NLP Benchmarks - Can You Just Get the Best Data?

2min Snip

00:00

Play full episode

Summary

Transcript

Episode notes

The field is going significantly more towards what sort of new things can we build with this by just varying the data, because that's where the money is. We in our group at the NLP group at DeepMind were big fans of the benchmark driven development of the field. I would have liked it to be more recognized at the time. There's a lot of room for well designed benchmarks to move the needle of science. And that's what we tried to do with this pragmatic benchmark I mentioned earlier.

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