This paper is about how AI vision models don't have human level performance, but also this: GPT-4 without any vision capabilities at all still does remarkably well just faking it, suggesting that, much like with words, there is more regularity in the visual world then we realize
MULTIMODAL AI
-
AI Breakthroughs: Music, Chips, ChatGPT Updates, Mistral LLM
By
–
Top stories in AI today: -Udio levels up AI music creation
-Meta unveils powerful new custom AI chip
-Edit images in ChatGPT with new interface
-Mistral releases massive new open-source LLM
-6 new AI tools & 4 new AI jobs Read more: http://
therundown.ai/p/musics-chatg
pt-moment
… -
Udio Becomes Music Industry’s ChatGPT Moment with AI Advances
By
–
The music industry just had its 'ChatGPT for music' moment with Udio. Plus, more developments in AI today from Meta, Mistral, Hubspot, ElevenLabs, and Microsoft. Here's everything you need to know:
-
GPT-4 Turbo Vision et Whisper timestamps pour synchronisation multimodale
By
–
Thanks for sharing! We just launched GPT-4 Turbo with Vision, and we recently added word-level timestamps in Whisper for precise text-audio synchronization. We’re excited about multi-modality, and we’d love to hear if this fits your use case!
-
Udio: AI Music Generation App with Text Prompting
By
–
— swyx 🐣 (@swyx) 10 avril 2024
Introducing Udio, an app for music creation and sharing that allows you to generate amazing music in your favorite styles with intuitive and powerful text-prompting. 1/11
-
AI Music Generation Quality Approaches Spotify Indistinguishability
By
–
I've been playing with @udiomusic for a few days now.
— Matt Shumer (@mattshumer_) 10 avril 2024
The quality of the music it generates is insane.
Stop whatever you're doing and listen to this.
It feels like we're less than a year away from AI music that is indistinguishable from what you hear on Spotify. https://t.co/xgdRKG2R8X pic.twitter.com/tn6gxIKt6HI've been playing with @udiomusic for a few days now. The quality of the music it generates is insane. Stop whatever you're doing and listen to this. It feels like we're less than a year away from AI music that is indistinguishable from what you hear on Spotify.
-
AI Model Trained on Visual Aesthetics of Poor Things
By
–
This model was trained on the visual and aesthetic sensibility of the movie "Poor Things." A lot of the objects in the scene are molded around that interpretation, which is something you just can't do with traditional text to image.
-
GPT-5’s reasoning and memory will accelerate AI agent development
By
–
The GPT-5 Moment and Its Implications for AI Agents – GPT-5 will have better reasoning, long-term memory, multi-modality, and emerging properties – As soon as GPT-5 is released, agent builders will go nuts, and development speed will be impossible to keep up with – Proper
-
EverArt AI Fine-Tuning Achieves Indistinguishable Photorealistic Image Generation
By
–
Our fine-tuned pipeline and the quality of output on @everartai have become so good that I can no longer tell whether these images are AI-generated or real. These are from our Poor Things fine-tuning. Also, you truly can't get these images in DALL·E or Midjourney.
-
GPT-4 Turbo with Vision: Function Calling Now Supports Image Inputs
By
–
We just launched GPT-4 Turbo with Vision, so you can now use function calling with image inputs. Looking forward to seeing what you build with it!