For one thing they're already at human performance.
For another, i don't think you realize how expensive inference can get. Microsoft with no scant amount of available compute is struggling to run gpt-4 such that they're rationing it between subsidiaries while they try to jack up compute.
So saying, it would be economically sound if it cost x10 or x100 what it costs now is a joke.
This tells me you haven't really stress tested the model. GPT is currently at the stage of "person who is at the meeting, but not really paying attention so you have to call them out". Once GPT is pushed, it scrambles and falls over for most applications. The failure modes range from contradicting itself, making up things for applications that shouldn't allow it, to ignoring prompts, to simply being unable to perform tasks at all.
We have given it extensions, and really the extensions do a lot of the work. The tool that judges the style and correctness of the text based on the embedding is doing much of the heavy lifting. GPT essentially handles generating text and dense representations of the text.
Still waiting to see those plugins rolled out and actual vector DB integration with GPT 4, then we'll see what it can really do. Seems like the more context you give it the better it does, but the current UI really makes it hard to provide that.
Plus the recursive self prompting to improve accuracy.
How are they at human performance? Almost everything GPT has read on the internet didn‘t even exist 200 years ago and was invented by humans. Heck, even most of the programming it does wasn‘t there 20 years ago.
Not every programmer starting from scratch would be brilliant, but many were self taught with very limited resources in the 80s form example and discovered new things from there.
GPT cannot do this and is very far from being able to.
Because it performs at least average human level (mostly well above average) on basically every task it's given.
"Invest something new" is a nonsensical benchmark for human level intelligence. The vast majority of people have never and will never invent anything new.
If your general intelligence test can't be passed by a good chunk of humanity then it's not a general intelligence test unless you want to say most people aren't generally intelligent.
I would argue some programmers do in fact invent something new. Not all of them, but some. Perhaps 10%.
Second the point is not whether everyone is by profession an inventor but whether most people can be inventors. And to a degree they can be. I think you underestimate that by a large margin.
You can lock people in a room and give them a problem to solve and they will invent a lot if they have the time to do it. GPT will invent nothing right now. It‘s not there yet.
For one thing they're already at human performance.
For another, i don't think you realize how expensive inference can get. Microsoft with no scant amount of available compute is struggling to run gpt-4 such that they're rationing it between subsidiaries while they try to jack up compute.
So saying, it would be economically sound if it cost x10 or x100 what it costs now is a joke.