Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

AI chatbots are not a replacement for search engines

iai.tv

Post reply on HN
Print view View original

371–380 of 415 posts

Re: AI chatbots are not a replacement for search engines

#371

Post by astrange »

solarkraft wrote: post #202
cortesoft wrote: post #142

Earlier quoted context omitted.

How is this different than regular intelligence? We grow up trained on a polluted dataset, being taught things that are subtly or completely false, and we often struggle to tell which things we have learned are false.

Once in a while somebody does some actual on-the-ground research and people are convinced to change their minds through evidence and reasoning. So far AI can't do that, it'll just parrot whatever it hears more often, (apparently) with some degree of randomness. Maybe it'll learn to adopt a new view, but for now it won't be through actual reasoning.

GPT has had up/downvotes for years. They were used to train ChatGPT. It’s not just “what’s the most likely answer”.

(The most likely answer is the letter “e”.)

Re: AI chatbots are not a replacement for search engines

#372

Post by astrange »

k__ wrote: post #359
creatonez wrote: post #356

Earlier quoted context omitted.

> This model cut its teeth writing articles about the discovery of unicorns in the Andes This is a different model, called GPT-3, released in 2020. ChatGPT has the same number of neurons, but is trained with both human feedback and web data, rather than just web data like GPT-3. What you're saying still holds up, though -- ChatGPT still hallucinates information.

Isn't ChatGPT just a conversational interface for GPT-3?

Yes, the human feedback model is called InstructGPT and has been the official “GPT-3” for some time now. ChatGPT is only somewhat more trained, but unlike the playground, it’s free and has a Markdown renderer.

Re: AI chatbots are not a replacement for search engines

#373

Post by arcturus17 »

echelon wrote: post #104
lolinder wrote: post #86

Earlier quoted context omitted.

The problem with ChatGPT as a replacement for Google is that it was not designed to produce accurate facts, and it shows. This model cut its teeth writing articles about the discovery of unicorns in the Andes[0] for goodness sake! It's a language model, and a very impressive one at that, but language is used to express falsehoods and fiction just as regularly as it is used to express truth. This doesn't mean that it…

These are early days. Everything is going to move faster than you can keep up. You'll see head-turning headlines on a daily basis. Next year will see AI begin to disrupt so many businesses you'd thought were absolutely impenetrable. I don't think anything is safe anymore. By the end of the decade, I think even manufacturing and robotics will be automated. Everyone from McDonalds to Apple and Ford will feel the pressu…

> Next year will see AI begin to disrupt so many businesses you'd thought were absolutely impenetrable

Why, because of LLMs in particular? What is the ground-breaking research that will be fueling an AI revolution?

> There are so many people working on AI-first products now. Incumbents will be overwhelmed and washed away.

Incumbents like Google have been doing AI for ages. Microsoft has a very important investment in OpenAI

> By the end of the decade, I think even manufacturing and robotics will be automated. Everyone from McDonalds to Apple and Ford will feel the pressure.

McDonald's, Apple or Ford sell products, not manufacturing services. They will be delighted if manufacturing is automated.

Re: AI chatbots are not a replacement for search engines

#374

Post by astrange »

jchmrt wrote: post #304
the_black_hand wrote: post #266

Earlier quoted context omitted.

RIP google. There's a joke that the best place to hide a murder is on page 2 of google search. The fact is GPT has at bare minimum parsed the top page of google and reasoned about the top results to give the best answer with as much context of why it reasoned so. GPT is basically a superset of google with added reasoning. I can't see it giving worse results than Google. Both will produce a bogus answer if you ask a b…

I think it's really problematic to use the word "reason" to describe what GPT does. As others have commented in this thread, GPT is a language model , which in no way reasons about anything it writes. It is simply writing whatever fits statistically to the prompt and context. Attributing reasoning or too much intelligence to this could prove quite dangerous.

> As others have commented in this thread, GPT is a language model, which in no way reasons about anything it writes. It is simply writing whatever fits statistically to the prompt and context.

Those people are wrong; statistics are a kind of reasoning (intuition) and if you try it, it obviously does the other kinds of reasoning too. I have all kinds of answers from it that are logically coherent and can’t possibly have appeared on the internet already.

(Try asking it to write new SCP wiki entries.)

Re: AI chatbots are not a replacement for search engines

#376

Post by astrange »

tsimionescu wrote: post #321
the_black_hand wrote: post #266

Earlier quoted context omitted.

RIP google. There's a joke that the best place to hide a murder is on page 2 of google search. The fact is GPT has at bare minimum parsed the top page of google and reasoned about the top results to give the best answer with as much context of why it reasoned so. GPT is basically a superset of google with added reasoning. I can't see it giving worse results than Google. Both will produce a bogus answer if you ask a b…

> but GPT will never produce a bogus answer when google has the correct one. What are you basing this assumption on? GPT can always generate a bogus answer, because its only notion of whether an answer is good or not is whether your prompt + that answer are statistically similar to text in its training corpus. The more text from the training corpus it resembles the better GPT will think the answer is. For more obscur…

> GPT can always generate a bogus answer, because its only notion of whether an answer is good or not is whether your prompt + that answer are statistically similar to text in its training corpus.

“Statistically similar” is doing a ton of work even for the original GPT model - just printing the letter “e” is “statistically similar” to English text, if it can even be grammatically correct you should be surprised.

And it’s certainly not true for the last two generations (InstructGPT and ChatGPT) which weren’t trained with that objective at all but used human raters.

Re: AI chatbots are not a replacement for search engines

#377

Post by astrange »

kosh2 wrote: post #341
echelon wrote: post #104

Earlier quoted context omitted.

These are early days. Everything is going to move faster than you can keep up. You'll see head-turning headlines on a daily basis. Next year will see AI begin to disrupt so many businesses you'd thought were absolutely impenetrable. I don't think anything is safe anymore. By the end of the decade, I think even manufacturing and robotics will be automated. Everyone from McDonalds to Apple and Ford will feel the pressu…

Well, there's one thing that humans are much better than robots for the forseeable future: Energy efficiency. Try building a robot that contains a general purpose AI and the motoric functions and sense as humans that runs on the energy of three meals. Now of course humans want and need more than just 3 meals but still pretty good. But if most of your work consist of interfacing with computers, you might be in trouble…

Also, since humans aren’t slaves, they do their own maintenance. And they don’t do exactly what you ask, which is good (when they do it’s called “working to rule” and is a kind of strike).

It’s very rare for automation to cause unemployment to the point it possibly never happened.

Re: AI chatbots are not a replacement for search engines

#378

Post by mrkramer »

bandrami wrote: post #255
lolinder wrote: post #240

Earlier quoted context omitted.

For the same reason why it might produce a bogus answer: it's a language model, not a general AI. Prompt it for footnotes and it will generate footnotes for books that don't exist or that don't actually contain the citation. They'll match the form of a footnote, because that's what the model was trained to do, but the citations won't resolve. For example, I asked who the first president of Argentina was. It gave the…

OTOH I could imagine a hybrid situation where the AI Bot searches Google for you and restates the millions of results Google gives it

Google already does this, it is called Knowledge Graph[0].

[0] https://support.google.com/knowledgepanel/answer/9787176?hl=...

Re: AI chatbots are not a replacement for search engines

#379

Post by epolanski »

lolinder wrote: post #86
gukov wrote: post #22

The way I see it: - Today's search engines will give you links to millions of documents - ChatGPT, if asked correctly, will instead generate one perfect document based on millions of the documents To me, that's a clear evolution of the search engine, especially with all the SEO & ad spam that's plaguing Google and others currently. I wasn't ready to pay a monthly fee for an ad-free Google. I am ready to pay for somet…

The problem with ChatGPT as a replacement for Google is that it was not designed to produce accurate facts, and it shows. This model cut its teeth writing articles about the discovery of unicorns in the Andes[0] for goodness sake! It's a language model, and a very impressive one at that, but language is used to express falsehoods and fiction just as regularly as it is used to express truth. This doesn't mean that it…

Chatgpt is is beta, we can safely assume the situation may change from now to 2024/2025.

Re: AI chatbots are not a replacement for search engines

#380

Post by Retric »

epolanski wrote: post #379
lolinder wrote: post #86

Earlier quoted context omitted.

The problem with ChatGPT as a replacement for Google is that it was not designed to produce accurate facts, and it shows. This model cut its teeth writing articles about the discovery of unicorns in the Andes[0] for goodness sake! It's a language model, and a very impressive one at that, but language is used to express falsehoods and fiction just as regularly as it is used to express truth. This doesn't mean that it…

Chatgpt is is beta, we can safely assume the situation may change from now to 2024/2025.

Not at a fundamental level. Even if you exclusively trained one on some reasonably authoritative source like Wikipedia it would still happily tell you complete nonsense.

These models have no way to differentiate that elves are fictional while Indians actually exist.

Post reply on HN