common sense, causal reasoning, and any model of physical reality."
#DavidWilliamSilva, 2026
@strypey its great that he is finally saying this … but it is a shame that only a few years ago he was attacking people who were saying the same thing - see, eg gary marcus - https://garymarcus.substack.com/p/the-false-glorification-of-yann-lecun . scientists are allowed/encouraged to change their minds when new facts/evidence comes to light; but he owes an apology to the people he once criticised, and whose message he has appropriated for his own personal aims. he is a giant in the field so its important he acts ‘ethically’
Apologies for the SS link, but I thought it was worth it ; )
"There is a meme in Brazil that goes:
'How do you manage to do so much in a single day?'
'I do everything very poorly.'
That is #AI in one exchange.
AI does incredible things, very fast, in the sloppiest possible way. It is the most productive intern you’ve ever had, one who never sleeps, never complains, and many times fails to check whether the work is actually correct.
And we are building entire industries on top of this."
#DavidWilliamSilva, 2026
There’s a joke I tell my kids sometimes:
I can do multiplication very quickly. Give me a 4 digit number:
8437
And another 4 digit number:
3201
That’s 12
See how fast that was? Amazing
@strypey That being said, they can do a variety of useful things better than conventional search techniques.
@mike805
> they can do a variety of useful things better than conventional search techniques
Before I can tell whether I agree with this or not, you'll need to give some details and maybe some examples.
