Not sure how this would be different from Bing Copilot. I use copilot a bit and it's great for some general queries about current events or things that are written in many articles. For instance, how many children does Drake have or what's up with Eurovision (referenced in dozens of spammy articles).

But I noticed it overly relies on search. If I ask a technical question, it relies on those same crappy search results, compared to gpt which answers correctly without referencing anything external.

I don't know if you can bridge that divide but I'm curious what openai comes up with and how its different from bing copilot

I wonder if you could prompt engineer a more useful result by asking it to say what it knows and then add synthesis of recent urls as well.

> “People come to us because we are trusted,” said Raghavan.

Well I don’t. Kagi is much better. Thank you, Kagi.

I shouldn’t be emotional about this, but I find I am getting very angry with these large companies with business models that use their users.

There doesn’t seem to be any way for them to avoid getting worse, given their market caps and multiples have years of continued innovation in user surveillance and manipulation priced in.

(And in general, I am not against success.)

I hope the AI industry keeps a healthy paid tier going. Because mass free, ad-driven AI is not going to be pretty.

There's no reason to reserve you ire for large companies, any business that chooses to make their product irritating for some incremental increase in profitability deserves our scorn.

I think irritating vs. insidious are different.

And pervasive vs. easily avoided are different.

And I can think of other differences.

With respect to surveillance especially: Scale has a quality all its own.

I grant that I do notice when any service or product has obvious unaddressed flaws. But that is a different tier from companies globally warping the fabric of society.

I think it's fair to be angry that Google has used their power to, frankly, ruin large parts of the web by incentivizing bad behavior.

I'm skeptical about this, mostly based on my experience trying Perplexity recently, which is I assume roughly what OpenAI's future product will be like. My impression was that it essentially does a web search and then uses an LLM to summarize the top few results.

It's useful for certain types of general knowledge questions, but for anything involving finding something specific, which is probably 90% of my Google usage (for example, finding the URL of some particular project on Github), it tends to fall flat.

Completely agree. Google has the upper hand because you still need a great traditional search engine to weight/rank the content the LLM should summarize. Otherwise you end up spending a lot of extra time sending follow up questions.

Here's an example asking perplexity something I've already researched a lot. I asked what's the best PC monitor for work. It cited several sites including Rtings and Forbes. It then told me the best was a Dell monitor based on the Forbes site. Rtings is definitely the better site for monitor reviews and didn't even rank that monitor as one of the best. I then asked Perplexity, what's the best website for pc monitor reviews. It responded Rtings was the best, listed a few others, and didn't even mention Forbes.

Kagi’s ”?” operator (I’d expect openai to do better than it) suggests RTings, then Display Ninja, PC Magazine, and finally, reddit.

I don’t think Google has much of an advantage in this fight. Presumably, OpenAI will be able to partner with bing if they want.

Thanks to the antitrust suit, we know for sure that Google has been intentionally tanking search relevance in order to prop up their ad display numbers. I’m guessing none of the smaller search engines do that (since they all desperately want to grow their market share).

I think Google has 3 huge advantages.

1. They're the incumbent with almost monopolistic market share. People aren't going to switch to a new product that's just roughly as good. It has to be way better.

2. Their search algorithm is straight up better than their next biggest competitor, Bing. The latter has tried for years to overtake google and failed.

3. They have copious amounts of user specific data across their products/services. That enables tailored AI responses.

You have a fair point about google propping up ad buyers. That is their Achilles heel.

4. Default search on Chrome, Android...

5. Default deals with Apple, Samsung...

fair point - but i suspect satya gave the keys to bing to his buddy sam

Is Google a "great traditional search engine" nowadays? It's a great SEO search engine, but it has been taking me longer and longer to find relevant results on the rare instances I use it.

> My impression was that it essentially does a web search and then uses an LLM to summarize the top few results.

Yes that's a good short summary. More details on how we do it here: https://blog.mojeek.com/2024/04/mojeek-search-summary.html

I expect it would be something entirely different - OpenAI with their resources could build a new search from scratch using dedicated llms/embeddings.

Otherwise it would be not much better than the current bing+gpt4

the AI will continue until morale improves

"This goulash is inedible!"

"Ah, comrade, you are just not eating enough!"

OpenAI has proven to be WAY too opinionated about information access for me to trust it as a search engine. I believe they COULD provide a high quality search experience but they definitely will NOT.

Example queries?

Search for: openapi paternalism

Many, many articles have been written on the implications of the ChatGPT censorship stuff.

I asked gpt this and here are the results:

https://chat.openai.com/share/e/fc40c207-6e8c-430d-9435-c8d9...

Or did you mean that you don’t have specific queries that you did yourself that were too biased, but you’ve read some people had?

I don't use it too much anymore personally as too many attempts literally just failed at dispatch and never made it to the model. I don't want an opinionated search engine. I use Kagi now and it's been pretty good so far.

I also don't know why you seem to be implying to the other user that it's some conspiracy OpenAI is biased. All models are biased. I personally want to be able to search information that I want without a roadblock in the middle determining if I can complete my request or not.

I honestly think Google is completely vulnerable in search currently, the actual results you get back from Google and Image Search are shockingly poor, even get it zclaiming it has no results now for some queries which I just know isn't true because 10 years ago you could type anything in and there would be at least 6 figures if not 7 figures of results for any term.

Actually think even if a western company just launched Google from 10 years ago again today people would switch over because it would be better than what their offering today is.

Not sure if OpenAI will get converts but it's definitely the right time to be taking on those stagnant and rotten products Google has, well neglected isn't the right word but more spent the past few years optimizing the wrong metric to the point of breaking the product.

Wow. I agree, people would sign on for an old version of Google in a second.

I suppose there is no external access to any legacy Google code to make this A/B test. But it would be wonderful!

As an SEO I'm a bit excited and interested to see how they'll be "ranking" information, or whether every source will be paid-ad only.

It will be essentially a bing competitor. They are already doing LLM on search. What is the strategy to compete with Google except ads?