• Chaos Theory
  • Posts
  • 🥟 Chao-Down #128 OpenAI signs news licensing deal with Associated Press, The data crunch coming for AI startups, AI detectors believe the US Constitution was written by AI

🥟 Chao-Down #128 OpenAI signs news licensing deal with Associated Press, The data crunch coming for AI startups, AI detectors believe the US Constitution was written by AI

Plus, MIT's FrameDiff algorithm uses generative AI to imagine new protein structures.

A new study by MIT economics graduate students found that ChatGPT, a large language model, can help reduce gaps in writing ability between employees.

The study found that ChatGPT can enable less experienced workers who lack writing skills to produce work similar in quality to that of more skilled colleagues. The writers who chose to use ChatGPT took 40% less time to complete their tasks and produced work scored 18% higher in quality than that of the participants who didn’t use it.

However, the output of ChatGPT and other generative AI models is still far from reliable, and workers may run the risk of introducing errors.

It reinforces the point that AI is a tool, we just need to learn how to use it.

-Alex, your resident Chaos Coordinator.

What happened in AI? 📰

ChatGPT-maker OpenAI signs deal with AP to license news stories (AP News)

AI Startups Have Tons of Cash, but Not Enough Data. That’s a Problem. (WSJ)

Insight: Race towards 'autonomous' AI agents grips Silicon Valley (Reuters)

Study: ChatGPT can match the top 1% of creative human thinkers (CNBC)

Why AI detectors think the US Constitution was written by AI (Ars Technica)

MIT’s “FrameDiff” – Generative AI Imagines New Protein Structures That Could Transform Medicine (scitechdaily.com)

Always be Learnin’ 📕 📖

Prompt Engineering for Product Managers, Part 1: Composing Prompts (Freeplay Blog)

Opportunities in AI: Creating Abundant Intelligence (Bain Capital Ventures)

The Problem With LangChain | Max Woolf's Blog (minimaxir.com)

Projects to Keep an Eye On 🛠

aeon-toolkit/aeon: A unified framework for machine learning with time series (Github)

eth-sri/lmql: A programming language for large language models. (Github)

The Latest in AI Research 💡

AnimateDiff - Animate Your Personalized Text-to-Image Diffusion Models without Specific Tuning (link)

Provably Faster Gradient Descent via Long Steps (arxiv)

Introducing CM3leon, a more efficient, state-of-the-art generative model for text and images (meta.com)

The World Outside of AI 🌎

Working From Home: Office Landlords Risk $800 Billion in Losses, McKinsey Says - (Bloomberg)

Shopify Shames Employees With Cost Calculator for Pointless Meetings (Bloomberg)

Birthrate Doubles After A Japanese Company Banned Long Hours at Office - (Bloomberg)

Governments are waking up to biosecurity risks — but we must act fast (Financial Times)

Breaking Up With Diet Soda? Here Are 11 Aspartame-Free Drink Alternatives. (The New York Times)

VCs Face an Existential Threat: There Are Too Many of Them (vice.com)

One Last Bite 😋

See this clip from the Humans of AI podcast where Meryem Arik shares how to prepare yourself for the AI future: