Collaboration Is a Very Good Way to Do Science

3min Snip

00:00

Play full episode

Summary

Transcript

Episode notes

Daniel: I think a number of our guests recently have spoken to this aspect of collaboration whether it's on an individual level the people you might want to write papers or do a project with but then also the institutional level between perhaps different colleges things like that. We need more communities like that so I definitely check out c4ai cohere for AI which is producing a fantastic contribution in that space. You can subscribe to the gradient on sub stack to receive not just this podcast but also our articles and newsletters directly to your email. Visit us at the gradient.pub where you'll find all of that as well as more information about the gradient and how you could even contribute if you're interested.

In episode 62 of The Gradient Podcast, Daniel Bashir speaks to Ed Grefenstette.

Ed is Head of Machine Learning at Cohere and an Honorary Professor at University College London. He previously held research scientist positions at Facebook AI Research and DeepMind, following a stint as co-founder and CTO of Dark Blue Labs. Before his time in industry, Ed worked at Oxford’s Department of Computer Science as a lecturer and Fulford Junior Research Fellow at Somerville College. Ed also received his MSc and DPhil from Oxford’s Computer Science Department.

Have suggestions for future podcast guests (or other feedback)? Let us know here!

Subscribe to The Gradient Podcast: Apple Podcasts | Spotify | Pocket Casts | RSSFollow The Gradient on Twitter

Outline:

* (00:00) Intro

* (02:18) The Ed Grefenstette Origin Story

* (08:15) Distributional semantics and Ed’s PhD research

* (14:30) Extending the distributional hypothesis, later Wittgenstein

* (18:00) Recovering parse trees in LMs, can LLMs understand communication and not just bare language?

* (23:15) LMs capture something about pragmatics, proxies for grounding and pragmatics

* (25:00) Human-in-the-loop training and RLHF—what is the essential differentiator?

* (28:15) A convolutional neural network for modeling sentences, relationship to attention

* (34:20) Difficulty of constructing supervised learning datasets, benchmark-driven development