• AI Time to Impact
  • Posts
  • . . AI: ChatGPT turns 1, Alibaba Cloud launches huge LLM, and much more (11.30.23)

. . AI: ChatGPT turns 1, Alibaba Cloud launches huge LLM, and much more (11.30.23)

Author

Marshall Kirkpatrick
November 30, 2023

Friends, today ChatGPT turns 1, Meta’s Fundamental AI Research (FAIR) turns 10, and a whole bunch of cool new projects were born - just today!

As always, this is a curated collection of the top stories in AI according to a weighted analysis of AI community engagement. Every one of them should offer some inspiration, perspective, or utility, I think. But give them a scan and see what you find!

Thanks for reading and sharing this newsletter,

First impacted: AI users and developers
Time to impact: Short and Long

ChatGPT, OpenAI's chatbot, celebrates its first birthday today, with CEO Sam Altman and the AI community discussing its influential role over the past year. Can you believe it's only been 1 full year? [via @sama] Share by email

First impacted: AI developers, AI researchers
Time to impact: Short

The Alibaba Cloud team has introduced two open-source LLMs, Qwen-72B and Qwen-1.8B, which immediately caught the attention of developers watching the largest open source models. These are really big. There's a very fun chat prompting riddle game on the Qwen HuggingFace page too - how many levels can you complete? [Qwen] Share by email

First impacted: AI researchers, AI developers
Time to impact: Short to medium

Meta's Fundamental AI Research (FAIR) team is marking its 10th anniversary, having worked on various projects such as pytorch, detectron, and llama. To commemorate this milestone, they are showcasing three new research projects: the Ego-Exo4D dataset of video perception for skilled human activities, the Seamless Communication family of AI models for vocal style preserving near-real tme streaming language translation, Audiobox, a foundation model for audio generation of voices and sound effects. [via @AIatMeta] Share by email

First impacted: AI developers, AI researchers
Time to impact: Short to medium

The PyTorch team says they've created an LLM that's nearly 10 times faster than industry standards, using PyTorch's own optimizations and established techniques like int8 weight-only quantization to address memory bandwidth issues. [Accelerating Generative AI with PyTorch II: GPT, Fast] Share by email

First impacted: Artificial Intelligence Researchers, Optical Illusion Artists
Time to impact: Short

University of Michigan scientists have developed a technique for creating multi-view (eg flipped upside down) optical illusions using pretrained diffusion models, which involves calculating noise in various image views and averaging them. The demo of this is amazing! It's hillarious that there's a machine that can just pump these things out now! [Visual Anagrams: Synthesizing Multi-View Optical Illusions with Diffusion Models] Share by email

First impacted: AI lobbyists, European citizens
Time to impact: Medium to long

The EU's AI Act, aimed at regulating AI, is under pressure from lobbyists seeking to loosen controls on AI foundation models and move regulation to the application layer. Despite resistance from France, Germany, and Italy, a YouGov poll indicates that over two-thirds of European citizens back the regulation of these models, with Spain, the current EU presidency holder, advocating for a revised AI Act mandate. [Experts urge EU governments to regulate AI instead of companies] Share by email

First impacted: Content creators, Social media marketers
Time to impact: Medium

Forbes: "Former Snap software engineer Joshua Xu believes AI-generated video is about to have a moment like Snapchat or Instagram had in the early days of the mobile photography revolution. "As early proof of that, he points to his own company HeyGen. After launching its AI-powered video creation app last September, HeyGen reached $1 million in annual recurring revenue in March, then $10 million in October. Today, that number is up to $18 million, Xu, cofounder and CEO, told Forbes." [AI Video Startup HeyGen Launches Near-Instant Avatar Generator, Adds $5.6 Million In Funding] Share by email

That’s it! More AI news tomorrow!