AGI may not have a crisp, agreed-upon definition, but it's generally understood to be a level of AI that matches human intelligence, i.e. more than just working autonomously on tasks.

One of my concerns here is that re-purposing existing terminology causes even more confusion than already exists in this area.

Another concern is with making claims that can't be reasonably substantiated, like that "most experts currently agree". They don't. The ones at Anthropic may agree, but there are lots of experts, and most don't agree that the sky is blue.

I like Anthropic and use Claude myself. Personally, I consider Anthropic much more trustworthy than, say, OpenAI. But all of these companies are fueled by investment dollars, which makes their claims inherently suspect.

I try to balance out commercially-motivated claims with academic work on the subject, which can really shine a light on the potential and the limitations of the technology.

Anyway, sorry for snapping at you. It was uncalled for. Have a good one.