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

391–400 of 415 posts

Re: AI chatbots are not a replacement for search engines

#391

Post by turmeric_root »

TOMDM wrote: post #97
oefnak wrote: post #94

Earlier quoted context omitted.

I wanted to convert a string to be used in an URL. I gave him the source string, the target string (spaces replaced by %20). It told me it was called URL encoding and gave me the code to do it.

Another great example. I'm sure most people who have messed with APIs long enough could have told you that, but it's an annoying thing to learn the first time if you don't have someone to talk to. To me chatgpt is that over confident friend you can talk a problem through with, except they've used every language and library under the sun. It's like speaking to someone who is an amateur at _everything_ but a little too…

While I've had fun using it to meme, I don't disagree with people calling referring to chatgpt as 'mansplaining as a service'. It feels like the model straight up ignores part of my prompts sometimes.

Re: AI chatbots are not a replacement for search engines

#392

Post by dinkumthinkum »

spiderice wrote: post #223
dinkumthinkum wrote: post #107

Earlier quoted context omitted.

It gives literally the same argument for 42 being prime as 23.

I'm not sure what you're talking about. I just asked it to argue that 42 is a prime number, and it clearly knew it wasn't and even said multiple times that the agument is just pretend/hypothetical: ________ Me: Make an argument for why 42 is a prime number ChatGPT: 42 is not a prime number. It is a composite number, meaning it is the product of two smaller numbers. Me: But pretend it is a prime number and make that a…

This has been documented many times, you can Google it or just ask ChatGPT again.

Re: AI chatbots are not a replacement for search engines

#393

Post by machiaweliczny »

NikolaNovak wrote: post #195
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…

Today I have zero confidence in chatGPT, OR my ability to evaluate and correct it's inaccuracies strictly from within its interface. I feel I can compensate for search engine and arrive at answer I seek within my confidence interval via search engine. It's a tool that can get me there. That may change in the future. Based on my limited understanding on how chatgpt works,I don't see how though - it explicitly is not d…

For me it’s better than search engine. If promted correctly it can summarise and synthetise hours of work.

Re: AI chatbots are not a replacement for search engines

#394

Post by sibflporb »

ChatGPT works very well as a tech demo but nothing like it will ever work at scale.

There simply is no business model for something as useful as ChatGPT, and there never will be. Let's assume for a minute that OpenAI will manage to improve their models such that the accuracy

There will probably be a business case for the chatbot Google will turn its search engine into, but it will be as far removed from ChatGPT as today's Google is from the good old Google of PageRank fame from 1999, which would show you exactly what you asked for, often enough to be considered genuinely useful.

The Google search engine from 1999 was a bit like ChatGPT is today: The demonstration of a new way of doing things that seemed, at the time, to be the way of the future. Except we don't. PageRank is an awesome example of how computer science nerds can disrupt a market with their superior math but it didn't change the basic laws of capitalism and this is why it succeeded in turning Google into one of the letters of FAANG, but it failed at being sustainable. A Google-killing descendent of ChatGPT would have a hard time turning a profit without also turning into the same sort of thing Google is today: A convenient way of giving people ads instead of answers.

Here is my prediction:

1) ChatGPT and similar "AI" chatbots will never be more than tech demos

2) The Google of the future will have some shortish chatbot-copypasta for search results that resemble prompts to write essays

3) Google will sell ads for highly specialized and expensive services for the few people who need to copy-paste longer essays (which might or might not have a sufficient relation to reality) and are able to pay for it

4) The end-result will be virtually indistinguishable from the Google of today, because for every honest chatbot-aaS, there will be an uncountable amount of dishonest vendors of "free" auto-generated propaganda and spam, copy-pasted by school-children all over the world, because that will be the results Google will show below their own copypasta, which is just too short for this week's essay because the teacher knows about Google, too. And on HN, people will long for the good old times of 2022, when ChatGPT was, briefly, surprisingly awesome.

5) Your crush won't believe you that you wrote that touching love poem all on your own.

Re: AI chatbots are not a replacement for search engines

#395

Post by layer8 »

noitpmeder wrote: post #361
layer8 wrote: post #90

Earlier quoted context omitted.

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'"…

Yes, that‘s the conclusion I drew after a number of interactions. “It tried to convince me” was shorthand for that it persistently argued for that. I don’t attribute any actual agency to ChatGPT, of course.

The lack of self-reflection, of logical reasoning ability and associated memory is what makes ChatGPT useless for the use cases I’d actually find interesting.

Re: AI chatbots are not a replacement for search engines

#396

Post by rtkwe »

grugagag wrote: post #118
rtkwe wrote: post #70

Earlier quoted context omitted.

We used to have just Internet grep but it was awful because sites just slammed tons of keywords into invisible text to get fake traffic. That's what spawned the original innovation at Google of PageRank and thus the SEO wars started. Going back to that wouldn't solve the issue with modern search.

If we only had a downvote or don’t show me results from this source this problem would be solved. Google didn’t want to give us that feature because of … different incentives, no other reason comes to mind. I’ve witnessed google search become worse and worse and now it’s barely useful

That would cut down on some of the concentrated SEO spam centers but I think you'd still be fighting an endless battle with more distributed SEO spam sites that are more focused on one point.

Re: AI chatbots are not a replacement for search engines

#397

Post by doliveira »

brookst wrote: post #387
doliveira wrote: post #136

Earlier quoted context omitted.

AI is orders of magnitude cheaper and faster, not to mention amoral.

Disinformation is moral when created by people, because it costs more? I’m new to this idea that something can be acceptable when done by people but unacceptable when automated. I don’t really see it.

Amoral, not immoral. I mean the AI won't make any moral judgement itself, and using it for whatever task doesn't carry additional moral implications other than those of the task itself, opposite to humans in which you have to care at least a little bit about what you're telling them to do.

About the task itself, it may be immoral by nature, and now it can be executed orders of magnitude faster and without the moral concerns concerting the workers themselves.

Re: AI chatbots are not a replacement for search engines

#398

Post by mrkramer »

schmorptron wrote: post #365
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.

Google finally stared to pay licencing fees to the news publishers for using their news articles for creation of headlines and summaries for Google search results and for the Google News[0]. If AI chatbots want to thrive they need to employ this business model in which they would pay licensing fees to the content creators in order to use their content to feed the AI algorithm. And ofc they(AI chatbots) would need to link(quote) content sources in hope and in order for users to follow the links and drive traffic to the original content creators. That's how Wikipedia operates; volunteers are crowdsourcing Wikipedia pages with information gathered from all around the Web(plus books and research papers) and they quote information sources. But in the AI chatbot case the selection and the analysis of information is done by the AI unlike Wikipedia which uses human volunteers to do the job. AI chatbot and its content providers would be incentivized by profit while Wikipedia and its volunteers are motivated by altruism. Which brings me to the conclusion that AI chatbot would actually provide you with better information than Wikipedia(taking in consideration for-profit competition between content creators) or any other non-profit but Wikipedia is still useful and relevant with the regard to quality.

[0] https://blog.google/around-the-globe/google-europe/google-li...

Re: AI chatbots are not a replacement for search engines

#399

Post by 1vuio0pswjnm7 »

This is a highly editorialised title. The title chosen by the authors is "All-Knowing Machines Are A Fantasy". There is nothing in the blog post about "AI chatbots" versus search egnines.

Due the editorialised HN title the top comment takes the thread on a tangent about ChatGPT versus Google web search.1 There is nothing in the article about Google specifically, other than to mention Google has their own language model for dialogue ("AI chatbot") along with OpenAI and Facebook. Nor is there any discussion of search engines, e.g., Google's, except to point out at the end that "AI" and search engines are similar because neither is under the control of the user. They are each offered by corporations that, regardless of the input, shape the output to benefit the corporation's own commercial interests. Neither offers "neutral" results.

A true improvement in search technology would be one that provides neutral results. Providing that improvement means the "tech" company cannot survive. It has no viable business model except operating as an intermediary (middleman), selling ad services.

1. Incidentally no user ever verifies that either Google or ChatGPT searches through millions of anything. The corporation will tweak the results to make sure it stays in business and maximises profit. The user is never allowed to see the tweaks.

Re: AI chatbots are not a replacement for search engines

#400

Post by Tepix »

matusp wrote: post #330
Tepix wrote: post #319

ChatGPT has its uses, but it can be dangerous: The other day it told me that the legal limit for Plutonium in drinking water is 0.1 micrograms per liter.

And what is the true legal limit?

Inhaling 40 nanograms is enough to reach the yearly radiation limit.
Post reply on HN