Which AI Models Are the Most Transparent?

Plus Universal sues Anthropic

The AI Breakdown First Five - Wednesday October 19, 2023

Today on the First Five:

  • 5. FBI Director Says AI Has Been Used to Amplify Terrorism

  • 4. Amazon’s Latest Impressive Robotics

  • 3. OpenAI’s New MidEast Partnership

  • 2. Universal Music is Suing Anthropic

  • 1. Stanford’s AI Transparency Ratings

Amazon’s Digit

5. FBI Director Says AI Has Been Used to Amplify Terrorism

Terrorist propaganda has gotten worse thanks to AI. That’s according to FBI Director Christopher Wray who said at a recent intelligence summit. He said that non state organizations are figuring out ways to jailbreak AI, and also had some choice words for China.

4. Amazon’s Latest Impressive Robotics

Did you know that Amazon now has more than 750,000 robots working alongside human employees in their factories? They’ve just announced a new order fulfillment robot named Sequoia as well as a bipedal robot named Digit.

3. OpenAI’s New MidEast Partnership

Even as the US is clamping down on AI chip sales to the Middle East (for fear of them ending up with China), private AI labs are expanding their relationships in the region. OpenAI has announced a partnership with G42 to deliver OpenAI models in industries including energy, healthcare and financial services.

2. Universal Music is Suing Anthropic

According to Universal Music, Anthropic is ripping off lyrics from their artists like The Rolling Stones and Beyonce. Accusing the AI lab of widespread copyright infringement, UMG has filed a $75M lawsuit. AI safety advocates don’t need an official pause; legal complaints around copyright are going to slow things down all on their own!

1. Stanford’s AI Transparency Ratings

One of the challenges about LLMs is that we just don’t exactly know how they work. Some are even more opaque than others, and with that in mind, Stanford has released a “Foundation Model Transparency Index” which rates 10 large models including GPT-4, PaLM 2, LLaMA 2 and others on 100 criteria ranging from disclosure around sources of training data to hardware used to train it. Click here to see which was the most transparent.

BONUS: Midjourney upscales update

The Latest AI Breakdown Episodes

Thanks for reading! How did this edition hit you?

How was today's newsletter?

Login or Subscribe to participate in polls.