People talk about 'AI' as if stackoverflow didn't exist. Re-inventing the wheel is something that programmers don't do anymore. Most of the time, someone somewhere has solved the problem that you are solving. Programming earlier used to be about finding these solutions and repurposing them for your needs. Now it has changed to asking AI, the exact question and it being a better search engine.
The gains to programming speed and ability are modest at best, the only ones talking about AI replacing programmers are people who can't code. If anything AI will increase the need for more programmers, because people rarely delete code. With the help of AI, code complexity is going to go through the roof, eventually growing enough to not fit into the context windows of most models.
> Now it has changed to asking AI, the exact question and it being a better search engine.
Except that you get mostly the wrong answers. And it's not too bad when it's obviously wrong or you already know the answer. It is bad and really bad when you're noob and trying to ask AI about stuff you don't know yet. How would you be able to discern a hallucination from statistics bias from truth?
It is inherent problem of LLMs and no amount of progress would be able to solve it.
And it's only gonna get worse, with human information rapidly being consumed and regurgitated in 100x more volume. In 10 years there will be no google, there won't be the need to find a written article. Instead, you will generate a new one in couple clicks. And we will treat as truth, because there might as well not be any.
The gains to programming speed and ability are modest at best, the only ones talking about AI replacing programmers are people who can't code. If anything AI will increase the need for more programmers, because people rarely delete code. With the help of AI, code complexity is going to go through the roof, eventually growing enough to not fit into the context windows of most models.