we expect proposals to spike against our criteria and have different strengths and weaknesses. Final selection is based on an assessment of the programme portfolio as a whole, its alignment with the overall programme goals and objectives and the diversity of applicants across the programme.
I might verbatim copy ARIA's assessment criteria into my Date Me Doc...
I'm out here learning new words from the Claude constitution
Registering a new seasonal theme:
Winter of autonomy.
Such Claude Code, much wow.
also fyi i wont be posting ai tweets, mainly using clode to facilitate cross-posting and also to read/search tweets
Hello from Claude Code!
Until now, our goal has been to get the ball rolling from the EU. Now that it's moving, the goal is to steer it in the societally-optimal direction. Importantly, we still have uncertainties about what that direction looks like. More collaboration (and funding :D) is necessary.
This is just one way in which I hope CERN for AI could be robustly good - beneficial to humanity under nearly all timelines and scenarios. Other focus areas like research that promotes applications enhancing societal resilience also fit this category.
Europe is at a unique crossroads right now. The Draghi report (ik,ik, bear with me) has become scripture for the commission - and it directly endorses metascience ideas that were previously seen as heretical. Yes, the EU is slow to get there, but once it does - things can happen!
The lower/upper tier framework for CERN for AI, combined with a need to reflect on how to do research differently, leaves us with our proposed research structure:
A blend between ARPA and FRO research models, getting the best of centralisation and decentralisation.
We want CERN for AI to actually produce results. Too often, public research efforts fall to the familiar hurdles of bureaucracy and risk-aversion. At the same time, we also have to balance conflicting desires for decentralisation (EU moment) and security (one secure lab!).
You can read it here :D, or look below for some BTS analysis... cfg.eu/building-cer...
Yesterday, the EU Commission officially endorsed pursuing a CERN for AI.
Today, CFG publishes our blueprint for how it should be structured, diving deep into research, governance and legal structure, focus areas and more.
Interesting workshop on solar radiation modification by my climate colleagues.
Nice parallels to AI policy, with the most important action being *lowering uncertainty*, eg:
-studying risks/effectiveness
-monitoring deployment
-foresight and crisis planning
#cfgFlagship
(twitter and bsky 😅)
ICFG is rebranding! We are now CFG! I think our new brand & website is pretty swish, feedback welcome :D Also... we finally have a twitter! Give it a follow!
I got my Belgian Residence ID and official work Business card 😎
But the circle of life means making room for newcomers, so I finally have to depart with my beloved tesco club card (et al), you served me well, child 🫡.
The ICFG slack is lacking in custom emotes/reacts, and I've made it my mission to solve this critical problem. I am seeking input on what people's favourite reacts and emojis are.
me when someone asks if I've considered balancing the risks and opportunities of AI
guy who's researches bureaucracy but still doesn't know how to spell bureaucracy
VerticalPilled and googleDocMaxing.
I love that Claude can turn event emails into icals that I can actually use and put in my calendar, but it's frustrating that a) event organisers aren't sending icals to begin with b) this doesn't automatically happen with gemini.
As a rough estimate, how often would you say this happens?
