Anthropic upkeeps a huge document they call Claude’s Constitution which is used when training Claude in order to shape its behavior intentionally. It is meant to shape Claude’s behavior and personality. The major change in the new constitution is a shift from being a list of rules to more explanations of broad principles and reasons why they should be followed- such as being honest and helpful. The explanations of why Claude should follow these principles reflects Anthropic’s “uncertainty” that Claude is “conscious or has a moral stance.” Yes, they actually are not sure if they are creating something with self determination.

OpenAI is exploring new pricing models, a sign they are growing their focus on revenue growth and profit. Last week, they announced they plan to test advertising. And at Devos the OpenAI CFO said during her talk that during use cases of ChatGPT like drug discovery that OpenAI could get “a license to a drug that is discovered.” This way of monetization would offer a new pricing model for enterprises where instead of paying a regular subscription they pay in partial ownership of the value created in their use of OpenAI products. Looks like the $1B in ARR added to OpenAI in past month is not enough.

Price Per Token now is active on social media, you can follow here:

X: https://x.com/pricepertoken

BlueSky: https://bsky.app/

Threads: https://www.threads.com/@pricepertoken

The Playground has now been online for about a month. It is getting decent usage so I am sharing the past models on it during the past week below:

That’s all for now!