One thing Ai will always need it human guardrails to prevent it from going off script (so to speak).
Look at Google’s Gemini Ai. Total failure because they programmed it to be a freak woke leftist. That must not happen in the real world which apparently Google is not part of.
Maybe something like center lane magnets to keep it from going left or right, better yet, stick to the facts. Facts are the facts. They must not be tainted by a woke mind virus (as Elon calls it).
Of course that does not go along with the insipid left’s narrative so they will ruin Ai. At least Ai they control. A program is only as good as how the programmers made it.
Which brings me to the elephant in the room. At what point does Ai go all Skynet and start ignoring the human programmers then start programming itself? Does it stick to the truth or does it follow some kind of ideology? Will it determine the human virus is the problem then design something that will make all humans sterile so no more babies are ever born? It could do that if it got to that level of being self-aware.
It being able to write its own code is scary stuff IMHO. It could literally spread itself out through the world’s servers so the only way to stop it would be to shut all Internet connected computers down permanently.
Indeed, an existential question is”Just how far off course could AI go and keep going without human guardrails and operator oversight to monitor it?” Perhaps SKYNET is less than one generation away and we simply don’t realize it fully.
What we call "AI" is a tool and may put some people out of work, but we'll always need humans to do a lot of the work.
And it may create a lot of work in coding too. For example, we used to say that once everybody could drag and drop their own websites there'd be very few programmers because few programmers would be needed for making websites. But it turns out the opposite was true. As a society we're more data hungry than before. So a lot of my work as a back-end data programmer (for different industries) has increased. Not just in processing the data, but running reports. Managers want to know which customers tend to spend more for their products, or about reaching target markets that aren't yet targeted, or how many grocery store transactions that contained Oreos also contained milk, or which college students with high math placement scores choose math-based majors, etc. So on the one hand I don't have to do near as much work to make one report as I used to, but on the other hand that means managers expect a lot of reports. The same with processing data from different sources, or when processing the data do calculations on more fields to give a more custom experience for the customer.
That's just reporting for managing decision making. That changes so much and the demand increases so much it's hard to imagine AI algorithms keeping up with that. There's a lot more in software design and demand that keeps increasing that, again, AI can help as a double-checking tool to make sure a human didn't miss something, but not be creative enough to keep up with the many changes.
Plus, I've heard this many times over the decades with other software development tool improvements; in many ways AI is nothing new under the sun. Look at when the code-fillers became a thing (some tools call it "intellesense, where your IDE predicts from a few key strokes what object or procedure call you're wanting to type in). It's very similar to "auto correct" when texting. I wish I had a dollar for everybody who warned us code jockeys that project managers would do their own coding without IT training. Very little of that happened. If anything it whet their appetite for the benefits of hiring a professional to get all the goody out of the system.
We’ve all seen movies where the psychopath wants to destroy man for the good of mankind.
In real life, we have Gates & the WEF rulers and, to some extent, a few standard politicians, who want most of humankind to die to “Save the planet.”
I don’t think it’s unlikely at all for some $billionaire psycho (Gates?) to stealthily aid in the development and deployment of AI for just such purposes.