Journal shanen's Journal: Should we have a worst-ever contest? 6
Many mornings I have one weird thought of the day, but it seems to be three this time:
1. The generative AI paradox: If you give a generative AI the slightest hint what kind of answer you want then it'll mentally jump backwards through hoops to give that answer to you. But if your question is too vague so as to avoid giving any hints, then the genAI will misunderstand the question and thereby mangle the answer. (But even if your question gives nothing away, if you are in some way "logged in" and recognizable, then the AI will probably use any background information it can find about you to "pander to the user", as I originally described the problem to some CS researchers I was working with about 20 years ago.)
2. What are the most difficult skills for human beings? Perhaps developing proper skepticism and real impartiality towards what they want to believe. But the generative AIs are experts in telling you exactly what you want to believe. And as plausibly as possible. Of course they hallucinate and generate LOTS of AI slop.
3. I know records are made to be broken, but I don't remember praying to live into an era of most-ever still-alive worst-ever super-greedy villains setting continuous records for being even worse than they were the day before... With morality-free support from the biggest and most powerful and, I dare say, worst-ever AIs ever created. Most people are nice enough, but they are NOT making the crucial decisions now.
In particular, I don't see how the American Constitution can survive this mess. And in general the near-term future of humanity looks rather bleak to me. Linking back to my thoughts on the Fermi Paradox resolving in negative ways...
But maybe I'm overreacting? Just read an interesting piece from Umberto Eco, published a few years before he passed away. The title as translated into English is "On the Disadvantages and Advantages of Death". Funny joke about human idiocy before coming down on the pro side.
Overestimating the effect of AI (Score:2)
AI sucks ass, not just morally but effectively.
Employers have tried to replace workers with AI, but found it does not work.
AI, like most computers, appears to be good at a bunch of things that humans suck at (spelling, grammar, math, finding references to back their believes, use of the em-dash, ability to draw naked people, ability to draw things besides naked people,....) . But Ai also appears to be really bad that the things that count. Not lying/hallucinating, noticing incongruities and investigating/
Re: (Score:1)
Partial concurrence, but at least two major points of disagreement. One is the temptation aspect. People want oracles and that's how AI feels. The other is that the people who are making the crucial decisions are dangerous idiots who refuse to see the coming crash but just keep stomping harder on the accelerator. I actually think that most of them are clever enough to see that there will be a crash but they are also gamblers who are stupid enough to think that they will survive and become the biggest winner
Re: (Score:2)
Really? More moderation abuse for the annual transition? Hope that it's part of the end of a bad year rather than the start of another bad one? Since the abusive moderation was "Redundant", I guess that justifies replicating the reaction? So whatever happened to ol' Scott Adams?
Re: (Score:2)
Not lying/hallucinating, noticing incongruities and investigating/fixing them, morality/ethics, truly original creative ideas, coming up with new solutions, etc.
Unfortunately right now there are a lot of people making careers out of doing exactly the opposite of those things. While one could hope those people would be replaced by AI - or even better be replaced by people who follow those principles - that does not seem to be priority right now.
Re: (Score:2)
My version would be that most people are okay, even nice, but they are not the people making the big decisions these years.
I also see "the human condition" as a kind of software compatibility problem. We have various programs running at various levels and as the levels go higher things get more and more out of control. This time I'll use eating as an example. We have a low level program about stuffing things in our mouth to see if they taste good and spitting them out if they don't. Later on we developed pr