Season 2 Ep 23 Twitter Q&A with Geoff Hinton

[@karnikram]

0:00 Intro and season wrap-up 3:53 What were some of the hardest times on your path to getting DL to work? 5:14 Are you concerned that AI is becoming too successful? 6:15 Is DL hitting a wall? Can we achieve AGI with scale? 8:16 What are three questions that keep you up at night? 8:57 Advice for contrarians trying to produce the next AlexNet moment 9:47 What is the connection between mania and genius? 11:34 What childhood experiences shaped you the most? 12:27 Thought process to solve a research problem? 13:35 Next big thing in AI and advice for PhD students? 14:42 Regrets about missed research opportunities? 15:09 How important is embodiment for intelligence? 17:37 Why do you do what you do? 18:09 Curiousity-driven research vs application-driven research 19:04 Research in different settings 20:29 Batch of questions about the brain 33:50 Batch of questions about what is next in AI 35:58 Thoughts about the bitter lesson 36:41 How do you read research papers? 37:36 Role of formalism and intuition in research 39:27 How did you transition from psychology to AI? 41:43 How conscious are today's neural networks? 44:38 Can ML help advance pyschology?

[@TheRobotBrainsPodcast]

Thanks for this breakdown!

[@prabhavkaula9697]

This session was such a treat! Thank you Geoff and Pieter.

[@TheRobotBrainsPodcast]

Thanks for listening!

[@peterdavidfagan]

The explanation for the term consciousness in this talk is really great. It makes a lot of sense that it is a word whose ill definition reflects our own lack of understanding or framework for modelling and measuring it. I wonder if there will be a day in the future where someone explains how we used the word consciousness similar to the cars and oomph explanation?

[@이규빈-q7s]

I've never seen anyone thinking of walking on two legs like him in my life(energy efficient time sharing wheel). amazing.. amazing interview. thank you pieter n geoff!!!

[@hhhghggfh8818]

Ģ è

[@jeeding]

I am considering applying for a related PhD program, this episode was really inspiring, and to express my gratitude, I would like to offer my thanks in 3 languages! Thank you Geoff and Pieter. תודה לך ג'וף ופיטר. 感谢杰夫和彼得。

[@usinani]

Great Q&A session with Geoffry and podcast, thanks for sharing!

[@TheRobotBrainsPodcast]

Glad you enjoyed it!

[@fuchuichui]

非常的鼓舞人心！

[@rb8049]

Spiking is not a requirement, but our brains have figured out how to use spiking. Spike timing can be used for encoding a lot of information in a network. We don’t have to go full analog to use spiking. We can generate spikes with a clocked system. Analog might be more dense and energy efficient, but digital might be just as good. We can design digital systems faster today.

[@jimmylovesyouall]

17:19 , Geoff said: a long time ago David Bxx (a name)... What is this name? Does anyone know? Thanks.

[@cbord4139]

Dana Ballard

[@jimmylovesyouall]

@@cbord4139 Thank you so much. Do you know what paper was Geoff referring to by Dana?

[@cbord4139]

@@jimmylovesyouall not sure, but I assume his 1991 paper entitled "Animate vision"

[@arvisz1871]

Nice episode 👍 it has a lot of umph inside

[@abhinavchauhan7864]

Great stuff pieter🙌🏻. AI is getting really scary but at the same time it gives us hope about a lot of stuff such finding cures of stuff like cancer and brain damage. One suggestion why dont you add translated subtitles to your videos so even the peole who dont understand english can benefit from them and your channel would grow as well

[@iandanforth]

I don't buy Geoff's "humans have wheels" argument, but a really great session, thanks for putting it together!

[@FelixWatts]

Yea. Legs are actually much better than wheels. Wheels need roads. When the first fish came out of the sea, they didn't find a surface conveniently criss-crossed with a network of roads.

[@sivasakthis7087]

HAHAHAHAHAHAHAH the sa # save the turtels

[@mbahsugimin]

Ingin gambar yang telanjang bulat

[@commentselements2534]

comment 1662132484325