I actually think the people developing AI might well not get rich off it.
Instead, unless there's a single winner, we will probably see the knowledge on how to train big LLMs and make them perform well diffuse throughout a large pool of AI researchers, with the hardware to train models reasonably close to the SotA becoming more quite accessible.
I think the people who will benefit will be the owners of ordinary but hard-to-dislodge software firms, maybe those that have a hardware component. Maybe firms like Apple, maybe car manufacturers. Pure software firms might end up having AI assisted programmers as competitors instead, pushing margins down.
This is of course pretty speculative, and it's not reality yet, since firms like Cursor etc. have high valuations, but I think this is what you'd get from the probably pressure if it keeps getting better.
Instead, unless there's a single winner, we will probably see the knowledge on how to train big LLMs and make them perform well diffuse throughout a large pool of AI researchers, with the hardware to train models reasonably close to the SotA becoming more quite accessible.
I think the people who will benefit will be the owners of ordinary but hard-to-dislodge software firms, maybe those that have a hardware component. Maybe firms like Apple, maybe car manufacturers. Pure software firms might end up having AI assisted programmers as competitors instead, pushing margins down.
This is of course pretty speculative, and it's not reality yet, since firms like Cursor etc. have high valuations, but I think this is what you'd get from the probably pressure if it keeps getting better.