🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Ahem... Clojure

During my PhD, I set my goal as "finish one paper I am really proud of" each year. This is still my measure of success as a tenured faculty at a top university. It adds up quickly!

I came to do my PhD in the UK (and stayed to eventually pay more taxes than 99% of Brits) only because my partner could move with me. As a Cambridge academic, I am losing out on great students, top global talent, who choose Germany because they have a partner.

I'm teaching a new course on AI Alignment this term at the University of Toronto. The first half will cover idealized models of future AI systems (optimal planners, universal induction, etc.), and the second half will cover practical alignment techniques in the context of LLMs.

Der jüdische Historiker Hans Kohn zitiert 1964 in seiner Autobiographie "Bürger vieler Welten" die Memoiren von Simone de Beauvoir, um die Appeasement-Stimmung der 1930er Jahre zu beschreiben, in der die Intentionen des Nationalsozialismus verkannt wurden. 🧵

Our new paper auto-distills machine-learned algorithms from blackbox neural networks into Python code. The RNN below is seen to have rediscovered the "ripple " algorithm for long addition that we learned in school. We develop an integer autoencoder to convert the RNN into a

Depois de muito aprendizado sobre controle de concorrência, testar implementação com 3 bancos de dados diferentes (Datahike, Datalevin, PostgreSQL) e com direito até a contribuição para projeto open source, lancei a minha submissão para a Rinha de Backend

New paper: how to do flow matching on discrete data. Flows give a simple generative framework and better performance than discrete diffusion models. Discrete flows are easily combined with continuous flow matching for multimodal models. arxiv.org/abs/2402.04997 A thread 1/7

Unveiling the next chapter in our quest towards the elusive theory of *all* neural networks: Categorical Deep Learning! 🐈🤖 categoricaldeeplearning.com See thread 🪡 for a bucket list 🪣 of concepts our framework can handle! w/ Bruno Gavranović P. Lessard Andrew Dudzik Tamara von Glehn @_joaogui1

I am really excited to reveal what Google DeepMind's Open Endedness Team has been up to 🚀. We introduce Genie 🧞, a foundation world model trained exclusively from Internet videos that can generate an endless variety of action-controllable 2D worlds given image prompts.

Wow. Here’s Robert De Niro’s full statement about how Donald Trump should NEVER be president again: “I’ve spent a lot of time studying bad men.  I’ve examined their characteristics, their mannerisms, the utter banality of their cruelty. Yet there’s something different about

I have been working on this over the past 9 months with my colleagues from PLAI Group. It is now deployed in multiple AWS data centers. Let's see how quickly it grows...

The #OSI² ONE #OpenSource #MRI #scanner: Press: ptb.de/cms/presseaktu… Web: opensourceimaging.org GitLab: gitlab.com/osii Build workshop movie: youtube.com/watch?v=1q0dXg… Workshop slides: zenodo.org/records/100796… #DIYbio #DIYchem #lab #instruments #analytics #MRT #medicine

We should call models like Llama 3, Mixtral, etc. “open-weight models”, not “open-source models”. For a model to be open-source, the code and training data need to be public (good examples: GPT-J, OLMo, RedPajama, StarCoder, K2, etc.). Weights are like an exe file, which would be

The revolution will be simulated.

We are happy to present Simformer, an all-in-one amortized Bayesian inference method. Simformer is flexible and efficient and opens up new possibilities for simulation-based inference. Accepted as oral at ICML 2024. By Manuel Glöckler Michael Deistler Christian Weilbach [email protected] Frank Wood Jakob Macke

Piden transparencia para las #EleccionesVE2024🗳 Argentina🇦🇷, Costa Rica🇨🇷, República Dominicana🇩🇴, Ecuador🇪🇨, Panamá🇵🇦, Paraguay🇵🇾, Perú🇵🇪 y Uruguay🇺🇾 pidieron "un #conteo de los #votos transparente, que permita la verificación" en Venezuela🇻🇪 (ama) 🔗⬇ dw.com/p/4iqGi

Announcing LeanAgent: the first life-long learning agent for formal theorem proving in Lean. LLMs have been integrated with interactive proof assistants like Lean for theorem proving with 100% accuracy. So far, these LLMs are static, cannot learn new knowledge online, and

youtube.com/watch?v=-YJgcT… very nice demonstration of copilot jailbreak dangers by mbg

Next time I will meet you all in person again. Thanks for organizing LAFI!