I wonder how this will compare to Arc's "Browse for me" feature–at least for a single search (I don't thing the chat-based interface will be much of a value add here beyond just running a second "Browse for me"/SearchGPT). I really like the "Browse for me" feature, but only in specific cases, and I don't think I'd ever use it if it was a standalone app and not a built-in feature of Arc that also lets me run a quick google search.
OpenAI Announces SearchGPT
121–130 of 203 posts
Re: OpenAI Announces SearchGPT
#122Perhaps no-one could answer my question in [0] because many (including me) knew OpenAI would release their own search engine; eventually in [0]. "Who can tell me how and why is Perplexity.ai is worth $1BN? How much revenue are they making vs the amount of money they are burning? What is the justification of this valuation?" [0] At this point with this unsurprising announcement, Perplexity is worth [0] https://news.yc…
It is worth $1BN because people who invested in it, think it is worth that much.
Re: OpenAI Announces SearchGPT
#123Earlier quoted context omitted.
If that's the case, I think it's fair to say that we can skip websites and just host a service that chatgpt can talk to. If you're a restaurant, user can actually order right from the chat. or voice.
> If you're a restaurant, user can actually order right from the chat. or voice. So, instead of giving 30% fee for DoorDash or Wolt about visibility, we start giving that fee for "some-AI-search-tool", and they don't allow selling food cheaper than you get by using the AI search process ordering. I don't like this era.
That's assuming you don't have any customers. Those will just directly name your restaurant. I guess the only way to attract new customers would be word-of-mouth, or some other form of advertising.
Re: OpenAI Announces SearchGPT
#124Search engines for me is keyword based. I want to type "pizza near " or " " and get sensible result. Anytime a search engines tries to interpret my query with something far semantically and ignoring my own keywords, it's a worse experience.
I have no data to back this up, but I strongly suspect 90% of search traffic consists of dumb searches for things like, "facebook", "weather tomorrow", "gmail".
While people do search for things that could benefit from some comprehension, I don't think that's a common feature.
For example, my most recent searches and I'm probably a bit of an outlier given my usage:
"given when then" "[some project] github" "[some person] wikipedia" "[some person] wikipedia" "act of supremacy wikipedia" "NVDA stock" "django docs onetoonefield"
Perhaps one of those Wikipedia searches could have been done better as an AI search since I wanted to know something specific, but other two were just from wanting to read generally about a topic.
The benefits I get from ChatGPT and similar tools are more conversational than search like. Eg, I might be trying to solve a coding problem and want some suggestions about how I might go about it. I might act for libraries, example code, and pros and cons of different approaches. I basically use it as a replacement for another senior engineer which I can bounce ideas off of, it's not for search / knowledge type stuff and I can't see why I'd ask an AI for that. If I want to know something I can just type a few keywords into google and find a reputable site that for that info.
Re: OpenAI Announces SearchGPT
#125Earlier quoted context omitted.
> I'm glad that most of them seem to respect robots.txt. https://github.com/ai-robots-txt/ai.robots.txt/blob/main/tab... Some of them identify themselves by user agent but don't respect robots.txt, so you have to set up your server to 403 their requests to keep them out. If they start obfuscating their user agents then there won't be an easy solution besides deferring to a platform like CloudFlare which offers to pla…
The entry here for Perplexity is the one that got a lot of attention but it's also unfair: PerplexityBot is their crawler, which uses that user agent and as far as anyone can tell it respects robots.txt. They also have a feature that will, if a user pastes a URL into their chat, go fetch the data and do something with it in response to the user's query. This is the feature that made a big kerfuffle on HN a while back…
No, thats illogical.
The action is indeed prompted by a human, but so is any crawl in some way. At some point they either configured an interval or other trigger to send the script to the Web host to fetch anything it can find.
It's inherently different to extensions such as adblockers that just remove elements according to configuration.
After all, the users device will never even see the final DOM now. instead it's getting fetched, parsed and processed on a third device, which is objectively a robot. You'd be able to make that argument only if it was implemented via an extension (users device fetched the page and posts the final document to the LLM for processing).
And that's ignoring the fact that adblockers are seen as illegitimate by a lot of websites too, and they often try to block access to people using these extensions too.
Re: OpenAI Announces SearchGPT
#126Perplexity style is definitely better than Google. I can't remember the last time I googled something.
Having it part of the OpenAI subscription, and assuming it will be accurate, I think Perplexity will have a problem.
Let's see how long the waiting will last, still waiting for Sora access!
Re: OpenAI Announces SearchGPT
#127This is in a way very good, as it will be another differentiator between smart people and others: average Joe will indulge himself in hallucinations while a few will retain the knowledge of querying a search engine, citing a source and critically think of what they read.
You can do things "the right way" (find sources, critically think about the info) in either (or other things like books), the common denominator is the person using the tool not going off the first thing they read from the tool as gospel.
I don't know how fully I agree with this, to be honest. Or, yes, I do - but we obviously outsource a lot to the tools we use. Rightly or wrongly we expect our search engines to source good sources and high quality information. And that has been sort of inbred into people through years of internet use.
Degrading the quality of the output of tools that were once considered reliable and safe is going to create problems, and the onus is not fully on the people using those tools, but on those creating them.
Re: OpenAI Announces SearchGPT
#128> Getting answers on the web can take a lot of effort, often requiring multiple attempts to get relevant results. Don't get me wrong, search has become extremely problematic, ... but how much effort does it take really? Compared to writing a letter, reading a map, walking half a mile etc? The google results for "music festivals in boone north carolina august" are completely adequate. If not then you search again, sur…
Current search is great for facts, alright for generic questions, and annoying for answering something. AI has almost the inverse balance, pairing the two is a decent savings (and also what Bing/Google are trying to do from the opposite end of things). Whether this gets AI out of being a money pit remains to be seen.
> Current search is great for facts, alright for generic questions, and annoying for answering something.
I feel like Google used to be pretty great at answering questions until they started pushing ads more aggressively into their results, and then intentionally gimping their results (for instance if you put quotes around a word Google will happily ignore it, whereas this used to actually be an enforced rule).
I'm not sure AI is the solution to this problem at all. I think it's about misaligned incentives and trying to shoehorn a new cash cow into the mix.
Re: OpenAI Announces SearchGPT
#129Too bad it's only a waitlist. iFixit apparently published fake repair guides [0] that then got crawled by ChatGPT for training [1] [0] https://www.ifixit.com/Guide/Data+Connector++Replacement/147... [1] https://chatgpt.com/share/e52dc4dd-77e6-48a5-a7ca-77e3dfa39e...
>then got crawled by ChatGPT for training
Not really, from the example you provided I think it's pretty clear that a custom system prompt was used and ChatGPT is using its own creativity, it doesn't have anything in common with the iFixit guide.
Re: OpenAI Announces SearchGPT
#130Earlier quoted context omitted.
No, and that is a big problem, search doesn't make money either. People will not actually pay the cost for AI once they have to.
They might pay for it indirectly. For example, Apple just signed a deal with OpenAI and I could imagine a future where users of Apple devices get free access to some AI because Apple and that company made a deal.
Well, maybe, but I think the huge thing missing from this assumption is that Apple are not paying OpenAI anything, and are developing their own in-house and on-device models.
And a lot of what those on-model devices can do are what the average person will want.
And there's going to take a lot to move people meaningfully away from google.