Gary Marcus: Limits of Deep Learning | AI Podcast Clips

preview_player
Показать описание
(more links below)

Podcast full episodes playlist:

Podcasts clips playlist:

Podcast website:

Podcast on iTunes:

Podcast on Spotify:

Podcast RSS:

Note: I select clips with insights from these much longer conversation with the hope of helping make these ideas more accessible and discoverable. Ultimately, this podcast is a small side hobby for me with the goal of sharing and discussing ideas. For now, I post a few clips every Tue & Fri. I did a poll and 92% of people either liked or loved the posting of daily clips, 2% were indifferent, and 6% hated it, some suggesting that I post them on a separate YouTube channel. I hear the 6% and partially agree, so am torn about the whole thing. I tried creating a separate clips channel but the YouTube algorithm makes it very difficult for that channel to grow unless the main channel is already very popular. So for a little while, I'll keep posting clips on the main channel. I ask for your patience and to see these clips as supporting the dissemination of knowledge contained in nuanced discussion. If you enjoy it, consider subscribing, sharing, and commenting.

Gary Marcus is a professor emeritus at NYU, founder of Robust.AI and Geometric Intelligence, the latter is a machine learning company acquired by Uber in 2016. He is the author of several books on natural and artificial intelligence, including his new book Rebooting AI: Building Machines We Can Trust. Gary has been a critical voice highlighting the limits of deep learning and discussing the challenges before the AI community that must be solved in order to achieve artificial general intelligence.

Subscribe to this YouTube channel or connect on:
Рекомендации по теме
Комментарии
Автор

(more links below)

Podcast full episodes playlist:

Podcasts clips playlist:

Podcast website:

Podcast on iTunes:

Podcast on Spotify:

Podcast RSS:

lexfridman
Автор

This guy makes me appreciate human thinking abilities.

miked
Автор

On behalf of myself and international hearing disabled listeners: Very good recordings, professional microphones and enough light for interpreting by lips movement. The accent, though, may be easily understandable for an american listener, but to me many of the vocal sounds are produced in such a way that the words are hard to separate from each other. I'm trying hard to understand by repeated listening and unfortunately there are no subtitles or automatic captions. I would love a transcript! So many thanks for these series! Greetings from Norway!

tomhummel
Автор

Wondering what are his thoughts on GPT-3.

kalp
Автор

Deep learning: What is this object?
Gary Marcus' learning: What does the object do? 🤔

seasong
Автор

How does him know that symbolic abstractions cannot be built out of correlations and deep learning? We are far from perfect at symbolic abstractions, we make errors, which to me suggests we don't have symbolic abstractions "built" in our brains if not through something similar to what ANN do

MnulF
Автор

Gary Marcus reminds me of the meme: "He's right you know". Enfant terrible of the deep learning community, not himself a contributor to the state of the art or a practitioner, an outsider poking his nose where it doesn't belong... and yet Marcus (and Pearl, Pinker, Chomsky, et al) makes points that cannot be avoided.

snippletrap
Автор

I felt happy watching this and I don't know why

ahmedrebei
Автор

Unsupervised learning is much better than human input. Nothing to do with laziness. That's how AlphaGo worked.

Cool-gkmc
Автор

My two cents: Coming from symbolic processing of cognitive psychology and criticizing current deep neutral network architectures is a bit unfair. IMHO development of concepts with DNNs can be seen in the systems playing Dota or StarCraft. A fair common ground would be neuroscience, when we can describe how human concepts and symbols are formed with biological neurons we could have a blueprint for AGI. Vica versa is also possible: Look for biological correlates with machines that play Dota, Go or StarCraft.

DrFerencAcs
Автор

Would have been nice to ask him his opinion on where autonomous driving is within these limits

jeremylink
Автор

He is asking for abstraction of new phenomena without prior knowledge, because is a "like human intelligence thing". Because you know, when someone know archery, that should be translatable to firing guns, it's just that i replaced the arrow for a bullet, and the bow for a gun. "Common weapon firing sense", minimal or no training needed.


Also his example of odd and even numbers is very telling, number systems in humanity are fairly recent, parity as an exact definition is even more recent. So where those other cultures/people in history lacking "common knowledge" .

jotatsu
Автор

What your thoughts on self aware robotic arm created by Colombia university PhD student
Guys please respond

avadhutd
Автор

Lex: « This podcast is a small side hobby for me ».
I wonder what will happen when it’ll be just a small hobby or just a side one... ;-)

michelinstarschallenger
Автор

In order to make the mechanical-man, you must presuppose man as mechanism.

In order to make the computer-man, you must presuppose man as computer.

Both ideas are false.

irlserver
Автор

This conversation didn't age well.

palealeable