Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!

 



Forgot your password?
typodupeerror
×

Comment Now who would have guessed (Score 2) 199

I (as well as anyone with half a brain) has said so from the get go. That people who have the skills, have the projects to point to, have the certifications and have the bite will just flip them off and move on. What stays is the dregs, the bottom feeders who know that they can't jump ship because they have nothing to show for, haven't improved their skills since they were hired and know that nobody would hire them, so they have to bite the bullet.

We've been preaching this for what now? 2 years?

Maybe I should start consulting these dimwits hiding behind fancy three letter titles starting with C in something other than security. Apparently I'm superior to their knowledge in business administration as well. Likely because I never studied BA and thus evaded the mandatory lobotomy in the third semester.

Comment Re:How good is it? (Score 2) 28

That's one part of the problem.

Another, far more serious one, is that the input quality is deteriorating with every generation of AI. The first AI models only had human generated input to digest. Granted, some of that was complete drivel, but in general, the information level was pretty good. Sure, you also had conspiracy nuttery running rampart, but it was clearly labeled as such because conspiracy nutters usually label it THE TRUTH or some similar bull, so there's a consistent pattern that AI can latch onto.

The output AI generated was, well, hit or miss. It may be ok, it may be good or it may be one of the dreaded "hallucinations". Output that looks ok at face value but when you read on, you notice that it's complete garbage. Not just when it comes to accuracy, but simply weird, random ramblings of a madman. Something you'd get from the diary of an inmate of a mental asylum. It was hard to tell that from the rest, though.

And what's even harder is to tell AI generated content from human generated content. It's very hard to detect it with automated tools (like, say, AI), as we have seen with the difficulties universities had with students using AI to write their papers.

What adds to the problem is that AI is way faster at generating content than humans. Actually, faster even than humans could audit and vet it. Flooding the internet with AI garbage has become a realistic threat.

And newer models of AI will now use that drivel as input for the next round of AI model learning. And the quality will go down.

With a hint of bad luck, we'll wake up in a world where reality and what is being said about it has nothing to do anymore since most content is AI generated, based on the fever dreams and hallucinations of prior AI generations, with far too little "real" input to be more than a statistical noise element, eliminated by an AI model that considers that insignificant portion of diverging information the error rather than the last vestiges of actual information.

Slashdot Top Deals

Successful and fortunate crime is called virtue. - Seneca

Working...