ChatGPT has meltdown and starts sending alarming messages to users::AI system has started speaking nonsense, talking Spanglish without prompting, and worrying users by suggesting it is in the room with them
ChatGPT has meltdown and starts sending alarming messages to users::AI system has started speaking nonsense, talking Spanglish without prompting, and worrying users by suggesting it is in the room with them
Its being trained on us. Of course its acting unexpectedly. The problem with building a mirror is proding the guy on the other end doesnt work out.
To be honest this is the kind of outcome I expected.
Garbage in, garbage out. Making the system more complex doesn’t solve that problem.
Bamalam
The development of LLMs is possibly becoming self defeating, because the training data is being filled not just with human garbage, but also AI garbage from previous, cruder LLMs.
We may well end up with a machine learning equivalent of Kessler syndrome, with our pool of available knowledge eventually becoming too full of junk to progress.
God I hope all those CEOs and greedy fuckheads that fired hundreds of thousands of people wayyyyy too soon to replace them with this get their pants shredded by the fallout.
Naturally they’ll get their golden parachutes and land on their feet even richer than before, but it’s nice to dream lol
I mean, surely the solution to that would be to use curated/vetted training data? Or at the very least, data from before LLMs became commonplace?
Yes but that only works if we can differentiate that data on a pretty big scale. The only way I can see it working at scale is by having meta data to declare if something is AI generated or not. But then we’re relying on self reporting so a lot of people have to get on board with it and bad actors can poison the data anyway. Another way could be to hire humans to chatter about specific things you want to train it on which could guarantee better data but be quite expensive. Only training on data from before LLMs will turn it into an old people pretty quickly and it will be noticable when it doesn’t know pop culture or modern slang.
Pretty sure this is why they keep training it on books, movies, etc. - it’s already intended to make sense, so it doesn’t need curated.
The funny thing is, children are similar. They just learn whatever you put in front of them. We have whole systems for educating children for decades of their lives.
With AI we literally just plopped them in front of the Internet, with no guidelines on what to learn. AI researchers say “it’s a black box! We don’t know why it’s doing this!” You fed it everything you could and gave it few rules on what to do. You are the reason why it’s nuts.
Humans come hardwired to be a certain way, do certain things. Maybe they need to start AI off like that, some basic programs that guide learning. “Learn everything” isn’t working.
That’s a good point. For real brains, size and intelligence are not linked. An elephant brain has 3 times the amount of neurons as a human brain, but a human brain is more intelligent. There is more to intelligence than just the amount of neutrons, real or virtual, so making larger and larger AI models may not be the right direction.
True. Maybe they just need more error correction. Like spend more energy questioning whether what you say is true. Right now LLMs seems to just vomit out whatever they thought up, with no consideration of whether it makes sense.
They’re like an annoying friend who just can’t shut up.
Damn.
Thank you VERY much for that insight: AI’s version of Kessler-syndrome.
EXACTLY.
Damn, damn, damn, that gets the truth right in its marrow.
_ /\ _
I am happy to report I did my part on feeding it garbage. I only ever speak to chatGPT thru a pirate translator. And I only ever ask it for harry potter fan fic. Pay me if you want me to train it meaningfully.
The solution is paying intelligent people to interact with it and give honest feedback.
Like, I’m sure you can pay grad students $15/hr to talk to one about their subject matter.
But with as many as they’d need, it would get expensive.
So they train with low quality social media comments, or using copywritten text without paying the owners.
It’s not that we can’t do it, it’s just expensive. So a capitalist society wont.
If we had an FDR style president, this would be a great area for a new jobs program.
It appears, that with the increase in popularity of machine learning, the percentage of people who properly source and sanitize their training data has steeply decreased.
As you stated, a MLAI can only be as good as the data it was trained on, and is usually way worse. The popularity and application of MLAIs built with questionable practices scare me, though, at least their fuckups will keep me employed and likely more busy than ever.
LLM’s are not “machine learning”, they are neural-networks.
Different category.
ML is small potatoes, ttbomk.
Decision-tree stuff.
Neural-nets are black-boxes, with back-propagation training of the neural-net to get closer to ( layer by layer, training-instance by training-instance ) the intended result.
ML is what one does on one’s own machine with some python libraries,
ChatGPT ( 3, 3.5, or 4, don’t know which ) cost something like $100,000,000 to rent the machines required for mixing the training-data & the model ( I’m assuming about $20/hr per machine, so an OCEAN of machines, to do it )
_ /\ _