In this special edition, Dr. Tim Scarfe, Yannic Kilcher and Dr. Keith Duggar speak with Professor Gary Marcus, Dr. Walid Saba and Connor Leahy about GPT-3.
We have all had a significant amount of time to experiment with GPT-3 and show you demos of it in use and the considerations.
Do you think GPT-3 is a step towards AGI? Answer in the comments!
00:00:00 | Connor's take on LinkedIn
00:00:47 | Show teaserĀ
00:20:02 | Tim IntroductionĀ
00:26:55 | First look at GPT-3, python sortingĀ
00:31:05 | Search strategy in LMsĀ
00:38:28 | Character analogies and Melanie MitchellĀ
00:44:27 | Substitution cipherĀ
00:47:21 | Database promptĀ
00:53:00 | Broader Impact GenerationĀ
01:02:47 | Gary Marcus Interview (Robust.AI)Ā
01:29:11 | Connor Leahy Interview (Eleuther.AI)Ā
01:32:29 | Connor -- Tabular dataĀ
01:33:41 | Connor -- other surprising examples?Ā
01:34:54 | Connor -- Is interpolated stuff new?Ā
01:37:43 | Connor -- structure of the brain / How GPT worksĀ
01:41:21 | Connor -- Why cant GPT-3 reason?Ā
01:46:30 | Connor -- Missing information problem and ideas on our our brains workĀ
01:54:28 | Connor -- Topology of brain/modelsĀ
01:58:49 | Connor -- Hardware lottery / LSTM / TransformerĀ
02:01:41 | Connor -- NNs are just matrix program searchĀ
02:10:32 | Connor -- Google -- information retrieval, the new paradigm, how to extract info from GPT-3, RL controller on top?Ā
02:19:38 | Connor -- Database example / "pattern matching is Turing complete"Ā
02:23:55 | Connor -- Did gpt3 understand?Ā
02:26:30 | Connor -- Are the GOFAI people right?Ā
02:27:40 | Walid Saba on GPT-3Ā
02:30:41 | Walid -- What is understanding and pattern recognitionĀ
02:35:56 | Walid -- Chomsky would be happyĀ
02:42:13 | Walid -- Redefining successĀ
02:46:05 | Walid on HintonĀ
02:47:34 | Walid on software 3.0Ā
02:53:11 | Keith -- We use machine learning because we cant write code to do the same thingĀ
02:59:36 | Keith -- What is pattern recognition and understandingĀ
03:14:06 | GPT-3 trials -- Turing DialogĀ
03:15:35 | GPT-3 trials -- Mary Enjoyed a SandwichĀ
03:16:19 | GPT-3 trials -- BBC has five offices in Germany.Ā
03:16:55 | GPT-3 trials -- Database promptĀ
03:20:23 | GPT-3 trials -- PythonĀ
03:20:31 | GPT-3 trials -- PatternsĀ
03:21:01 | GPT-3 trials -- Database againĀ
03:25:11 | GPT-3 trials -- GPT-3 experiment -- the trophy doesnāt fit in the suitcaseĀ
03:27:32 | GPT-3 trials -- Scrambling wordsĀ
03:30:41 | GPT-3 trials -- PDF cleanup example (Gwern)Ā
03:35:03 | GPT-3 trials -- Word breaking and simple text patternsĀ Ā
03:37:16 | GPT-3 trials -- Typing of entitiesĀ Ā
03:38:30 | GPT-3 trials -- Basic Python appendĀ
03:39:07 | GPT-3 trials -- Automatic programming?Ā Ā
03:42:31 | GPT-3 trials -- Passive aggressive dialog inputĀ
03:44:39 | GPT-3 trials -- symptoms of depressionĀ
03:45:43 | GPT-3 trials -- Red shirts reasoning challengeĀ
03:49:59 | GPT-3 trials -- Binary encodingĀ Ā
03:50:36 | Concluding statements from Walid, Tim and Yannic