Weekly Digest: April 17, 2024
This week, big news from OpenAI and Adobe, why so many AI companies love Japan, a wave of new multimodal models, and more.
Welcome to Salt AI's Weekly Digest, where Alex Duffy, our VP of AI, summarizes the latest updates you should know across the AI industry, ComfyUI, and Salt AI. This week, big news from OpenAI and Adobe, why so many AI companies love Japan, a wave of new multimodal models, and more. Watch the video above and follow along with Alex's notes and links below:
Big News
OpenAI Japan - first office in Asia and releasing a GPT-4 custom model optimized for the Japanese language
Earlier this year Sakana raised $30m and Japan passed a law allowing training models on anything, copyrighted or not
Meanwhile countries like France are also establishing themselves with HuggingFace and Mistral but the EU just passed the EU AI Act
They also announced batch pricing, a 50% discount on all tokens and updates to their assistants API
Stanford AI Index — a 300-plus-page report tracking 2023’s worldwide trends in AI released
Represented in 13 charts , I’d also include this one as far as shipments from NVIDIA
AI in Design
Adobe adding generative AI including directly implementing external companies like OpenAI, Pika Labs and Runway
Coinbase using AI for design
Reminder, last year Figma announced acquisition of Diagram with a post called AI: The next chapter in design and since have been teasing that they’re working on tools to turn designs into code with AI & integrated it with FigJam
New in LLMs
Multimodal continuing to push forward
Twitter releases multimodal Grok and Elon Xeet’s about how they are posed to collect quality data with Tesla self driving & touches on synthetic data, the idea of using LLM’s to generate or improve data
Microsoft released and then unreleased their open source models that had people excited, but the cat is out of the bag with their license
HuggingFace releases their mutlimodal model fully opensource with public dataset
Shanghai AI lab released theirs too
With some convincing, GPT can be persuaded that your pretty awesome
Continuing the thread of evaluation, one is official medical boards
Some really great prompting skills wins a competition & $10,000
New in Generative AI
Naval released a social app where you can only talk as input
New viral videos:
C3P0 x Childish Gambino made with AI music, images, video, and more
For those that havn’t seen it there were some other classics like Harry Potter Balenciaga and Lord Of The Rings by Wes Anderson also all AI
AI replacement of people created a viral meme that was made with comfy!
New In ComfyUI
Create viggle person replacement and community working on improving it
Voice cloning in Comfy!
Workflows that use videos to drive content
Head to our Discord to let us know what you think of this week's newsletter and what you'd like to see included in future editions.
