Pushpak Kedia (@pushpakkedia) started following @samuel_colvin on Oct 2, 2024

RT @jxnlco: Check it out if you need observability in your Python applications. https://t.co/d3N946aDZZ

RT @jxnlco: Congrats to closing series A Pydantic :) https://t.co/Dc2ipE0VXt

RT @Hersh_Desai: We're incredibly excited to announce the latest batch of startups joining AI Grant! https://t.co/C5VKJAuYmy

RT @pdhsu: Excited to welcome the newest batch of @AIGrant companies! https://t.co/yuiMezxMMQ

Love the diversity of companies in here. I can also only *assume* based on my own conversations, that every single founder in here is using @meetgranola for all their meetings. 80% of all founders I've talked to in the last 2 weeks are using it, and if not I tell them to 🗒️💻🤖

Always exciting to see what the crew @neo is cooking, thanks for a fun day @suzannexie, @apartovi and @_shriyabhatt! https://t.co/gxBN48fjN7

RT @NVIDIAAIDev: 🎉 News from #NVIDIAResearch at #ECCV2024 in Milan 🎉 From embodied #AI to foundation models, our researchers are showcasing…

RT @nvidia: On October 7-9 policy and industry leaders will gather in Washington, D.C. for #AISummit to discuss how #AI addresses some of s…

Samsara Inc $IOT Deep Dive https://t.co/5L57THltjm https://t.co/85fYkc7YiE

$GPN Global Payments Inc Deep Dive and 2024 Outlook https://t.co/cqXoBMEDDH https://t.co/XyQjmNP8rt

Overall, our paper provides a theoretical and empirical investigation on how scale impacts model forgetting. All code to reproduce our experiments is provided in our GitHub. Huge shout out to my coauthor @VihanL! 🧵8/8 https://t.co/nNh2yoQ1uE

We explore other connections between architecture and forgetting as predicted by our theoretical framework. In fact, our theoretical framework and experiments demonstrate that as the number of tasks seen increases, forgetting increases linearly. 🧵7/8 https://t.co/RkjmjJ7fz2

RT @JonSaadFalcon: What is the best way to spend your inference compute budget to create LLM systems greater than the sum of their parts?…

RT @etash_guha: Excited to discuss my ICLR 2024 work! Our paper “Conformal Prediction via Regression-as-Classification” can simply predict…

NotebookLM is a reemergence of what made Google great. A few people got together to ship an idea that’s cool + useful and then oops - PMF.

RT @Oscurart: I made a new version of the Trinitron CRT Display. #b3d #blender #blender3d #eevee #cycles https://t.co/MGL8bp0q8r

When they say the NotebookLM team really "cooked" were they referring to our evals https://t.co/x1u4tq73io

Aside from TPUs running hot today, five things from Notebook HQ: 1) Thanks for all your feedback on AOs so far. I’m copy pasting everything into a Notebook so I can listen to a Deep Dive and search it later. We’re going to launch some immediate tweaks to make it less repetitive,

We hit 100 Stars ⭐on github https://t.co/nLSFnBguBU

Any one from @huggingface is it possible to sponsor a GPU zero space for this demo?

This whole tweet thread was written with Flow. Start today for free 👉 https://t.co/ObLJvkmRI4

Flow + Cursor is a deadly combination. Watch @mckaywrigley make a Perplexity clone in 8 mins. https://t.co/2NiFic5Jzi

RT @Azaliamirh: Very excited to introduce Archon, a framework for LLM inference-time architecture search! Paper: https://t.co/wIxxoxhez7 C…

7/7 Paper: https://t.co/125VOobesy Code: https://t.co/SG9yl2GnBV Project: https://t.co/CG7St7HAZ2 It’s been a lot of fun collaborating with @adriangamarra__, @Shlok_Natarajan, @nahuum_maru, @ht_todorov, @etash_guha, @ekellbuch, @MayeeChen, @NeelGuha, @HazyResearch, and

Last time I posted about the specific model was the Phi family release - it was a breakthrough. Today - another one, please take note. @SebastienBubeck himself in our private correspondence said: "The Liquid models are strong. Wow." - sorry, Seb, oversharing, I know

Shopify did. The integration of Ads and LLMs has been solved - and it is a win for consumer. In the technical parlance it is called CPA ads, aka "Cashback and Coupons". "This is the best product for you, this one is slightly worse, but I have a coupon".

I really hope OpenAI can work these issues out, with or without Sam. AGI is a profound and incredibly consequential technology, and I want all companies working on it to be as sensible and well-functioning as possible.

Like, regardless of whether you think x-risk is a bunch of sci-fi nonsense, that is just not a well run company, and it is the job of a board to do things about that.

really exciting work led by @JonSaadFalcon!!

RT @AnkaReuel: Technical AI Governance research is moving quickly (yay!), so @ben_s_bucknall and I are excited to launch a living repositor…

RT @rosedotai: Excited to announce our new partnership with @CMEGroup! This collaboration brings comprehensive futures and options data to…

RT @abcampbell: Incredibly proud to announce you can now get @CMEGroup data in Rose! One of many amazing financial data vendors we've gott…