AI Roundup 01/26 -> 02/02/2023

Last week's most interesting AI news in <2 minutes

Text-to-audio is the latest modality to gain traction:

1. Google Research introduce MusicLM, a transformer-based text-to-audio model capable of producing tracks of varying genres, instruments, and concepts (Paper's website)

2. Google researchers introduce SingSong, a system that generates instrumental music to accompany vocal inputs (Paper's website)

3. Anonymous researchers introduce Noise2Music, a diffusion model generating high-quality 30-second music clips from text prompts (Paper's website)

4. Researchers from ByteDance AI Lab introduce Make-An-Audio, a text-to-Audio diffusion model capable of transforming image and video to audio thanks to robust generalization (Paper's website)

5. The German Max Planck Institute releases Maûsai, a text-to-music generative model capable of long-context high-quality music (Paper)

---

6. Meta researchers introduce MAV3D (Make-A-Video3D) the first generative model capable of rendering 3D dynamic scenes from image and textual prompts (Paper's website)

7. Microsoft researchers present BioGPT, a generative pre-trained transformer model trained on biomedical literature outperforming other scientific LLMs on par with human expert performance (Twitter)

8. OpenAI introduces a web-based tool to distinguish between AI-generated and human-written texts to fight plagiarism and harmful campaigns at scale (OpenAI)

9. Salesforce Researchers and researchers at UCSF and Berkeley introduce ProGen - an LLM capable of generating functional protein sequences (Salesforce Blog)

10. Salesforce Research presents BLIP-2, a state-of-the-art vision-to-language model capable of zero-shot image-to-text generation and VQA (Paper)

11. Google is testing a ChatGPT-like competitor called Apprentice Bard powered by Google's LLM LaMDA, which unlike ChatGPT has knowledge of recent events (Business Insider)

12. Microsoft researchers introduce FLAME, a Formula-Language Model for Excel capable of repairing, auto-completing, and syntax reconstruction of formulas (Paper)

13. OpenAI publishes a paper on scaling laws for singal-agent reinforcement learning (Paper)

14. Madrona releases a comprehensive overview of the Generative AI space with an outlook to the future of Foundation Models (Madrona blog)

15. Microsoft plans to embed GPT-4 along with an elevated ChatGPT version into Bing's search, challenging Google's dominance in Search (TechCrunch)

16. CarperAI releases a new diffusion model, capable of generating code changes with commit messages (CarperAI blog)

Last week’s AI Tidbits