🥟 Chao-Down #237 Researchers find that Google Gemini is not as good as GPT-3.5 Turbo, AI models still have trouble reading SEC filings, The next generation of chips to power the AI wave

For readers of Chao-Downs who have been following, the Semantic Kernel team (the project I work on) just released the 1.0 SDK! It’s a huge milestone for us as it signifies our conviction that we’ve built a framework that developers and enterprises can rely on going forward.

It’s the outcome of a year-long effort that started with just a few of us in the Office of the CTO iterating on ideas on a whiteboard as we looked ahead to what the future of AI would look like and has grown immensely since.

Thank you to all our friends at Microsoft, our many enterprise partners, and the amazing open-source community for getting us to this milestone.

Check out the full blog announcement here. I’m excited to see what the community builds with Semantic Kernel heading into the new year!

-Alex, your resident Chaos Coordinator.

What happened in AI? 📰

Google Gemini is not even as good as GPT-3.5 Turbo (VentureBeat)

New generation of chips will drive the AI wave (ft.com)

GPT and other AI models can't analyze an SEC filing, researchers find (CNBC)

Tesla’s Self-Driving Tech Has Competition (WSJ)

You Can’t Truly Be Friends With an AI (The Atlantic)

Stability AI announces paid membership for commercial use of its models - The Verge

Always be Learnin’ 📕 📖

Spotify Wrapped: 6 psychology principles that make it go viral every year (growth.design)

What is a distributed database and when should you use one (Fauna)

The best growth advice of 2023 (growthunhinged.com)

Practices for Governing Agentic AI Systems (openai.com)

Projects to Keep an Eye On 🛠

Introducing Ego-Exo4D: A foundational dataset for research on video learning and multimodal perception (meta.com)

jbexta/AgentPilot: A cross-platform desktop app to create, manage, and chat with AI agents! + Multi agent chat, branching chat and multiple API providers (Github)

Writesonic/GPTRouter: Smoothly Manage Multiple LLMs (OpenAI, Anthropic, Azure) and Image Models (Dall-E, SDXL), Speed Up Responses, and Ensure Non-Stop Reliability. (Github)

The Latest in AI Research 💡

DriveMLM: Aligning Multi-Modal Large Language Models with Behavioral Planning States for Autonomous Driving (arxiv)

HyperRouter: Towards Efficient Training and Inference of Sparse Mixture of Experts | (OpenReview)

LLM in a flash: Efficient Large Language Model Inference with Limited Memory (arxiv)

Mathematical Language Models: A Survey (arxiv.org)

EdgeSAM: Prompt-In-the-Loop Distillation for On-Device Deployment of SAM (arxiv.org)

StemGen: A music generation model that listens (arxiv)

The World Outside of AI 🌎

All the changes coming to Google Play and sideloading following $700M settlement | TechCrunch

Why Japan's Nippon Steel is buying US Steel for $15 billion (qz.com)

NASA highlights first commercial delivery service to moon | Digital Trends

Here’s why the fediverse is the future of social networks, and the web - The Verge

Britain Ruined One of the Best Healthcare Systems in the World (The New York Times)

Older Americans are working more, earning more — and propping up economy (axios.com)

One Last Bite 😋

Why houses are so expensive, explained in one chart (axios.com)