A Radical Plan to Make AI Good, Not Evil

> [Anthropic's] constitution includes ... "content that is ... in exceptionally poor taste, ..."

Imagine trying to sell any conventional product, like a car or a toothpaste, but before you get to the real sales pitch, you must convince your customers that it's not an evil car, or an evil toothpaste. Or selling a pen, but only if the customer promises they won't write anything with too poor taste.

Kind of doesn't matter. Anyone can download an LLM and train it do OSInt to spearfish high value assets at companies at a huge scale, ransomware and all. Or make millions of phone calls from desperate loved ones asking for cash. Or generate hundreds of thousands of hate filled comments on every forum imaginable. Or create millions of propaganda news articles to tip elections to favor an adversarial nation state. Or turn your social media pictures into pornography as black mail. Get people to sign horrendous contracts. The list goes on and on.

One company doing something ethically correct doesn't really matter. We need regulations and the infrastructure to enforce them. Otherwise a lot of people will be harmed so a few billionaires can gain more wealth.

It's Anthropic public relations/ marketing:

"OpenAI competitor Anthropic says its Claude chatbot has a built-in “constitution” that can instill ethical principles and keep systems from going rogue."