AI Roundup 02/09 -> 02/16/2023

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

1. Stability AI introduces MedARC, an open-source research organization focusing on developing foundation models for medical AI research (Twitter)

2. World's 7th largest law firm announces it will start using Harvey, a generative AI software that builds custom LLMs for law firms to draft contracts, client memos, etc. (Bloomberg Law)

3. Richard Socher, the founder of the search engine YouChat, launches a multimodal conversational AI that can generate text answers, videos, and other data in response to textual prompts (VentureBeat)

5. Researchers at Stanford University introduce ControlNet, an open-source neural network structure with the goal of fine-tuning Stable Diffusion models (Paper)

6. Researchers from Meta AI introduce Toolformer, a language model capable of teaching himself when and which external tools such as calculators and Wikipedia to use to generate accurate answers, outperforming GPT-3 on selected tasks (Paper)

7. Stanford researchers announce H3 (Hungry Hungry Hippos), a generative language model that outperforms GPT-Neo-2.7B in parameter scale and a much longer context (Paper)

8. Nvidia researchers introduce Re-ViLM (Retrieval-augmented visual language model), a novel approach that boosts performance for image-to-text generation tasks, especially for zero-shot and few-shot generation (Paper)

8. Researchers from Baidu introduce ERNIE-Music - the first text-to-music generation model in the waveform domain using diffusion models (Paper)

9. Google researchers showcase ViT-22B - the largest vision transformer model at 22B parameters (Paper)

10. Google researchers present MaskSketch, a method that uses sketch guidance to define the desired image structure, demonstrating highly realistic images for a given sketch-to-photo (Paper)

11. Researchers from UC Berkeley propose Hindsight Instruction Relabeling (HIR), a novel algorithm for aligning language models with instructions (Paper)

Last week’s AI Tidbits