🥟 Chao-Down #283 ChatGPT can now read its answers aloud in 37 languages, India reverses its AI stance and requires government approval for product launches, How social media influencers shape AI

I’ve been diving more into the latest Claude 3 model and will be building more apps on top of it given impressive early results.

As evidenced by the benchmarks provided below, we’re already witnessing more records be broken with Claude 3 in its several variants (Opus, Sonnet, and Haiku). In particular, if Claude 3 can be a strong reasoning engine, it can pair its already impressive 200k+ context window to that to process

My colleague and I recently explored building agent architectures through our work with Microagents and found that context truly is king as a bigger token window allowed for more tools and reliable agent execution. We’ll have to see how much better Claude 3 is over GPT4 in the coming weeks.

Expect OpenAI to ramp up its GPT5 efforts.

-Alex, your resident Chaos Coordinator.

What happened in AI? 📰

ChatGPT can read its answers out loud (The Verge)

India reverses AI stance, requires government approval for model launches (TechCrunch)

AI Warfare Becomes Real for US Military With Project Maven (Bloomberg)

Wix’s new AI chatbot builds websites in seconds based on prompts (The Verge)

A Few Social Media Influencers Are Shaping AI (IEEE Spectrum)

Willy Wonka experience’s AI debacle is start of new advertising reality (fastcompany.com)

Always be Learnin’ 📕 📖

How LinkedIn Serves 5 Million User Profiles per Second (quastor.org)

How startups beat incumbents (asmartbear.com)

A Deep Dive into the Underlying Architecture of Groq's LPU (substack.com)

Projects to Keep an Eye On 🛠

bruin-data/ingestr: ingestr is a CLI tool to copy data between any databases with a single command seamlessly. (Github)

hundredblocks/transcription_demo - Using Claude 3 for video summarization (Github)

🧞 Genie: Generative Interactive Environments (Google)

The Latest in AI Research 💡

Panda-70M - A Large-Scale Dataset with 70M High-Quality Video-Caption Pairs (snap-research.github.io)

The current state of artificial intelligence generative language models is more creative than humans on divergent thinking tasks (Nature)

Why do Random Forests Work? Understanding Tree Ensembles as Self-Regularizing Adaptive Smoothers (arxiv)

The World Outside of AI 🌎

IRS Free File users jump nearly 10% from last tax season, IRS says (USA Today)

Phone screen protectors are becoming pointless (qz.com)

How to avoid PFAS in water, food with ‘forever’ chemicals (The Washington Post)

Millennials will soon be the wealthiest generation (qz.com)

Why your auto and home insurance premiums are surging (NPR)

How scientists are using sound waves to hack the brain (Free Think)

One Last Bite 😋