Is AI just statistics? | Yann LeCun and Lex Fridman

Please support this podcast by checking out our sponsors:

GUEST BIO:
Yann LeCun is the Chief AI Scientist at Meta, professor at NYU, Turing Award winner, and one of the seminal researchers in the history of machine learning.

PODCAST INFO:

SOCIAL:

[parthokr]

Loved his straightforward answer. "yes".

[midnight]

Hi Lex, not sure if you read these comments. But I have been listening to your podcasts for a while now and you have inspired me to learn AI/ML at MIT no less! I come from a medical background. It’s been difficult and rewarding at the same time. Thank you for inspiring me and pushing me out of my comfort zone.

[rkalla]

I appreciate the fast answer - no fumbling around an ego-validating answer.

[iorekby]

"I have made no effort to understand what AI is so I'll insert a faux-profound comment about ethics or Skynet reference because everyone needs to know what I think about it anyway"

Basically 99% of the comments on YT about AI. Sadly a lot of them have already started on this video's comment section too I see.

[joao_aguilera]

Man, as a data engineer and scientist, i can say: Your service for the world will not be forgotten. Watching your show is own if the things I most enjoy in life!

[dntinpalevo]

Finally a guest said to Lex the right comment: those are too many questions!

Lex I love you and all the things you do BUT for the love of God pick one question at a time and keep it short and simple!

On all your interviews I have noticed that your guests always answer the very last question you ask, and it's a shame because the previous questions were so good but got wasted because of your rambling.

[NeilS.]

If I dare throwing my two cents in the pool of ideas, I would say that the essence of intelligence is built around the logical axiom of survival. A living being would try to do anything in any circumstances in order to survive. It is that core parameter that we need to emulate in order for a neural network to "evolve" into something that could create its on solutions on any reality presented to it. I would add that the axiom of survival is also deeply connected to WHY a living being needs to survive. So that's why I think AI research is such an interesting topic for both mathematics, engineering and philosophy.

[SMAKSHADE]

bless up lex i enjoy your show big time love from canada

[westganton]

Life is like a model of models that have been trained over countless iterations and under countless conditions for the sole purpose of survival. It's arrogant to think that we can out-engineer life while we have so much left to learn about ourselves

[tonechild]

yes, artificial/transistor-based neural nets are just using statistics, unlike organic neural nets. There's plenty of information on this, AND this is taught in ML / AI classes. In fact the term coined "neural net" by AI researchers is hugely misleading.

[Flamingpiano]

AI is statistics guided by a set of designed goals, it's internal weighting of it's statistics comes from what it is trying to achieve.

[mazu]

Everything has a cause and even if there is no cause of the existence of the universe that is also a cause, you cannot have effects without a cause.

[Drakyry]

"Is physics just statistics?"
"Is reality just statistics?"

[S.G.Wallner]

I'm Focusing on one phrase Lex invokes like many other neuroscientists or computer scientists, "when we look at the brain." One subtle point to take very seriously is that when we "look at the brain, " we aren't accurately seeing what the brain is actually doing. Set aside the fact that even our best cutting edge brain imagining is still a simplified and highly processed representation of one aspect of brain activity, but more importantly even if we could hypothetically see what the brain is doing in totality, our perception of that brain activity is still not necessarily what the brain is doing in and of itself. Our observation of brain activity can not extend beyond our limited and fundamentally subjective perception of it. There is no direct evidence that brain activity causes consciousness or intelligence.

[AlexanderMoen]

You could probably say that evolution is entirely statistics playing out. And, perhaps, once we evolved to the level of developing consciousness that it just took everything to a higher level. We're thinking about numerous scenarios without actually going through them, then implementing, then recording the results and adjusting. And, AI can do this even more quickly and in parallel. Super interesting thought I'll have to ponder over more.

[intheshellify]

My god man are you trying to end the world? Who would make an ai based on cat brain? We already have serial killers.

[trukxelf]

My cat is also excited about neural cat networks and AI

[ruffyistderhammer]

I agree with trying to reach cat level reasoning first its probably easier and would show a lot to reach human level

[eSKAone-]

To get something like a digital human I don't think we have to go the cat route. I don't think cat's are lesser humans, they are just different. Every animal is specialized.

[andrewcutler]

From his comments about current AI not even coming close to a cat, I gather he thinks projects to protect us from AGI are silly (such as OpenAI).