Yann LeCun: Self-Supervised Learning Explained | Lex Fridman Podcast Clips
Ғылым және технология
Lex Fridman Podcast full episode: • Yann LeCun: Dark Matte...
Please support this podcast by checking out our sponsors:
- Public Goods: publicgoods.com/lex and use code LEX to get $15 off
- Indeed: indeed.com/lex to get $75 credit
- ROKA: roka.com/ and use code LEX to get 20% off your first order
- NetSuite: netsuite.com/lex to get free product tour
- Magic Spoon: magicspoon.com/lex and use code LEX to get $5 off
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:
Podcast website: lexfridman.com/podcast
Apple Podcasts: apple.co/2lwqZIr
Spotify: spoti.fi/2nEwCF8
RSS: lexfridman.com/feed/podcast/
Full episodes playlist: • Lex Fridman Podcast
Clips playlist: • Lex Fridman Podcast Clips
SOCIAL:
- Twitter: / lexfridman
- LinkedIn: / lexfridman
- Facebook: / lexfridman
- Instagram: / lexfridman
- Medium: / lexfridman
- Reddit: / lexfridman
- Support on Patreon: / lexfridman
Пікірлер: 38
aw man where is the rest?!? such a clear explanation so far of the hard problem. i want more!
@Jfantab
10 ай бұрын
in the description, you can find the link to the full podcast episode
That is because images, videos contain way more information than text. If you have trained to draw, you'll definitely know how big the gap is between description and the final image you are gonna draw.
@eelcohoogendoorn8044
2 жыл бұрын
Not clear that this should make it easier rather than harder though. Frankly I am a bit surprised that video hasnt made more progress on this front i the past few years; a shame the clip cuts there.. now im going to have to watch the whole episode to see what more yann had to say on the topic.
If we cannot create an AGI then can we use the human brain circuitry itself to generate a NN with human comparable intelligence?
what about Pulsechain
We also have several quantum dimensions that our brain uses. Do you dimensionally train AI?
DOJO will be a paradigm shift. Maye the force
We don’t have true L5 driving is because we are stuck on old concepts of how the neuron works and how the brain works. Once we stop thinking this way and open our minds, it will come overnight.
We have self driving horses but we don't have self driving cars.🤣
@thomasgilson6206
Жыл бұрын
Horses have 5 billion years of self-supervised training. (Evolution).
Lex is so smart he is half asleep explaining everything the professional is saying but in a more simple and smother sense.
They need to eject the lithium electro-ceramic bio-filter. should eject the subatomic vacuum node.
This is the best.
Humans learn while sleeping. Maybe we just run a lot of examples in our head 🤔
@lukassteindl1914
Жыл бұрын
think so too but you dont let it run 1 million times. when you dream, it feels more like a single realtime replay per night, so it must be much more efficient than reinforcement learning type episode replay. and sometimes the mental video shows strange things that are unrealistic. might be some calibration / regularization going on while one sleeps.
@christopherprobst-ranly6357
Жыл бұрын
@@lukassteindl1914 Well, .. remember Inception and the in-sleep time ;-)
TALK ABOUT SAFEMOON!!!!!!!!!!!
I don’t see consciousness as mysterious. We are just a very complicated quantum machine and consciousness was just the next logical step in evolution. Being aware of the past and future. I think this is the self learning mechanism which is the key.
@Saad-xo6yc
2 жыл бұрын
lol
@linkin543210
2 жыл бұрын
"quantum machine"...are you a marketer ? 😂
@nicktheodorou3474
2 жыл бұрын
@@linkin543210 Lol! The brain and body is composed of atoms, and atoms follow the laws of quantum physics 😉. But probably not the best way to refer to something like cognition.
@linkin543210
2 жыл бұрын
@@nicktheodorou3474 I'm sorry I was a bit rude, and I totally agree that conciousness was the next level in evolution to survive.
I'm supposed to believe this guy is a genius? He spits out like 4 words a minute.
@thelonespeaker
2 жыл бұрын
Both speak at a reasonable pace YLC is French btw
@JohnSmith-bb1sv
Жыл бұрын
I don't mean Lex any harm but he is not a genius not by a long shot. He's slow, he asks dumb questions, and he is obtuse. His interview with Vladimir Vapnik was what did it for me. The interviewer needs to move at the same intellectual pace or faster than the audience. Machine Learning Street Talk is a much better podcast and they interviewed Prof LeCun twice.
All you need to do us create a system with desire, as soon as an AI actually WANTS to learn something it will need to be programmed and therefore can never be as smart as us.
Man with the specs, talking about meta learning, meta learning is different from self supervised learning
Why don't you make longer vid???too short not wordy of my time...
@samosa9488
2 жыл бұрын
Watch the full interview on his other channel. Lex fridman
@artpinguin1176
2 жыл бұрын
@@samosa9488 Okey thnx friend
Building world model is done trough an interacting physical body. To duplicate our intelligence I beleive that for training we will need to provide a phyical body with same sensors. A humanoïde robot
Why does lex sound so sleepy and boring? Can't he make an effort and learn to articulate and project some energy and enthusiasm?
Everything about this channel is a JRE rip-off.