FR/US/GB AI/ML Person, Director of Research at @GoogleDeepMind, Honorary Professor at @UCL_DARK, @ELLISforEurope Fellow. All posts are personal.
Nov 19 • 11 tweets • 2 min read
🌶️(?) take: Agents are somehow hot right because people realized that LLM output can be interpreted as a DSL which directs side effects in the world (e.g. tool calls) rather than just returning text in a chat/autocomplete sense. What are the open challenges? A 🧵... [1/11]
Setting aside the fact that people are literally re-inventing a bunch of terminology around agents when we have several decades of framing from RL (POMDPs, anyone?) that we can rely on to describe what we're doing, the main challenges are as follows, working backwards. [2/11]
Dec 29, 2020 • 16 tweets • 9 min read
2020 retrospective time. It's been a rubbish year for productivity, but I've been privileged to have that being the main impact of the pandemic in my life. I feel sorrow and sympathy that not everyone can say the same…
Here's some highlights and related thanks.
[1/14]
It seems like the work was done years and years ago, but happy to have been able to revisit our work on RTFM with @hllo_wrld when he presented it at @iclr_conf 2020. arxiv.org/abs/1910.08210 [2/14]
Dec 31, 2019 • 19 tweets • 16 min read
To end 2019, I want to express thanks for all the great people I've worked with this year. Naturally, this is an opportunity to brag a little, but all this work is due thanks to leadership/effort from fantastic interns, collaborators, etc, so hopefully you will indulge me (1/16)
It's been a delightfully short working year, thanks to @Facebook's generous paternity leave (4mo) thanks to which I spent (along with PTO) over 40% of the year at home with my infant daughter. Both practically, and in terms of morale, this has been a huge help. (2/16)
