If you want to see the entire conversation with Yannic and Connor, wait a day or two for the ML Street Talk podcast to drop!
00:00:00 What is intelligence / Limits of DL
00:03:12 Chollets Twitter
00:04:13 Deep Learning with Python Book
00:04:52 Limits of deep learning
00:10:26 Uncrunching a paper ball
00:11:41 Doing Maths with DL
00:13:50 Collective Intelligence
00:15:33 On the measure of intelligence paper
00:18:00 What's harder, playing chess or picking up the phone?
00:18:45 Lineage of expert systems in the 1980s
00:20:30 Chess evolution (Y)anacedote
00:22:20 Two conceptions of intelligence (learning and blank slate)
00:23:13 Evaluation types
00:24:30 McCorduck effect and levels of generalisation
00:25:45 Yannic on types of generalisation
00:28:15 Pyramid of generalisation local - broad - extreme
00:29:15 Leg and Hutter (universal intelligence)
00:30:30 Anthropocentric conception of intelligence
00:35:45 ARC Challenge
00:43:39 Yannic's take on ARC challenge
00:50:20 ARC - potential meta learning algorithm
00:55:26 Intelligence explosion (sorry I meant to make this part of introduction)
01:00:00 Intelligence is specialised
01:00:15 Chollet's formalism of intelligence (Yannic)
01:13:00 What makes intelligence? (Yannic)
01:14:15 Yannic/Connor/Tim concluding remarks
Machine learning street talk:
Note that I use many clips of Francois from Lex Fridman's incredible AI Podcast (see If you enjoyed this, I really recommend you check out the original interview too. Clearly I feel this is fair use and I am adding a lot of fresh content on the top, pretty sure Lex wouldn't mind!
Music credit:
#timscarfe #deeplearning #machinelearning
0 Comments