Martin Signoux
@MartinSignoux
Followers
3K
Following
11K
Statuses
2K
Red-teaming innovation, fine-tuning regulation ——— AI policy @OpenAI
Paris
Joined June 2016
Enter the age of intelligence !
0
0
1
RT @sebkrier: Too many people in AI talk about decentralisation and centralisation as if this was a binary choice rather than a spectrum. B…
0
5
0
RT @aleks_madry: To better quantify LLM reliability, we construct platinum benchmarks by (manually) cleaning subsets of 15 classic “saturat…
0
4
0
Well done @kyutai_labs team !
Today we release Hibiki, real-time speech translation that runs on your phone. Adaptive flow without fancy policy, simple temperature sampling of a multistream audio-text LM. Very proud of @tom_labiausse 's work as an intern.
0
0
4
Deep !
“Steve Jobs called computers a bicycle for the mind; this, however, is an airplane.”
0
0
1
RT @natolambert: A fun example of OpenAI Deep Research attempting what would be a very useful task for me in the @interconnectsai discord.…
0
9
0
Deep Research achieving SOTA on GAIA and Humanity’s last exam 👀
0
0
4
RT @fchollet: The key is really this: AI usefulness scales logarithmically with inference time compute. Right now for many use cases the am…
0
310
0
@gchampeau L'inverse est vrai aussi. J'aurais personnellement bien aimé voir plus de défenseurs de LUCIE soutenir Mistral.
2
0
5
@Miles_Brundage Where’s the playbook for government organisations ? (Something equivalent to the shared prosperity framework for PAI but for AI readiness more broadly)
1
0
1
In the future, AI literacy won’t be about learning how to use a tool; it’ll about learning how to manage a workforce of agents. The implications are overlooked
If models have full context on our work and environment and who we are and our preference + good human level capabilities on the task we do. We’ll be able to run them in parallel to simulate many outcomes for a given tasks. Our role might shift from doing the tasks to expressing our preference by picking the outcome that is best?
0
1
4
Advancing capabilities AND safety 🤝
Trading Inference-Time Compute for Adversarial Robustness
0
0
2
🚜Farmers using tractors replaced farmers who didn't 📊Accountants using spreadsheets replaced accountants who didn't So goes the history of technological progress. It rewards adaptation. 👩💻 That's why AI literacy should be a top priority for most organizations
The main risk for workers is that they will be displaced by workers at other firms using AI, rather than by AI directly, Philippe Aghion, @BunelSimon, and @XJaravel argue.
0
0
3
@alexvoica @tim @arthurmensch From the article in Le Monde: "Our discussions with Mistral began about a year ago, so it is not linked to Meta's decision," stated Mr. Fries, emphasizing his "diversification strategy."
1
0
0