Getting therapy via an improv comedy group sounds like the sort of thing that could actually be hugely effective.

Laws are social constructs. They are mere threats of violence

“in test of”???

Sue Kavanaugh

Oh, they had a reason alright. Just not a non-racist one.

So, interestingly, LLM companies now have to search for “clean” data, which is sometimes Internet writing created before a certain date, and they may have to do that forever.

Also! Because of cases like this, where the data itself might be LLM-generated, there’s a fear of “model collapse” by training an AI on AI data (which progressively makes the LLM predictions worse and worse), so there’s an increasing wariness of using anything after the popularity of AIs as LLM data

The models are pretty “baked in” with each release and can’t be changed. It can redirect sometimes to web searches and summarize, but the data it’s trained on is never recent.

That’s not how ChatGPT works