Monicaxie (@immonicax) started following @edwinarbus on Oct 1, 2024

OpenAI DevDay keynote at 10 AM. 🥵 https://t.co/pn1sbhbp85

we’re in the a̶r̶e̶n̶a̶ marina getting ready to ship some of your most requested features to the API https://t.co/rqv7598JhP

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,

This research direction is a big deal. T gates have for a long time been a big enough bottleneck that they were often the only operations counted in resources estimates. This is changing. As @CraigGidney notes: "Over the past decade the expected cost of a T gate has gone from

Will be in Austin next Monday hosting this quantum and physics tech event with @JurczakQris at @CapitalFactory See you there!

RT @joehenriod: Yoga class with Ai? A Krea Motion Brush Experiment. Photos by Midjourney. https://t.co/EgwC142UtX

RT @dvorahfr: AI allows all the madness 😎 @Kling_ai text to video https://t.co/Yc67IDSDnG

The talk covers our recent work on training vision-language-action models for robotics. I'll discuss how embracing LLMs/VLMs in robotics allows us to scale policy learning, but also cover key differences btw robotics and other multi-modal agents. Zoom: https://t.co/TdiESaUU1X https://t.co/9vXSfKfruc

I'll give a talk at the Multimodal Agents workshop at ECCV tomorrow Sept 30, at 2:20pm CET. Excited for my first talk at a vision conference: robotics is increasingly becoming a multi-modal sequence modeling problem w/ lots of potential for LLM/VLM researchers to have big impact! https://t.co/r3KH5qbxQo

DM me if you’re attending and would like to meet up

just wrapped up our first set of rehearsals for DevDay. this week will be an exciting one for developers! https://t.co/AR7buoTWZv

Another notable take from here: when a tech is truly disruptive, it can win over distribution. AWS, with over 60% share in public cloud, together with Anthropic, only takes <14% of total mkt of LLM API. Google’s mindshare in AI is still strong, even w/ all the complaints on GCP..

RT @ZeyuanAllenZhu: Now uploaded a 50-min video for Part 2.2, covering many technical details: https://t.co/2U1u1x37pm https://t.co/sDHQCKT…

SPC x Luma hack day is live! Swing by SPC SF at 5pm for pizza + demos https://t.co/lJoG9ik9zY

ALSO! Demos will be at 5pm at 380 Brannan St if you want to see what everyone built!

RT @hwchung27: Learning, if defined from first principles, shouldn't need to assume that a student is of a particular type (human, monkeys,…

RT @M3LWorkshop: 📡Join us at the 2nd workshop on Mathematics of Modern Machine Learning (M3L) at #NeurIPS2024! https://t.co/crLL1KQZMV Sub…

This is happening now! I will be poster #913 presenting this work. Happy to chat if you are interested!

Link to book here: https://t.co/IWMLwzONPP

The shock therapy used in Poland known as “Plan Balcerowicza” was heavily influenced by Jeffrey Sachs (recently in a @theallinpod panel to rave reviews). Balcerowicz’s memoirs are worth a read and hopefully someone will translate them. A win for liberal democracy nation building https://t.co/zMoCadN259

RT @amanmathur_: what does the inside of a website look like? leslie and I explored this in "translucent website" https://t.co/vDQyLVINF1

RT @thejamescad: Day 1: Profound HQ, Union Square NYC. Gonna be spending a lot of time here with @dbabbs. https://t.co/FRDjuGyhdw

RT @arjunkhemani: .@elonmusk (worth $250B) answers why he’s still working: I think it's a good question you asked, because it goes to, lik…

🦅🇺🇸

Claude Console 之前是我使用最高频的 AI 工具之一，主要是因为它的 Prompt Generator 功能，可以帮你直接优化生成一个生产级的 Prompt，之前也向 @oran_ge 兄大力推荐。后面 Claude 把它开源了，flowith 也在产品中集成了 Prompt Generator 模式，你现在可以随便输入一个简单/狗屎的 prompt，AI

Claude Console was my go-to AI tool, mainly for its Prompt Generator. Thanks to @AnthropicAI for open-sourcing it. @flowith_ai now has a Prompt Generator mode - type any prompt, let AI optimize it. Often, a great prompt + Claude 3.5 Sonnet still beats O1-Preview. ⬇️ https://t.co/fVjL9cSlR7

We found you on the app https://t.co/SjeQAULxnF

From the organizing committee and everyone involved we hope that you had an amazing time at ICML https://t.co/UIK68Wr2OZ

Code is a (graph) data structure. Storing it in text files is convenient for IDEs, but has many downsides: ❌ Import hell ❌ Large files ❌ Poor tree shaking ... etc. I think we'll soon eliminate the "file" abstraction and instead operate directly on the symbol graph.

Terence Tao is giving commentary on o1's math capabilities on Mastodon, and has mixed but overall optimistic takeaways World's foremost expert is seriously evaluating a model vs. his PhD students w/ mixed results. Rate of improvement is incredible. https://t.co/CsGf5weNX2 https://t.co/tjqJVT8dxt