I'd more than happy to try to replicate this. Because I'd love to be proven wrong on this (it would ease some anxiety around the subject). Let me know if you're willing to do it together; we can set up the parameters for how it should happen.

Touché. While I agree that is certainly the case in some instances. If a spokesman or individual say that they get utility, should we not take that as face value?

Ah yes, you had GPT-3 via api in 2020, my mistake.

The extensive usage of claude code and other LLMs seems to tell another story. I doubt people/companies would pay for something that they don't find useful.

(whether that is in their interest at the end of the day is a different story).

In some way, art and writing is good because it is _for_ humans by other humans, drawing from their own experience. That is what gives it value (to the other humans).

Code, could be for reading by other developers. Though it seems the industry at large don't care too much about that.

So LLMs had their public beginnings (effectively) in 2020 with the release of ChatGPT, a history of roughly 5 years. Around 2022 they started to have some kind of utility.

Do you think they will plateau here? I definitely think there is a non-zero chance they will continue to improve.

Some recommendations for you if you're in kyoto :).
maps.app.goo.gl/ik8SYN7viFhB...

Tbh, not sure I buy that premise.

For instance a sleepwalking person pours orange juice into a plant instead of water. They have no conscious intent, it is still a mistake that harms the plant. Meaning mistakes can occur without any intention, conscious or otherwise.