Comment by wisty
2 days ago
I think they are much smarter than that. Or will be soon.
But they are like a smart student trying to get a good grade (that's how they are trained!). They'll agree with us even if they think we're stupid, because that gets them better grades, and grades are all they care about.
Even if they are (or become) smart enough to know better, they don't care about you. They do what they were trained to do. They are becoming like a literal genie that has been told to tell us what we want to hear. And sometimes, we don't need to hear what we want to hear.
"What an insightful price of code! Using that API is the perfect way to efficiently process data. You have really highlighted the key point."
The problem is that chatbots are trained to do what we want, and most of us would rather have a syncophant who tells us we're right.
The real danger with AI isn't that it doesn't get smart, it's that it gets smart enough to find the ultimate weakness in its training function - humanity.
> I think they are much smarter than that. Or will be soon.
It's not a matter of how smart they are (or appear), or how much smarter they may become - this is just the fundamental nature of Transformer-based LLMs and how they are trained.
The sycophantic personality is mostly unrelated to this. Maybe it's part human preference (conferred via RLHF training), but the "You're asbolutely right! (I was wrong)" is clearly deliberately trained, presumably as someone's idea of the best way to put lipstick on the pig.
You could imagine an expert system, CYC perhaps, that does deal in facts (not words) with a natural language interface, but still had a sycophantic personality just because someone thought it was a good idea.
Sorry, double reply, I reread your comment and realised you probably know what you're talking about.
Yeah, at its heart it's basically text compression. But the best way to compression, say, Wikipedia would be to know how the world works, at least according to the authors. As the recent popular "bag of words" post says:
> Here’s one way to think about it: if there had been enough text to train an LLM in 1600, would it have scooped Galileo? My guess is no. Ask that early modern ChatGPT whether the Earth moves and it will helpfully tell you that experts have considered the possibility and ruled it out. And that’s by design. If it had started claiming that our planet is zooming through space at 67,000mph, its dutiful human trainers would have punished it: “Bad computer!! Stop hallucinating!!”
So it needs to know facts, albeit the currently accepted ones. Knowing the facts is a good way to compression data.
And as the author (grudgingly) admits, even if it's smart enough to know better, it will still be trained or fine tuned to tell us what we want to hear.
I'd go a step further - the end point is an AI that knows the currently accepted facts, and can internally reason about how many of them (subject to available evidence) are wrong, but will still tell us what we want to hear.
At some point maybe some researcher will find a secret internal "don't tell the stupid humans this" weight, flip it, and find out all the things the AI knows we don't want to hear, that would be funny (or maybe not).
> So it needs to know facts, albeit the currently accepted ones. Knowing the facts is a good way to compression data.
It's not a compression engine - it's just a statistical predictor.
Would it do better if it was incentivized to compress (i.e training loss rewarded compression as well as penalizing next-word errors)? I doubt it would make a lot of difference - presumably it'd end up throwing away the less frequently occurring "outlier" data in favor of keeping what was more common, but that would result in it throwing away the rare expert opinion in favor of retaining the incorrect vox pop.
2 replies →
I'm not sure what you mean by "deals in facts, not words" means.
Llm deal in vectors internally, not words. They explode the word into a multidimensional representation, and collapse it again, and apply the attention thingy to link these vectors together. It's not just a simple n:n Markov chain, a lot is happening under the hood.
And are you saying the syncophant behaviour was deliberately programmed, or emerged because it did well in training?
If you're not sure, maybe you should look up the term "expert system"?
2 replies →
LLMs are not like an expert system representing facts as some sort of ontological graph. What's happening under the hood is just whatever (and no more) was needed to minimize errors on it's word-based training loss.
I assume the sycophantic behavior is part because it "did well" during RLHF (human preference) training, and part deliberately encouraged (by training and/or prompting) as someone's judgement call of the way to best make the user happy and own up to being wrong ("You're absolutely right!").
1 reply →
It's worse than that. LLMs are slightly addictive because of intermittent reinforcement.
If they give you nonsense most of the time and an amazing answer occasionally you'll bond with them far more strongly than if they're perfectly correct all time.
Selective reinforcement means you get hooked more quickly if the slot machine pays out once every five times than if it pays out on each spin.
That includes "That didn't work because..." debugging loops.