🔥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

Very proud of our 📣 new paper on Measuring Anthropomorphism in LLMs! This new multi-turn evaluation & large-scale human study led by Lujain Ibrahim لجين إبراهيم is a key step in better understanding what factors make people perceive LLMs as more "human-like". 🧵

Today we introduce Sarvam-M, a 24B open-weights hybrid model built on top of Mistral Small. Sarvam-M achieves a new benchmark across a range of Indian languages, math, and programming tasks, for a model of its size. Here is a detailed technical blog on how we customize

UAE is planning to become the first country in the world to offer free access to ChatGPT plus to all its citizens and residents! OpenAI has likely made a deal with an entire country as its client. thearabianstories.com/2025/05/25/fre…

this paper quantiatively shows what someone told me: "everyone loves interdisciplinary but no one will give u a job if you are interdisciplinary" Hiring at top universities rewards disciplinary loyalty over interdisciplinary breadth. Things are changing. arxiv.org/abs/2503.21912

It's interesting how the major LLM API vendors are converging on the following features: - Code execution: Python in a sandbox - Web search - like Anthropic, Mistral seem to use Brave - Document library aka hosted RAG - Image generation (FLUX for Mistral) - Model Context Protocol

You know all those arguments that LLMs think like humans? Turns out it's not true. 🧠 In our paper "From Tokens to Thoughts: How LLMs and Humans Trade Compression for Meaning" we test it by checking if LLMs form concepts the same way humans do Yann LeCun Chen Shani Dan Jurafsky

I had an hour long conversation with Natasha Zouves from News Nation about the extremely complex topic of the impact of AI on jobs and careers The video is on YouTube - here's my Claude Opus 4 assisted index of the transcript, with links to key sections simonwillison.net/2025/May/30/ai…

I wonder why "age of AI" has become such a common phrase in recent research articles about AI recently. I don't remember "age of Covid" being a common academic phrase. How did "age of AI" come to be then? Was it injected by LLMs itself into academic prose? 🧐

lol!

People with ALS often use eye-gaze typing, reaching 8–10 wpm (compared to 150–200 wpm for neurotypicals). A new AI tool, SpeakFaster, lets users select just the first letter and predicts the rest. Early results: 29–60% faster communication. nature.com/articles/s4146…

One week later...

Good post from Balaji on the "verification gap". You could see it as there being two modes in creation. Borrowing GAN terminology: 1) generation and 2) discrimination. e.g. painting - you make a brush stroke (1) and then you look for a while to see if you improved the

🏆 Honored to receive the 2025 ATTW Graduate Research Award for my work on generative AI prompts in tech & professional communication. Grateful to the committee and mentors for their support! #ATTW2025 #AI #TPC #HCI #WritingStudies

I interviewed 108 machine learning researchers about their experiences with writing! From the abstract: Given the hype around artificial intelligence (AI), it is imperative to investigate how researchers of AI negotiate this hype as well as wrestle with it in their research.

I've published video, slides and a detailed annotated transcript from my talk at this week's AI Engineer World's Fair conference AI Engineer in San Francisco - "The last year six months in LLMs, illustrated by pelicans on bicycles"

If I had to learn Math for Machine Learning from scratch, this is the roadmap I would follow:

We're launching Veena TTS 🪕 on June 20 Our flagship text-to-speech model for Indian languages 🇮🇳 Natural, expressive, and actually sounds like us. We’re launching two models: Veena Lite >Open-source and lightweight >4 unique, natural-sounding voices >The first open-source

A key critique from Bender et al. (2021) was that LLMs lack linguistic diversity. That’s thankfully changing—Sarvam AI just released open-source models trained on 10 Indian languages including English. Check them out: sarvam.ai 🇮🇳📚