> Or is it all bogus (with justification), and we're fine?
This is mostly my take. We are at the stage of truck drivers ten years ago who might have been freaking out about self-driving taking their jobs, and here we are in 2022 with a truck driver shortage.
I don't think GPT can be useful to the point of replacing software engineering without a consistent mental model of itself, time, and the world, and I predict it will approach the limits of what advanced-search pattern-matching can do without getting anywhere near those AGI capabilities.
I think these are much different. Trucking depends on humans due to infrastructure and complex problems of law, vision, last mile details, etc. I agree about near term limitations of this kind of model, but I think highly contextual and refined models for software will prove to be powerful enough (edit: by powerful enough, I mean powerful enough to transform how people write software) in many cases.
Software reiterates a lot, and generated solutions can be quickly vetted. A truck driving to the wrong destination or locking up en route is a much larger issues than a few seconds spent determining that a generated solution has a bug or doesn’t match a spec exactly.
If AI can stand on the shoulders of giants and people can vet its outputs, I’m fairly sure it will become more capable and safer to implement than self driving trucks, much faster.
The question I have is that if we can build more, faster, will we run out of work or will more people simply make more things?
There are huge incentives for wealthy companies to run more and more code on their infrastructure. Can we do more business digitally? Will it scale to provide more programming work, even if it’s heavily AI-assisted?
Well sure. That's why we have Wordpress. Javascript frameworks. And ten thousand other things. All the plug-ins of the last ten years have made me a more productive developer. But it hasn't reduced the demand for developers.
This is mostly my take. We are at the stage of truck drivers ten years ago who might have been freaking out about self-driving taking their jobs, and here we are in 2022 with a truck driver shortage.
I don't think GPT can be useful to the point of replacing software engineering without a consistent mental model of itself, time, and the world, and I predict it will approach the limits of what advanced-search pattern-matching can do without getting anywhere near those AGI capabilities.