Twitter Beğeni Hilesi

We release PaliGemma. I'll keep it short, still on vacation: - sota open base VLM designed to transfer quickly, easily, and strongly to a wide range of tasks - Also does detection and segmentation - We provide lots of examples - Meaty tech report later! ai.google.dev/gemma/docs/pal…

Congrats to the GDM Astra team! Super impressive video, audio, and long-context capabilities - can't wait to play with it 😀 (in the meantime, love the demo, though can't help but feel some secondhand awkwardness on behalf of the coworkers 🤣)

Video and audio go really well together! Feels truer than ever now 😀

GPT-4o (o for “omni”) is the first model to come out of the omni team, OpenAI’s first natively fully multimodal model. This launch was a huge org-wide effort, but I’d like to give a shout out to a few of my awesome team members who made this magical model even possible!

GPT-4o was a whole team effort. Huge credit especially to Prafulla Dhariwal for the conviction to build an omni model, and helping see it through collaboration with many teams at OpenAI over the past 18 months. Also, he's excellent at trivia:

the first time gpt-4o spoke back to me in real-time, it became clear that we built something completely new – and that what we are building is the future of human-computer interaction. come build this real-time future with us. openai.com/careers/real-t…

We're hiring Real-Time Communications Engineers to help us scale and safely deploy multimodal voice models. Apply now:

I'll be in Seattle for #CVPR2024 (Tues June 18/Wednesday June 19) If you'll be there and interested in realtime video+audio, then we should meet! DM me if so 😀

I'm going to #CVPR2024 next week. If you are in Seattle and interested in broad multimodal intelligence or real-time multimodality, we should talk. DM me here or send an email.

Multimodal OpenAI is out here #CVPR2024 this week. We have one gptv/gpt4o talk tmrw (Tuesday) @9am. DMs are open. Come find us to chat about AGI, multimodal, vibes, or hiring. Deets about the talk and who's here in thread 👇.

Excited to announce that the Embodied AI team at Alana developed AlanaVLM, a new foundation model for egocentric video understanding. Best open-weight model which rivals GPT-4V on spatial reasoning questions in OpenEQA. Thread below! arxiv.org/abs/2406.13807 #AI #NLProc

Roman Huet, Head of Developer Experience at OpenAI, shows a live demo of how GPT-4o can interact with the world through a webcam, understanding what it sees and reads

congrats kyutai on the Moshi demo moshi.chat - the latency is impressive! Curious to read the tech report! Capabilities wise, the current model feels a bit like the early days of LMs 😅. So it'll be useful to have this model to build on + make benchmarks for.

Who all is going to be at ECCV in-person?

🍓is ripe and is ready to think, fast and slow: check out OpenAI o1, trained to reason before answering I joined OpenAI to push boundaries of science & reasoning with AI. Happy to share this result of team's amazing collaboration does just that Try it on your hardest problems

O1 is much better at code (and especially at writing docstrings, a trick Vlad Fomenko showed me!). super excited to preview it!

As a coach for the US IOI team, I’ve been motivated for a long time to create models which can perform at the level of the most elite competitors in the world. Check out our research blog post - with enough samples, we achieve gold medal performance on this year’s IOI and ~14/15