Why Both the Hype and the Fud over AI Are Overblown
medium.com
medium.com
Not really. It is sufficient that it makes most programmers orders of magnitude more productive. That will quickly restructure most teams across the industry, with most engineers dealing with AI pipelines end-to-end, which are as productive as an entire team. My current thinking is that the broad effect will amount to polarizing engineers into AI specialists and product specialists. It's still possible that the same amount of people are needed for this shift, but no doubt it will shake up the entire industry and force people to adapt. That said, I don't know how soon it will happen, but when it does it will probably happen relatively quickly.
It doesn't, do that though. It can currently only generate (bad) code for trivial problems.
Honestly, I don't think the main threat is AI generating code anyway. I think it's AI as the middleware layer. There are a lot of developers that are gainfully employed as JSON shepherds -- herding JSON from one system to another with some business logic sprinkled in. I think there are some systems where a hypothetical language model that was extremely fast and reliable would be perfectly capable of directly handling a typical API request in realtime. And the programming of these systems would require little more than crafting a prompt (or series of prompts). Of course, right now LLMs are too slow, too expensive to operate and they hallucinate too much so people won't use them for that use case today. But in 10 years? Who knows...
And clearly, AI won't replace EVERY developer but it does seem as if there might be a future where fewer developers will be needed (in the sense that the growth in demand for developers will slow...maybe to a crawl).
Any good coder will avoid writing code manually if at all possible. Our tool sets are filled with code generators that blow Copilot out of the water. We're constantly trying to automate our jobs wherever possible.