Is OpenAI building a search engine?
131
39k
Dec 31
77%
chance

This resolves as NO if by the end of 2024 there is no evidence of a search engine feature in ChatGPT. NO would expect an absence of evidence including any mention by @Sama, official Twitter announcements, blog posts, FAQs in Help content, a search feature in ChatGPT itself, partnerships with publishers for real-time information, or any other confirmation that would dispel the rumors of a search engine.

This resolves to YES if there is an official announcement, ChatGPT no longer relying primarily on Bing for results (60% or more searches using non-Bing), product features, partnerships, or other demonstrative materials indicating that ChatGPT sources real-time information for answering questions.

Search engine definition: A search engine is a software system designed to search for information on the World Wide Web by processing user queries and returning relevant results from its database or index of web pages.

As of May 2024, ChatGPT uses a browser tool with Bing to index potential websites. A key limitation is Bing’s unreliability in providing accurate answers about news events within the last 24 hours.

A simple test: Ask ChatGPT about a recent headline from the past 4 hours covered by major news outlets—for example, the death of the Iranian president, Scarlett Johansson’s lawsuit, Rabbit R1 fraud, signs of GPT-5 as AGI, or Tesla Optimus being sold exclusively in Texas.

Get Ṁ600 play money
Sort by:

Search engine definition: A search engine is a software system designed to search for information on the World Wide Web by processing user queries and returning relevant results from its database or index of web pages.

Under that definition, DuckDuckGo would be considered a search engine, while Kagi would not (even though they self-describe themselves "a premium search engine").

Are you sure that's what you want to resolve on?

As of May 2024, ChatGPT uses a browser tool with Bing to index potential websites.

I don't think that's true either. ChatGPT seem to use Bing (i.e. query Bing's index) for search terms provided by the LLM, can then visit sites using ["ChatGPT-User"](https://platform.openai.com/docs/plugins/bot), and of course crawls the web to train future GPT versions as ["GPTBot"](https://platform.openai.com/docs/gptbot), but nothing suggests that they're building a "classical" web search index at this point.

@lxgr lukas, you sound upset by this market. my best recommendation is for you to create a new one that you define as you'll be able to control the parameters more to your liking.

regarding your bing point. this is true – that chatgpt relies on the bing index – as shown when bing was down last week, leading to chatgpt being unable to search and cite websites in its answers during that period.

@0xSMW here is screenshot proof from that time period in case you aren't able to source it yourself

@0xSMW So in other words, you won’t clarify the significant ambiguities pointed out to you by multiple commenters, in a market you’re trading in yourself?

I’m not upset because I didn’t trade here, but I’m leaving this here for anyone considering to do so.

@lxgr You seem to just want to pick a fight. DuckDuckGo is considered a search engine by their users, most notably their biggest public fan, Joe Rogan. If you have legitimate questions about truly ambiguous information, feel free to ask. I placed a single trade after treating to showcase where my opinion stands but I’m happy to be wrong.

Like I said: if you don’t like the definition of a search engine in this market, you may create your own. I’m happy to support you and your market and trade in it as well.

DuckDuckGo is considered a search engine by their users, most notably their biggest public fan, Joe Rogan.

I'd consider both DuckDuckGo and Kagi search engines (although I'd maybe call the latter one a "meta-serarch-engine" for clarity)! Not sure why Joe Rogan's definition of a search engine should matter here, given that I don't think he's a subject matter expert in the ontology of hypertext information retrieval.

Your definition, however, would not include Kagi, since they don't operate their own index, yet you wrote:

A search engine is a software system designed to search for information on the World Wide Web by processing user queries and returning relevant results **from its database or index** of web pages. [emphasis added]

And to make it more confusing:

This resolves to YES if there is an official announcement, ChatGPT no longer relying primarily on Bing for results

You could do a lot to increase my faith in your willingness to resolve this on good faith by clarifying how this would resolve if

  • OpenAI switches to Kagi as their exclusive search provider (and Kagi does not develop their own index/crawlers in the meantime) but does not build their own index.

  • OpenAI starts their own web crawler for a few affiliated news sites only, and otherwise still defers to Bing and does not build their own index.

  • OpenAI starts multi-sourcing search results from various search providers other than Bing (e.g. Google, DuckDuckGo), but does not build their own index.

Would the Bing-replacement have to be something built by OpenAI for it to count?

If they're still using Bing, but it has realtime data does that count? If they're using some other search provider (e.g. DuckDuckGo), does that count?

Based on the title, I would have guessed that YES on this question meant OpenAI is building the indexer itself (so they're not relying on any 3rd party search provider), but based on the actual text of the description it sounds like it's just about whether they're going to have realtime data or not.

Cam you clarify which interpretation is right — is this about whether OpenAI is going to be running the indexer themselves, or is it about realtime data?

@rogs Can be either. It’s possible to use third party data and obfuscate this, so not something that is a hangup imo.

@0xSMW I don't understand this answer.

> It’s possible to use third party data and obfuscate this

Okay, so that sounds like you're saying they don't have to build the indexer for this to resolve YES. Is that right?

And then does that mean that it's just about whether they incorporate realtime data? If they have realtime data but nothing else about the current UI changes, would that get a YES resolution?

@rogs it's possible to source third party index data as part of their backend service and never showcase this to the end user

for the second part of your question, please read the description again. there's plenty of context beyond UI changing or not.

Needs a clearer no resolution criteria

@RemNi added

Resolves NO if no evidence as of EOY 2024?

@Nat added

Comment hidden

More related questions