To make #RealityCentricAI inspired ML models work, we need to make them trustworthy! Uncertainty Quantification is ESSENTIAL! But, we need to understand what UQ models flag as suspicious - Hao Sun, Boris van Breugel, Jonathan Crabbé, Nabeel seedat & I explore that for #NeurIPS!

After many attempts, finally landed in New Orleans to attend #NeurIPS 🛬✅ I will be presenting several cool papers on interpretability, data-centric AI and deep ensembles during the week, starting with the first poster session Tuesday morning. Stay tuned for details 🔜

🚨 Tomorrow, on Tuesday, don't miss my #NeurIPS2023 𝐏𝐨𝐬𝐭𝐞𝐫 #𝟏𝟓𝟐𝟏 on the robustness of interpretability methods @ 𝟏𝟎:𝟒𝟓-𝟏𝟐:𝟒𝟓. A great opportunity to chat about how model invariances are reflected (or not) in popular interpretability methods!

Seeing my students come into their own is one of the greatest pleasures in my professional life! Heads up to Jonathan Crabbé and his fantastic poster pulling a substantial crowd at #NeurIPS2023 - very well done!

🚨 𝐏𝐨𝐬𝐭𝐞𝐫 #𝟓𝟑𝟕 𝐖𝐞𝐝𝐧𝐞𝐬𝐝𝐚𝐲 𝟓𝐩𝐦-𝟕𝐩𝐦 🚨 We will challenge many intuitions on training deep ensembles at #NeurIPS2023. Super excited about this awesome collaboration between Alan Jeffares @ ICML 🇨🇦, Tennison Liu, Mihaela van der Schaar and myself. Come chat with us & check the 🧵

🚨Tomorrow (Wednesday) at 5pm #NeurIPS 🚨 If you are interested in ensembles, MoEs, optimisation, vision, or memes come visit myself & Jonathan Crabbé at poster #537!

Wednesday at #NeurIPS2023! Today we are presenting posters with Yuchao Qin, Changhee Lee, Alan Jeffares @ ICML 🇨🇦, Tennison Liu, and Jonathan Crabbé! Join us at the session to discusse Timely Outcome Prediction, Deep Ensemble Training, and everything #realitycentricAI!

Had an amazing time at #NeurIPS where we discovered that if you forget to take your poster down you may end up doing a whole second poster session the following morning feat. Jonathan Crabbé with an (appropriately?!) guilty expression 😅

Always a pleasure to engage with the ML community at conferences! Had an amazing time at #NeurIPS2023 and really interesting conversations with researchers from all around the world. Really worth arguing with Reviewer 2 🤩

Un rêve de pouvoir enfin porter les couleurs de la Karmine Corp au plus haut niveau et d'écrire un nouveau chapitre pour l'organisation. Rendez-vous dans 10 jours #KCORP

I'm very excited to share our papers that have been accepted at @aistats_conf2024! This year, we keep pushing our research on reality-centric AI - from advancing scientific discovery and adaptive experimental design to data-centric AI. Some amazing technical work by my students!

Excited by the generality of CLIP, but need more fine-grained details in your representation? Introducing SPARC, a simple and scalable method for pretraining multimodal representations with fine-grained information. 🥳 1/6

Latent Labs comes out of stealth today with $50M funding. Our goal? To push the frontiers of generative biology, giving partners instant access to tools capable of accelerating drug design. Every biotech or pharma company searching for the best therapeutic molecules understands

Introducing Latent-X — our all-atom frontier AI model for protein binder design. State-of-the-art lab performance, widely accessible via the Latent Labs Platform. Free tier: platform.latentlabs.com Blog: latentlabs.com/latent-x/ Technical report: tinyurl.com/latent-X

5 months from funding to shipping our first frontier model. Latent-X achieves state-of-the-art hit rates for macrocycles and mini-binders, with picomolar binding affinities — a breakthrough in de novo protein binder design. Available now on our no-code platform!

Really happy to be releasing our first model on a beautiful platform. It generates de novo binders with impressive affinity and speed. If you are curious, you can try it for yourself, more info on the thread👇

🧬 Latent-X is our first frontier model for protein design. For mini-binders, our AI model achieved picomolar binding affinities — the strongest reported vs prior methods in head-to-head lab validation. Explore our best binders here (no login required): platform.latentlabs.com

The first 3 days have been incredible. The response from our research community has been phenomenal. Many hundreds of users, tens of thousands of designs and amazing feedback flowing in.

Showcasing Latent-X's breakthrough macrocycle designs. Our AI achieved an unprecedented 91-100% hit rate with single-digit micromolar binding affinities—functional binders on every single target tested. Explore our binders without logging in: platform.latentlabs.com

Latent-X has been available for a little over two weeks, and we couldn't be more excited about the science that is already being done with it on our platform. Our CEO Simon Kohl joined Bloomberg News Daybreak Europe this morning to talk about why we're building it, and the impact