
I am honored to share that I've been endorsed by the Royal Society as a Global Talent in AI research, and the UK is officially home. 🇬🇧🎉 If you are in London, reach out; I would love to catch up! 🌞

•mle @cohere; LLM posttraining •i write code&songs&stories •🇹🇷🇬🇧 tweets •proud Bilkenter •chasing legacy✨ •views are mine, not my employer's. rt≠endorsement

I am honored to share that I've been endorsed by the Royal Society as a Global Talent in AI research, and the UK is officially home. 🇬🇧🎉 If you are in London, reach out; I would love to catch up! 🌞

you dont work on bleeding edge stuff if your slack isnt littered with this guy


ICLR reviews are really instructive. Now I'm thinking, what a shame I didn't run multiple seeds for every decision I've made in life to make sure the results were statistically convincing.

Why is AI slop so frustrating to read? AI writing often obscures main ideas and invents awkward phrases, making sentences more complex than necessary. Many of these slop patterns violate writing principles established in English writing books. In our benchmark, AI writing breaks these principles far more often than human writing, with Claude Code replies by far the worst offenders. We show examples of the most common slop patterns in this thread

cant even go on the tube without seeing ads about ai agents anymore 😁🐰 the adoption is insane!

also my inbox below up, im trying to catch up. please bear with me!

this role is remote friendly and we have offices in toronto, london, sf, nyc, paris and more! im working from the london office in hybrid mode


Appreciate reposts and leads!!

My teammates are all brilliant people and I believe we have a really nice and collaborative culture. I highly encourage applying! My DMs are open if you have questions.👩🏻💻

My team at Cohere is hiring for various roles. This is the perfect opportunity to work at a company training their own LLMs from scratch! Both roles are on the post-training side of the training stack and highly impactful. 1. Member of Technical Staff - Post-training (Data Infrastructure): jobs.ashbyhq.com/cohere/eef8… -> If you have experience in creating highly scalable and efficient data pipelines and if you are interested in solving state of the art problems at the intersection of large language modeling and engineering, this role is perfect for you! 2. Member of Technical Staff - Post-Training (Reinforcement Learning & Generalization): jobs.ashbyhq.com/cohere/8a73… -> If you want to make GPUs go brrr by creating new RL environments and algorithms that make the model more generalizable and robust, you should apply!

18 of the 20 best open models today are Mixture-of-Experts. Dense is basically over. And yet the know-how to train a good MoE still lives mostly in the heads of a handful of teams. So we spent 800k GPU hours running the experiments ourselves and wrote everything down. Super happy to finally release "The Ultimate MoE Training Handbook", free and open, organized around the questions training teams hit: - how many experts, how big, shared or not? - why is half my step spent in all-to-all, and how do I hide it? - why is my run 25x slower with no error in the logs? - softmax, sigmoid or sqrt(softplus) router, and why keep it in fp32? - aux loss or DeepSeek's loss-free bias? - upcycle a dense checkpoint or train from scratch? - can I serve with fewer experts per token than I trained with? ⚠️ Warning: it's long (~70k words, budget a few days, sorry) but with 55 interactive figures, code you can run and numbers from our own runs. Read it here: huggingface.co/spaces/huggin…

Wow, this blew up. Thanks so much for the support and your kind words! 💜 I got a lot of questions in my DMs so I decided to write a blog post about my experience. If you have questions about the global talent via process, drop them in the thread or dm me. I'll try to cover as much ground as possible. I hope that it will be helpful!
I am honored to share that I've been endorsed by the Royal Society as a Global Talent in AI research, and the UK is officially home. 🇬🇧🎉 If you are in London, reach out; I would love to catch up! 🌞

Can we sexualize normalizing

The sort of thing your agent writes tests for
The chances of being killed by a landmine delivered by a bear are low, but they are never zero.