/sci/ - Science & Math

>>15349010
As a person in tech who follows AI very closely, I think things are going to move slower in medicine because of all the regulations, and people not tolerating any mistakes.

The way it works today is that people do their work, and sometimes they ask GPT a question either to bounce some ideas off, or to answer a small thing.

The goal for the next couple years is to flip that around.
In a few years, we want people to collaborate with the bot to come up with a plan, then have the bot start doing 80% of the steps (all the tedious, boring, or easy to automate parts), and only consult the human when it's stuck on a small well-defined substep of the plan.
So instead of you doing the work, and asking questions here and there, the bot will do the majority of the work at high speed while you watch, and it'll and ask you to intervene here and there to solve the small things it doesn't know how to do yet.

Given that the current #1 trending code project is a hastily made first draft attempt to do just that (AutoGPT), at the current rate of progress we should have an actual working version less than 5 years from now. If it takes a full 10 years, that would be a surprise.

But all in all, you're probably pretty safe in the medical field, compared to other people. It's possible we'll have the first fully autonomous bots way before regulators approve of bots making medical decisions.
The fun part starts when the bot becomes able to earn money doing freelancer work on its own. Then it can start shell companies as one of the sub-steps of its plan, employ real people to build things in the real world. Then focus on growth, make your first million, fill in the blanks for what happens next.

That might happen before AIs take your job, in which case there's nothing to worry about, since the world will be too different for your career choice to matter a single bit.