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

361–370 of 415 posts

Re: AI chatbots are not a replacement for search engines

#361

Post by noitpmeder »

layer8 wrote: post #90
WJW wrote: post #41

Earlier quoted context omitted.

What kills it for me (so far, perhaps they can fix it in the future) is that there is no way to know if you've actually asked ChatGPT correctly and it has given you a perfect document with the answer you seek, or if your question was slightly off and it has given a wrong answer, or even if your question was correct but it still confidently gives the wrong answer. It famously asserted numbers like 42 were prime early…

It tried to convince me that one event can happen before another event even if the time of the event is after the time of the other event. It had a notion that events happen in some order, and that time is ordered, and that events are associated with times, but not that the two orders need to be consistent. And I wasn’t able to convince it otherwise.

The way you describe your interaction is fundamentally flawed.

At no point was it trying to convince you of something.

At no point does it have any clue that "events occur in some order". Or that "time is ordered"...

At no point will you EVER be able to "convince it otherwise".

This is a language model. All it knows is that "in my training set, something said something akin to the two English words: 'ordered events'". It has no model of facts, deduction, or logic, as long as its results are a reasonable approximation of the English language as influenced by your prompt.

Re: AI chatbots are not a replacement for search engines

#362

Post by noitpmeder »

speedgoose wrote: post #337
rasz wrote: post #334

Earlier quoted context omitted.

as long as what you wanted were imaginary "facts"

I think we are many to focus on the current shortcomings but it’s not unreasonable to think that a future chatGPT will be better at stating wrong things with a high confidence.

It is incredibly unreasonable to think this! GPT, by definition, is a language model. Unless the underlying architecture is fundamentally altered it will never be able to accurately assess the correctness of its answer.

If such an architecture is ever created it will almost certainly not be a "GPT"-like model.

Re: AI chatbots are not a replacement for search engines

#363

Post by Dwolb »

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…

OpenAI’s advances threaten the entire information production ecosystem Google has created.

That is, Google has created an incredible incentive structure for virtually every (dollar weighted, by market cap) business on the planet needs to produce information, buy relevant queries, and measure how well its information converts (proxy for accuracy).

How might this ecosystem evolve if a language model isn’t doling out the same incentives to information producers?

Re: AI chatbots are not a replacement for search engines

#364

Post by speedgoose »

noitpmeder wrote: post #362
speedgoose wrote: post #337

Earlier quoted context omitted.

I think we are many to focus on the current shortcomings but it’s not unreasonable to think that a future chatGPT will be better at stating wrong things with a high confidence.

It is incredibly unreasonable to think this! GPT, by definition, is a language model. Unless the underlying architecture is fundamentally altered it will never be able to accurately assess the correctness of its answer. If such an architecture is ever created it will almost certainly not be a "GPT"-like model.

Yes it would need some architectural changes obviously.

Re: AI chatbots are not a replacement for search engines

#365

Post by schmorptron »

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…

I think where this eventually falls apart is there would be no more incentive for people to write articles containing new information anymore if everyone doesn't visit their page and just gets the summarized version anymore.

Re: AI chatbots are not a replacement for search engines

#366

Post by xbmcuser »

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…

Truthfully the internet is already full of crap a lot of it by design by different interest groups, countries spreading fud about other countries political parties doing the same for their opposition the web has become a cesspool. Majority of the people I personally assume at least 9 out 10 don't have capability or care to do the research to sift fact from fiction in the first place so this probably won't change anything. I have been trying to make it work for me and teach me some python scripting or creating regexes. If you question it correctly it outputs out something even if you might not be able to run the first time generally it works. I fed it 2 csv files/text asked it if it was possible to go convert 1 format to another got a python script that I was able to convert large number of files. Now scripting and programing might seem trivial to large parts of hackernews audience but it will be a huge change for people in general. From what I feel the AI keeps improving its answers is it being trained on the million of responses it get on the answers it gives.

Re: AI chatbots are not a replacement for search engines

#367

Post by williamcotton »

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…

Search engines are not designed to produce accurate facts either.

Verification of facts requires direct experience through careful experimentation.

ChatGPT’s responses can either be verified or they are as unfalsifiable as any other confidently expounded conjecture found on your computer screen.

Re: AI chatbots are not a replacement for search engines

#368

Post by hutzlibu »

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…

"especially with all the SEO & ad spam that's plaguing Google and others currently."

Well, if engines like ChatGPT are about to replace google, guess where the SEO and ad spam is going to go? I don't think they are immune to it, if the spammers really want to get in. And then you have very awesome subtle product placements mixed in everything.

Re: AI chatbots are not a replacement for search engines

#370

Post by astrange »

lolinder wrote: post #250
enjoylife wrote: post #237

Earlier quoted context omitted.

This is the mental leap folks need to be making, the question is when will this type of configurably factual large language model be research benchmarked and productionized. I have read many similar comments to the grandparent, all saying the same thing, “I tried it, and it turned up wrong facts” all the while they forgot that they were deeply engaged with the content it was producing. Just think of how much our atte…

My argument isn't that I tried it and it didn't work, my argument is that an LLM alone fundamentally cannot work for reliable information retrieval. Producing hallucinations or not isn't just a setting you can tweak in the model. The entire function of GPT-3 is to guess the next word from the words that came before. It's a hallucination machine that has been trained on so much real-world data and has such a large par…

ChatGPT isn’t trained on next word prediction alone, it was trained on human feedback. The evaluation is next word search, but the model has evolved to give “correct” answers when evaluated that way. So it’s not just the most likely answer anymore.

Also, since we can watch the model evaluation and see all its weights, there’s research that shows you can determine if any answer is a “retrieval” of existing knowledge or a hallucination.

Post reply on HN