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

ChatGPT Search

openai.com

Post reply on HN
Print view View original

751–760 of 1001 posts

Re: ChatGPT Search

#751

Post by KoftaBob »

hn_throwaway_99 wrote: post #525
freedomben wrote: post #436

Earlier quoted context omitted.

I genuinely think Kagi has led the way on this one. Simplicity is beautiful and effective, and Kagi has (IMHO) absolutely nailed it with their AI approach. It's one of those things that in hindsight seems obvious, which is a pretty good measure of how good an idea is IMHO. Google could have done it and kind of tried, although they're AI sucks too much. I'm very surprised that OpenAI hasn't done this sooner as well. T…

> I'm also cautiously optimistic though. We'll get there, but it's gonna be a bit shakey for a minute or two. But I don't understand how all of these AI results (note I haven't used Kagi so I don't know if it's different) don't fundamentally and irretrievably break the economics of the web. The "old deal" if you will is that many publishers would put stuff out on the web for free, but then with the hope that they cou…

That's why you're seeing media companies making deals with companies like OpenAI to allow them to access their content for AI learning/parsing purposes, in exchange for the media company getting paid yearly royalties.

Since anyone creating content (whether that's a big media corp or a small cooking blog) holds copyright over their content, they get to withhold the permission to scrape their content unless these AI platforms make a deal with them.

Re: ChatGPT Search

#752

Post by visarga »

hn_throwaway_99 wrote: post #525
freedomben wrote: post #436

Earlier quoted context omitted.

I genuinely think Kagi has led the way on this one. Simplicity is beautiful and effective, and Kagi has (IMHO) absolutely nailed it with their AI approach. It's one of those things that in hindsight seems obvious, which is a pretty good measure of how good an idea is IMHO. Google could have done it and kind of tried, although they're AI sucks too much. I'm very surprised that OpenAI hasn't done this sooner as well. T…

> I'm also cautiously optimistic though. We'll get there, but it's gonna be a bit shakey for a minute or two. But I don't understand how all of these AI results (note I haven't used Kagi so I don't know if it's different) don't fundamentally and irretrievably break the economics of the web. The "old deal" if you will is that many publishers would put stuff out on the web for free, but then with the hope that they cou…

> fundamentally and irretrievably break the economics of the web

Good riddance, it is a surefire way to get slop by having misaligned incentives for publication.

Re: ChatGPT Search

#753

Post by blixt »

skydhash wrote: post #227
blixt wrote: post #108

One thing that is quite unfortunate with the state of SEO and the web in general today is that when I asked "what are the latest versions of common programming languages and when were they released?" a large amount of the sources were "13 Tools You Should Learn Now" and the like. This might be a solvable problem within the search API they provide to the LLM, but for now I wouldn't trust current LLMs to be able to fil…

> when I asked "what are the latest versions of common programming languages and when were they released?" The issue is with the query itself. You're assuming that there's some oracle that will understand your question and surface the relevant information for you. Most likely, it will use the word themselves as part of the query, which SEO sites will exploit. A more pragmatic search workflow would be to just search f…

It would be very odd if I have to speak to an LLM the same way I “speak” to Google. The point of this whole feature seems to be that the LLM chooses when and how to search to best respond to the user.

Re: ChatGPT Search

#754

Post by blixt »

notatoad wrote: post #210
blixt wrote: post #108

One thing that is quite unfortunate with the state of SEO and the web in general today is that when I asked "what are the latest versions of common programming languages and when were they released?" a large amount of the sources were "13 Tools You Should Learn Now" and the like. This might be a solvable problem within the search API they provide to the LLM, but for now I wouldn't trust current LLMs to be able to fil…

>what are the latest versions of common programming languages and when were they released? is this a real question you needed an answer to, or a hypothetical you posed to test the quality of search results? of course you're going to get listicles for a query like that, because it sounds like a query specifically chosen to find low-quality listicles.

It is a proxy to questions that pop up as part of other sessions in my daily use. The LLM chooses its query itself and as long as it is not fine tuned to avoid the “listicles”, or its underlying search engine are not putting importance on more factual responses, I don’t think the answer quality will be as high. It would be weird and redundant if I had to talk to the LLM as if it’s an old school search engine, wouldn’t it?

Re: ChatGPT Search

#755

Post by beambot »

mulmen wrote: post #486
righthand wrote: post #482

Earlier quoted context omitted.

> those visions, users simply asked questions and received reliable answers - nobody had to fact-check the answers ever. It’s a fallacy then. If my mentor tells me something I fact check it. Why would a world exist where you don’t have to fact check? The vision doesn’t have fact checking because the product org never envisioned that outlier. A world where you don’t have to check facts, is dystopian. It means the end…

If you can’t trust the result of a query how can you trust the check on that query which is itself a query? If no information is trustworthy how do you make progress?

Many of the "facts" you "know" are really just useful anecdotes, opinions, approximations or mental shortcuts rather than objective facts. These can still be wildly useful, even if they aren't "true" in the pedantic sense...

As a basic example: Newton's laws aren't a fact. They're extremely useful approximations; to get a more-complete picture you need relativistic effects, quantum effects, etc.

As a rule of thumb: Fact check in proportion to the cost of a mistake.

Re: ChatGPT Search

#756

Post by shinycode »

freediver wrote: post #241

Been thinking about this a lot [1]. Will this fundamentally change how people find and access information? How do you create an experience so compelling that it replaces the current paradigm? The future promised in Star Trek and even Apple's Knowledge Navigator [2] from 1987 still feels distant. In those visions, users simply asked questions and received reliable answers - nobody had to fact-check the answers ever. C…

Don’t you see it coming ? Contrary to google search the engine knows you and will put personalized ads on synthesized answers. Even generate a small video ad of a product or service, tailored to you on the spot. That, is the future. This is the endgame. It’s what made Google rich, why not use the same formula on steroids.

Re: ChatGPT Search

#757

Post by HeatrayEnjoyer »

_bin_ wrote: post #693
zbentley wrote: post #623

Earlier quoted context omitted.

I buy that there's bias here, but I'm not sure how much of it is activist bias. To take your example, if a typical user searches for "is ___ a Nazi", seeing Stormfront links above the fold in the results/summary is going to likely bother them more than seeing Mother Jones links. If bothered by perceived promotion of Stormfront, they'll judge the search product and engage less or take their clicks elsewhere, so it beh…

I can accept some level of this, but if a user specifically requests it, a model should generally act as expected. I think certain things are fine to require a specific ask before surfacing or doing, but the model shouldn't tell you "I can't assist with that" because it was intentionally trained to refuse a biased subset of possible instructions.

How do you assure AI alignment without refusals? Inherently impossible isn't it?

If an employee was told to spray paint someone's house or send a violently threatening email, they're going to have reservations about it.. We should expect the same for non-human intelligences too.

Re: ChatGPT Search

#758

Post by freedomben »

ahmedbaracat wrote: post #496
freedomben wrote: post #436

Earlier quoted context omitted.

I genuinely think Kagi has led the way on this one. Simplicity is beautiful and effective, and Kagi has (IMHO) absolutely nailed it with their AI approach. It's one of those things that in hindsight seems obvious, which is a pretty good measure of how good an idea is IMHO. Google could have done it and kind of tried, although they're AI sucks too much. I'm very surprised that OpenAI hasn't done this sooner as well. T…

Are you referring to Kagi Assistant? https://help.kagi.com/kagi/ai/assistant.html

I am referring to all of their AI stuff, kagi assistant included. Personally, the best feature is the quick answer. It essentially scans the top several hits and uses an llm, reads them to see if it answers your question, and will display a summary that also includes links to the full source. I find that feature to be wonderful. I will usually look through the quick answer and see if the site actually answers the question I have, and then we'll click through. If everyone implemented it like this, it's possible it could save the current model.

Re: ChatGPT Search

#759

Post by popcorncowboy »

rawgabbit wrote: post #415
whizzter wrote: post #300

This is probably Google's Altavista moment, by making their results crappier by the year in search of Ad dollars everyone has felt that there is a potential for search to be better and once that becomes available they'll be in a continious game of catch-up. Yes, Google has their own AI divisions, tons of money and SEO is to blame for part of their crappiness. But they've also _explicitly_ focused on ad-dollars over a…

For me the key difference between ChatGPT Search and Google is the feedback mechanism. With ChatGPT, I can give a thumbs up or thumbs down; this means that OpenAI will optimize for users thumbs up. With Google, the feedback is if I click on an Ad; this means that Google optimizes for clickbait.

No way that gets gamed

Re: ChatGPT Search

#760

Post by mr_toad »

umvi wrote: post #451
whizzter wrote: post #300

This is probably Google's Altavista moment, by making their results crappier by the year in search of Ad dollars everyone has felt that there is a potential for search to be better and once that becomes available they'll be in a continious game of catch-up. Yes, Google has their own AI divisions, tons of money and SEO is to blame for part of their crappiness. But they've also _explicitly_ focused on ad-dollars over a…

At some point it seems like Google switch to ML-based search instead of index based search. You can search for very specific combinations of lyrics and scenes: "eyes on me pineapple bucket of water house of cards chess time loop" and you won't surface a link to the music video featuring all of those things ( https://www.youtube.com/watch?v=AlzgDVLtU6g ), you'll just get really generic results of the average of your q…

> Google switch to ML-based search

The very earliest form of page rank used a form of ML.

Post reply on HN