Hacker News Reader

Comment by azinman2 | Hacker News Reader

It’s interesting you say incredible when I’m coming away with the opposite. It is not only overly polite, but constantly wrong yet so confident in its answers. And even when you iterate it still doesn’t get there, yet each time saying “oh you’re right” as if it finally got it.

View on HN

Being right and sounding right are two important but separate skills in natural language communication. It sounds like they've nailed one of them and the other is somewhat hit and miss right now. To me it's obvious that there's no reasoning going on behind the curtain, but in a lot of domains that doesn't matter. You don't need your answering machine to opine on philosophy questions just like you don't need your mechanic to invent and machine a custom part every time you bring your car in.

(Edit: Point being if you can back the chatty part with something that's actually an expert in the domain you need it to talk about, then you've got cold fusion.)

I don't think it's ready for prime time yet, but the way voice recognition and synthesis, ML, and scalable computing power have been going in the past several years, we're not far off. And that's incredible.

As you point out what’s missing is the philosophy of mind bit. It’s all just probabilistic NL models that will output _something_, whether it has any concept of rationale or not. What is absolutely astonishing is that ChatGPT can accomplish what it does with just these markov-chain-like methods. I never would have predicted it.

> It is not only overly polite

this is a deliberate goal of OpenAI's training, not an emergent feature of the underlying technology.

> constantly wrong

how close to right do you think it gets with the "Three Pints of Guinness" explanation?

i would have given full marks for and additional comment like "it's considered funny because the man himself claims to have quit drinking but he actually continues to do so via the tradition"

nevertheless, it does a remarkable job of parsing the structure of the joke and conveying this parsing via natural language.

to me, my "additional comment" looks like just a further iteration on presenting this parsing. i would say it gets very close to "completely right".

Interesting. My first thought upon reading that one was that "he says he quit drinking but is obviously currently drinking" is the first thing I'd explain about the punchline (then moving on to explain how that connects to the tradition).

I think we're all familiar with people who say something completely wrong, with full and absolute confidence that they are right :-)

Consumers lost $5.2 billion to phishing last year. This technology may not fool you, but there are enough folks who will be fooled by the basic conversational skills it has. Scammers can be 10x "productive" with this tech.

And I am completely surprised by people who think this tech today is not good, and fail to account that it can get way better in the future.

I’m guessing by people you mean me? I’m only talking about the state of ChatGPT as per the examples given. I’m not talking about the wider implications into the future or its other amazing capabilities.

Reply on news.ycombinator.com