"Historically, these sorts of disruptions lead to short-term issues, and long-term employment growth"
I think the pace of AI advances may turn these disruptions into long-term issues and not lead to employment growth.
Yes, ChatGPT can be a multiplier on human productivity. But that's assuming you can learn how to use it correctly. The author here seems to know how to scale the complexity of the prompts in such a way that they get meaningful output (a complete website). That sort of prompting is a skillset in itself.
The problem as I see it is: how long is that skillset relevant before an even more advanced LLM comes along, forcing you to re-learn how to interact with it. Now consider that the next advancement need not come from an LLM, but some entirely different system that doesn't work via prompts. That skilled prompter needs a new skillset.
Technological innovations of the past have always come slowly enough that people could learn and master them before something new came along. It seems like things are moving so quickly in the AI space that may no longer be true.
With all that said, an example that gives me hope is chess. The rise of chess AI's has not killed chess in any sense. The players of the last few years are actually much stronger than pre-AI. Can the best players in the world beat Stockfish or Alpha-Go? Not even close. And so far that hasn't mattered. People still prefer to watch two human players.
It's a very important point that the insanely rapid pace of change is itself a huge problem. We're still in the same (northern hemisphere) school year in which ChatGPT was released; teachers and students just started trying to understand its impact; and the underlying model has already been replaced with something much more powerful --- TWICE --- GPT4 and "tool-using" mode.
We’re just starting the first “performance cycle” at work since ChatGPT came out and I strongly suspect next year’s won’t be based on producing so many written documents (self reflection, peer feedback, etc).
I was thinking about the chess analogy too, and I don't think it holds. Chess is a pastime; people do it for fun. Once AI has consumed all our jobs, sure, you can manually tap out some code for fun, but no one is going to pay you for that.
I think the pace of AI advances may turn these disruptions into long-term issues and not lead to employment growth.
Yes, ChatGPT can be a multiplier on human productivity. But that's assuming you can learn how to use it correctly. The author here seems to know how to scale the complexity of the prompts in such a way that they get meaningful output (a complete website). That sort of prompting is a skillset in itself.
The problem as I see it is: how long is that skillset relevant before an even more advanced LLM comes along, forcing you to re-learn how to interact with it. Now consider that the next advancement need not come from an LLM, but some entirely different system that doesn't work via prompts. That skilled prompter needs a new skillset.
Technological innovations of the past have always come slowly enough that people could learn and master them before something new came along. It seems like things are moving so quickly in the AI space that may no longer be true.
With all that said, an example that gives me hope is chess. The rise of chess AI's has not killed chess in any sense. The players of the last few years are actually much stronger than pre-AI. Can the best players in the world beat Stockfish or Alpha-Go? Not even close. And so far that hasn't mattered. People still prefer to watch two human players.