Trapit Bansal

Thrilled to be joining @Meta! Superintelligence is now in sight 🚀

Imagine the next 9 months

It’s been so much fun working with the extremely cracked team at Meta on this! This is just the start, even more capable models coming in the future. ai.meta.com/blog/introduci…

1/ today we're releasing muse spark, the first model from MSL. nine months ago we rebuilt our ai stack from scratch. new infrastructure, new architecture, new data pipelines. muse spark is the result of that work, and now it powers meta ai. 🧵

Got flooded with “go back to your country” and “America doesn’t need you” on my H-1B tweet. America is 4.1% of the world. Even if Americans were 10X more likely to be geniuses, 70% would still be born elsewhere. Elon Musk, Jensen Huang, and Sergey Brin weren’t born American.

I’m excited to be the Chief AI Officer of @Meta, working alongside @natfriedman, and thrilled to be accompanied by an incredible group of people joining on the same day. Towards superintelligence 🚀

Introducing a research preview of Codex in ChatGPT openai.com/live

way back on friday, the high score on "humanity's last exam" was o3-mini-high at 13%. now on sunday, deep research gets 26.6%.

whatever else you think of openai it is now responsible for two separate 0->1 general intelligence scaling breakthroughs that required vastly different approach and mindset

🎄🎅starting tomorrow at 10 am pacific, we are doing 12 days of openai. each weekday, we will have a livestream with a launch or demo, some big ones and some stocking stuffers. we’ve got some great stuff to share, hope you enjoy! merry christmas.

@pablothee @OpenAIDevs @miramurati The instructions provided in the prompt do influence how the model thinks about the problem

@OpenAIDevs @miramurati Will there be ways to influence the middle steps for thinking?

We’re hosting an AMA for developers from 10–11 AM PT today. Reply to this thread with any questions and the OpenAI o1 team will answer as many as they can.

@GorvGoyl @OpenAIDevs o1 is a model trained in a completely new way that allows it to think

@OpenAIDevs is o1 just an existing LLM (like gpt4o) but with a feedback loop (chain of thought)?

@pixqc @OpenAIDevs My take is that this a fundamentally new scaling paradigm that is different from pretraining and the thinking capability allows inference time compute for responses, that wasn't previously possible

@OpenAIDevs What does this mean exactly? Previously capability was injected through pre-training and now inference can boost capability?

@amadad @OpenAIDevs We have seen it give really good responses for creative tasks as well, like when asked to generate poetry it is often more thoughtful and thinks carefully about rhyming scheme and the story it wants to tell

@OpenAIDevs How does the o1 model’s focus on reasoning, quantitative measures and benchmarks, influence its capacity for creative / lateral thinking? Is a form of abductive reasoning incorporated into o1? If not, can it be?

Letting the model come up with its own chain of thoughts can unlock System 2 reasoning -- so just let the parrot keep talking.

@bryonkuchML @OpenAI thank you! yes, hope you are good too!

@TrapitBansal @OpenAI Congrats! Hope you’re doing well 😊

Thrilled to see the launch of OpenAI o1! It's been an incredible (almost) 3-year journey working on this project with the amazing team at @openai to build AI models that take reasoning and problem-solving to the next level. Excited to see how you all use it! 🚀

@kalpeshk2011 @MohitIyyer @andrewmccallum @HamedZamani @YejinChoinka @Google Congratulations Kalpesh!

Happy to share that I've defended my PhD thesis! A big thank you to my advisor @MohitIyyer and my committee @andrewmccallum, @HamedZamani and @YejinChoinka for all their support and feedback! Next step: I am super excited to be joining @Google Bard as a Research Scientist!

Announcing GPT-4, a large multimodal model, with our best-ever results on capabilities and alignment: openai.com/product/gpt-4

Excited to share our new review paper “Tools for Extracting Spatio-Temporal Patterns in Meteorological Image Sequences: From Feature Engineering to Attention-Based Neural Networks ” w/ @Iebertu, yoonjinlee, kylehilburn @cira_csu; Link: arxiv.org/abs/2210.12310 🧵(1/8)