Item — HN Reader

Hold on, at least for me, 90% of the stuff is not just coherent-looking, but coherent. I do think that when it's wrong, it should give a heads up about not being certain about a subject. It's certainly tweaked to sound almost overconfident about every subject, which gives off a bullshitty vibe when the details of the explanation are wrong.

What sorts of subjects have you been trying it out with?

2 comments

I’ve had pretty complex discussions about Buddhism, physics, and other subjects and it was generally erudite and accurate. In its current unrefined form it’s more useful than google is at providing understanding on most subjects, especially because it will attempt to answer my direct questions rather than providing documents with terms in them. In fact I think it’s probably one of the most useful tools I’ve ever used.

Has it given me wrong information? Absolutely. But it’s always been pretty obviously wrong, and I often use it to introduce me to a subject then follow up with google to verify details. I further fully expect this to improve.

Yes, I've been going back and forth between Google and ChatGPT quite a bit lately. Sort of using Google as the verification step, after getting deeper into a subject.

For the haters, it seems incredibly likely that an IR system like google will be augmenting LLM and semantic reasoning systems to form a solution to the problems folks point. The problems chatgpt suffer from are solved and are complementary.

I think part of the point here is that ChatGPT has no clue what you might mean here by "certainty" or being "wrong," and the fact that people have the impression that it could have an idea of those concepts is indicative of people's poor understanding of what ChatGPT really is

Could you elaborate on this?

It definitely is able to decipher whether it has knowledge about the future, or some specific political events. This is obviously a pretty straightforward bonus layer on top of the model itself, but couldn't there be an extrapolation of that system where it's not binary, but rather a range between 0 and 1? I'd imagine this wouldn't be the model itself doing the crunching of the previous tokens here, at least not the same instance of it, as it could be stuck in whatever character or loop of reasoning it has going on at the moment.

Are you talking about the same people who couldn't agree about a vaccine? How is logical consistency going with humans?

Not to mention that a mere 700 years ago we were dying of bubonic plague and with all our general intelligence could not muster up the germ theory of disease. Not even to save our lives, you see, we are not generally efficient, it depends on century.

We are dependent on experimental results carefully constructed to verify our theories, theories which start like chatGPT's random bullshit initially, random words following a probability distribution in our heads. Even deep learning is often touted as modern alchemy - why don't we just understand?

Verification does wonders to language models. Humans have more verification and interactive experiences, so we think ourselves superior. But an AI could have the same grounding with us. Like AlphaZero who became a super-human Go player without ever looking at human games - learned it from lots of verification. And CICERO the model playing Diplomacy in natural language.

Just set AI up with a verification loop to see wonders. Predicting the next word correctly is just one of the ways AI can learn.