I'm not sure what would be a better bet? It's not like automation / the economy / capitalism aren't also coming for other workers. Maybe a good time to retrain as a bailiff?
ChatGPT was based on GPT3, which was released in 2020. The fundamental advance appears to have been InstructGPT which was 2022. GPT4 was 2023, three years after version 3. Id expect to see GPT5 sometime in 2026.
This thing is out in the public now and has massive mindshare. Funding and intellectual talent is flowing in at a much faster pace. The big incumbents are now under pressure from their shareholders to come up with competitors.
I'll be very, very surprised if the growth rate doesn't accelerate from here.
It depends on what improved GPT-4 relative to 3. If it was adding more data, then we'd expect progress to speed up in the short-term and then come to a screeching halt as we run out of text.
If it was algorithmic/model improvements then it could be much quicker to improve or much slower.
I think the big problem is that nobody (except Open-AI) knows what GPT-4 does and how it does it, so it's hard to predict.
I'm generally a fan of a naive model given the uncertainties around this, hence my prediction of three years.