Ed Grefenstette

1.7K Followers
63 Following
10 Posts
FR/US/GB AI/ML Person, Head of ML at Cohere, Honorary Professor at UCL DARK, ELLIS Fellow. Ex Oxford CS, DeepMind, Meta (FAIR).
An #introduction!đź‘‹ I'm a PhD student with @rockt and @egrefen interested in language and reinforcement learning. I think the next step for LLMs is improving their understanding of pragmatic language in context (e.g. the A: "I wore gloves" to the Q: "did you leave fingerprints?" means "No") and want to investigate more dynamic ways of learning language than next-word prediction
Belated #introduction: I am interested in open-endedness and reinforcement learning research. I'm an Associate Professor at the UCL Centre for AI where I am a PI for the UCL DARK (https://dark.cs.ucl.ac.uk/) lab with @egrefen (some of our students are here as well @LauraRuis @minqijiang @mika) and collaborate closely with @jnf. At DeepMind, I am leading the Open Endedness Team (follow @jparkerholder).
This is, of course, just the beginning. Grounded agents and greater customisability are all coming your way, courtesy of newer @[email protected] models in the pipeline. But ultimately, the power to build custom #ml-powered ConvAI solutions is in your hands TODAY:
https://github.com/cohere-ai/sandbox-conversant-lib
Back to #ml: I am having so much fun playing with the demos of persona-enabled conversational agents powered by @[email protected]'s models. Forget deep philosophical convos, it's awesome to see proper context tracking in a practical conversation...
https://conversant.streamlit.app/

Quick #introduction: my name's Ed. I live in London (Richmond, specifically) with my family. I'm Head of ML at Cohere, and Honorary Professor at UCL, where I co-lead the DARK lab with @rockt.

My research has spanned NLP, program induction/synthesis, meta-learning, open-ended learning, RL, and has somehow cycled back to NLP. I have a broader background in Physics and Philosophy, although I've forgotten most of the former.

Looking forward to a saner time on this platform than on Twitter!