Google devising radical search changes to beat back A.I. rivals
nytimes.com
nytimes.com
I have gotten increasingly frustrated with the results that even 6 years ago were more relevant.
When I put quotes around a specific phrase, I only want direct matches to those words first, before showing me anything else.
Yesterday an uncle was asking me if there are RV parks in SF or nearby. Google proceeds to show me a list of every dog park near my zip code instead. I quote "RV parks" and get similar. When I finally spelled out recreational vehicle parks, I got better results. And every result literally has "RV park" in the title. (Turns out there are 2 in SF. Candlestick park has bad reviews, and the link to the one in the Presidio is 404.)
In a year or two all of this AI stuff is going to calm back down when people finally realize that for other than a few product types, sprinkling "AI" onto everything like cheap MSG isn't going to make the core products any better.
I don't want a "personalized" search experience. I want accurate results for the damned thing I'm searching for. Those should be exactly the same results everyone else gets when looking for the same thing.
The reason the search results suck is because the web sucks. #1 search results for recipes crash iOS safari while scrolling past hundreds of ads to get to the ingredients. This is the result of Google's design. Google got us all into this mess, and they realistically have no way out with search as we know it.
While not perfectly accurate today, LLMs are able to troll through Google's mess and pull out the information people are actually looking for. Over time, Google will be forced to conform and will do the only thing they know how - use your personal information to tune an LLM just for you.
It's sad, but the data Google has on people will likely give them a massive edge over Bing and other competitors in the long term
I strongly believe this is a transient situation. LLMs are so useful today for generic search only because the scoundrels on-line got taken by surprise. Language models process websites differently than traditional search engines, and are able to see through all the usual SEO trickery. But it's only a matter of time before the scoundrels adapt, before SEO turns into LMO, before websites get redesigned to be full of subtle, plausibly deniable prompt injections, and the content itself optimized to maximize the amount of advertising payload that gets picked up as stowaways by LLMs, attaching themselves to the bits the language models extract.
Tuning the LLM "just for you" won't help you, because it'll still end up as an involuntary salesman for any vendor that applies LMO techniques. And even if it was possible, Google is the last company I'd trust with doing it.
They probably assume the same, but the problem with this assumption is that it's self-fulfilling. The reason power users are so small a group is, in part, because software and UI/UX design trends actively prevent it. Especially with SaaS, we live in the era of perpetual MVPs - features get offered the moment they clear the "demo" bar, and hardly ever improved afterwards, except to cut them down further. There is no space for mastery - what you learn to do in first 30 seconds is... all there is to do. More powerful functionality, if it existed, is already removed, after being first hidden away in a non-discoverable "advanced" menu, and then removed, citing low usage. Well, no shit, people won't be using things you go out of your way to hide from them.
And then there's the "attention economy", to which power users are anathema, as the attention economy's core idea is to make things as hard and as annoying as possible, and then make money on friction.
So honestly, I wouldn't trust such A/B tests alone - they generally cannot distinguish between something people like, vs. something people hate and have no chocice but to make do.
I'm not sure what they do to prevent learn-to-rank from over-optimizing the median case, but it seems to not be enough for the tastes of many power users.
recipe sites are terrible because the ones you find have been doing the google thing of increasing engagement, lots of info - blah blah.
No penalty for tons of ads from an ad company.
Google destroyed blog rolls, web rings, and so many other things that were user positive.
In some cases the google of years past was prodded to change (much of the censorship the hides so many results that were once popular back in the day are result of legal pushing or other outrage campaigns)
- however much of what they consider good to be in the top 10 is good for selling ads.
Google is the new yellow pages and that's all it will be so far as search goes imho - people will find what they want searching within portals and other apps more than using google for everything
- 'google it' - is no longer 'finding [obscure thing here'] - google it just means find what is curated... more and more it's 'go to reddit, amazon, tik, yout.. etc - google is known for being 'ad heavy' and prude.
Currently still good for finding accurate hours someplace is open, if you know the exact name and location of said place,
but if you are looking for a tow truck, locksmith, appliance repair, plumber, etc - and depend on google to find one - just like in the yellow pages - the first ones you find are over priced, and the deeper, smaller results may not be in business any longer -
cant wait for the less censored, more 'find the exact bit', less scroll through ads, 'definition of [thing you were looking for]' - ads, 'history of [thing you were looking for]' ads - 'part 1 of [thing you were looking for]' - ads, here's the 3 seconds of info you needed - ads
give me a different bot for each search, and less google and ads any day.
if brave browser or DDG can give me the answer without the 'history or, why you might want to bake a salmon, and 3 ads before I see it (in the search results), and 3 days before I get to the tiny info.. I could like that.
I've even enjoy a link to the info (that skips down the page to the info, auto-blocks any auto play video or audio or moving gifs/ html5 moving ads / banners) - then I'd actually enjoy searching, reading, and even clicking to verify context again.
Oh the web before google censored it into it's own desired ad filled prude image. It was a web and not a silo.
People who are actually passionate about a topic and write about it for the sake of sharing often don't have the resources or even the desire to invest in SEO and great mobile experience. Crappy commercial sites, meanwhile, can just throw money at SEO and developers to ace all of Google's technical metrics.
Google forgot that at the end of the day, it's the content that matters, not how the site looks or its domain authority.
[1] https://www.google.com/search?q=%22rv+parks%22+san+francisco
This is why "personalized" search is a terrible garbage idea.
The fact that it'll hardly return slow loading or older HTTPS-less (non-commercial) pages has essentially memory-holed a huge part of the internet, containing a lot of valuable and historically important litigation. I'm constantly frustrated in research in looking for material which is out there but which even knowing exact strings from it is difficult to get google (or Bing, which was bad in this way first and is often no better than google) to return.
But I think the main draw of Phind and similar alternatives isn't necessarily that it's that much better at getting the necessary information, it's just so much less annoying to use because it cuts out all the newsletter popovers, bait and switch blogspam, cookie consent popups, SEO dark patterns etc. You don't have to mentally filter out a bunch of nonsense.
Search engines are annoying. I say this as I spend time and effort developing one.
Have you incorporated any embedding model in your search engine?
FWIW, I personally enjoy using marginalia and like its current spartan + fast design.
Closest thing is explore2.marginalia.nu and similarity pagerank, but it's all very basic and unga bunga.
So, unless you and someone else are both searching over Tor (maybe you even need exit nodes in the same country), you're probably not going to get identical results for identical queries. This is what people are complaining about when they complain about Google "personalizing" search results.
Long-winded recipe pages are the canonical example, but last week I was just looking for info about a place I'd be traveling too, and had to scroll through pages of useless fluff to just get the info I was looking for.
It was surprising to realize how much I have stopped using Google entirely since ChatGPT will just tell me the answer or provide additional context, all from the safe, ad-free interface where I already am.
It's my hope that technologies like this will finally kill the ad-driven internet paradigm. More than social media, it really seems to be the root cause of all of the undesirable things we have seen sprouting like weeds across our relationships, media, and political institutions for the past 15 years.
There was a blatant "port list" to the results.
To find some commentary by a nominally centrist voice, I had to search on a blog.
More pure mechanism, less bias, please.
I mean, the post is directly in front of you. Might you quote accurately? Or were you too busy condescending?
Is it really so hard to talk about these things as they are?
they can't imagine a world where they aren't allowed to spy on everybody, but regulations are starting to have an impact there.
they can't imagine a world where ads aren't the way to make money, and have 25 years of failing to find a new way.
for all the "moonshots" they are still a one trick pony.
I found it exceptionally great.
As an aside, if you ever are in BC with kids, check out Kidtropolis. I did, thx to Bing+ChatGPT
Bing hasn't changed their underlying indexing tech. Adding a very fancy layer of frosting on top doesn't help the styrofoam underneath.
Google's results have gotten so bad these days that at least a couple times a day when I'm coding I try the same queries about an error message or something on both sites out of desperation. Even then I usually find Bing to be even worse. Then I'll try the old "site:reddit.com" trick on google and finally get something useful half the time.
I've also recently used GPT to help with an issue I was dealing with. I was trying to change the default location of where postgresql keeps its data directly. GPT gave a very convincing solution, except that after another few hours of tinkering and searching turns out to be completely wrong.
what you may not be doing is reading the explanation or pasting in your error messages. very rarely does it make an error that isn't trivial to overcome and in far less time than a hunt through "10 blue links".
Once you have everything set up and configured it is effectively an API that does arbitrary work for you.
OpenAI waiting lists for features and API access and message limits are separating classes of people of how productive they can be in life.
As for me I have a comfortable life but for some classes of people, like content creators or even programmers being on the waiting list vs having access to some features can be the difference between having a job or not and having a promotion or not.
That is not to say Google doesn't do a lot of bad/shitty things, it's that they are just "middling shitty" compared to all the scammers out there.
I stopped using Google Search years ago, not because they were doing nothing, but because they were doing too much. Very often I could not find a rare combination of terms because they refused to believe that I hadn't made a typo. They also killed the use of quotes and + signs to override, until it was their (useless) way or the highway.
Unfortunately DDG is following their steps lately.
Now I can only use search in very limited ways. That's a shame.
The theory is at least plausible given what we see. They'll kill many products that seem viable, and would help diversify their revenue stream. Does Google really place no premium at all on that?
For example, killing Stadia after only a few years, while predictable for Google, is still a head scratcher to me. Building a platform and letting other companies (in this case, game studios) take the big risks on what will sink and what will swim, is a hugely privileged position to be in. Why did they kill it when others are moving into the space?
Its like they built a global real time stock market for Ads. For any possible search query, or space on some any page you click where an ad can be displayed, there is an real time auction running in the background where bidders, your digital data profile, sellers are all brought together to decide whats going to get displayed. And it all happens by the time your page load.
On point, the path from search engine to AI chatbot, in case it does succeed, is quite dramatic in any case:
It might be economically challenging. My understanding is that an AI query is much more expensive to run whilst ad potential is lower. They can place an ad next to an AI answer, but Microsoft might not and subsidize it.
But that's not even the biggest threat. If the AI chat is as good as to give a direct answer, it means you won't hit the source website. Which currently runs ads. This might mean websites getting a fraction of their original traffic in the long run and ad impressions decimate.
Not to mention that even having a website becomes relatively pointless. Almost nobody would ever visit it, you're just a content slave for the big AI machine. You do all the work but get none of the rewards. The machine does no (human) work and gets all the rewards.
This breaks the content "contract" of the internet. The incentives were already bad, but its going to get a whole lot worse.
Engagement on individual websites has been tanking for over a decade, it's truly starving, a few exceptions aside.
The open web might as well cease to exist mostly, which might be compensated for with private spaces. That trend was already ongoing.
People are drawn to things that look cool and innovative for the sake of it, but we're not good at thinking ahead to how well they'd work in practice.
I've been searching the web for so long that it's like breathing. I type a few keywords, hop between the first few results very fast, get what I need and get out. Maybe I have to repeat. Few searches are more complicated than that.
I've tried replacing that flow with AI. It's slower to input, slower to get results, the results are opaque with regards to where they came from and how trustworthy they are. And the results aren't better, in my experience as someone for who finds search to have been solved and painless for many years.
Even asking one question of an LLM takes more time and effort than a traditional search, and the problem is compounded when I have to iterate on a search.
(Also it sucks to cede the kind of power to search providers that gives them leverage to force us to be logged in and monetize what's always been free, but those are just random complaints.)
So, going back to your question. My guess is that Google has always known this isn't a better overall experience, once you get over the initial wow factor. But maybe they underestimated how much that wow factor would allow true competition to materialize before the rest of us would figure it out and lose interest.
A search engine is a great peripheral for a language processor. Bing chat does show how you’re supposed to connect the two, as are ChatGPT plugins.
LLM can do synthesis and summaries of multiple search results. It is slow to get a response and the thing makes mistakes sometimes but we’re in the first quarter of the tech being deployed. It’ll only get better.
I've been using DuckDuckGo as my daily driver for years. It's generally OK. It sucks, but just barely works well enough to keep me going.
I started using Perplexity and ChatGPT for some searches that DDG was having trouble with, and it blew me away with the quality
Then I started going back and trying the same searches on Google, and I was surprised to find that it was just as good as the LLM-powered results after all. I don't know if phrasing things as questions was the key or just using Google instead of DDG did the trick, though
Call this a 90/10 split.
Of that 90, I would say about 75% of my questions can be answered by GPT* trivially. Most of this is due to the fact most of my problems are not novel.
Of that 10, the results are much more all over the place. I would say half or less.
The thing is, when you get crappy results from the LLM, they are genuine technical limitations (as long as you avoid politics and religion, if you are researching those with the LLM, good luck with that), not intentional badness put there to make someone else a few pennies off of wasting my time. There is a certain value there that is hard to quantify.
The other killer feature is being able to have a conversation about the results. This, for me, pushes the usefulness factor way above Google. The options for having conversations about technical topics with humans are comparatively god-awful.
You're probably not using it right. :)
Its utility as "search" is only partially overlapping with search. It's most useful for that when you can describe what you want but not with clear enough terms to search for it. For research now, I'll try the search engines-- if I don't get hits I'll ask GPT4. Even if it gives me a wrong answer it'll almost always use terms that let me form a more useful search, though often it just gives right answers.
Simple example, last night a old friend emailed me to ask if I remember the name of a 90s website that people used to brag about linux uptime and deployment. He couldn't find it via search. I remembered what he was talking about, couldn't remember the name, also couldn't find it via search. Asked gpt4 (by simply pasting in the email), it answered that it was probably the "linux counter".
It's possible that with more guessing and grinding on search terms I might have hit the Wikipedia article on it, but GPT4 gave an immediate and accurate answer and included the URL which is what I needed to load the page up on the internet archive (as the site is long gone from the internet).
The main thing GPT4 can't do is any kind of exact string search -- but unfortunately google more or less won't do that either anymore. In terms of figuring out what you intend from some vague text GPT4 usually does vastly better than google.
I think my ideal tool would in parallel give me an not-very-interpreted classical search result and a state of the art LLM result. It seems that Google's bard works kind of in that direction, but their LLM seemed extremely dumb and hobbled when I tried it even compared to GPT3.5 and the search elements are google. ... so worst of all worlds. :)
So far, the only use-case I've seen for LLMs are basically code completion.
There are dozens if not hundreds of other uses out there.
It would need the ability to contain infinite context to be genuinely useful in my opinion.
PS. Just read the Lemoine's transcriptions. Just wow.
This is all to say - Google could have not let it’s search product become a hopeless piece of garbage before the threat of real competition emerged.
My kids will just have a device that responds to their queries using AI.
If that device isn’t made by Google, it won’t use Google search.
Amazon, Apple, Microsoft, Nintendo will all aggregate data with their own AI front end.
Google’s saving grace is Youtube.
That’s all well and good, but with no internet of blogs made by people … what will the AI be trained on?
I’ll rephrase it, what will compel content creators?
For some authors, it’s the desire to pass on knowledge.
For others, it’s montetization.
Just as the internet decimated the print business, Generative AI will decimate the content business.
My dad bought Chilton’s automotive manuals.
I used Google/YouTube.
What will my children use?
Just tried with "Keto garlic noodles recipe" (something I've been playing with recently). 1 result, then shorts, 2 more results, then MORE shorts, then more results that have mixed relevancy. Oh yup and a pro-sugar anti-keto propaganda video mixed in because, why not.
Scroll down a bit more and yup, more shorts section. Keep scrolling and now YouTube is just injecting short videos in the results feed like normal videos instead of the horizontal short sections like higher up.
YouTube search is f-ing broken. Sorry for the language.
Unless they keep treating creators like trash and eventually a more creator friendly platform comes around that kids glom onto.
I've never seen a phase change in user behavior like this before, and it strikes me that it may have had a material impact already.
I guess that means it managed to get to her the long way around human society, since Japanese evening TV news is pretty much at the rear end of any cultural developments..
I general that describes what little I've used ChatGPT for.
I have it write something on a topic I know a fair bit about... "OK. that's pretty good. That's veering into weeds. That's not right. I'm going to work a bit on how this is structured; it's pretty formulaic. Needs some actual references and maybe a quote." If I had to start from not knowing anything, I'm not sure how useful it would be. But when it's "just" a matter of writing about a topic I know, ChatGPT can fill in a lot.
On the topic of languages, I had four years of high school French and have dipped in here and there (very dipping) since. If I wanted to put some effort in advance of maybe traveling some more in France, any recommendations? I know Duolingo but that seems to be more vocabulary.
How's that? It can't even give you pronunciations.
OpenAI is well positioned and has the lead, but the field is wide open at the moment.
The product and features mentioned here are certainly interesting and could be indeed useful.
I am surprised how their communication team has handled this, they appear to be calm, in this "we can do this, we can recover" mode, they aren't dismissive of competition or rumors as shown in their official responses in this article.
And unlike others here, I really think its googles game to lose, they have much more data than anyone, they can customize and improve the experience much more than Microsoft can. And I doubt they are truly as behind GPT-4 as LLaMDA showed.
Competition is definitely needed in the search space and we finally have it.
Two weeks? That’s a pretty slow response to an existential threat
I believe Google earnestly worked on this for many years, but about a decade ago, after losing the lawsuit with Google Books, this moved to the back burner.
The number one issue to making information universally accessible and useful is no longer technical, it's legal. It's copyright law. Google is making quite a lot of money from the status quo, and is choosing profits over their stated mission. I can't blame them. They carried the torch quite far.
Now OpenAI is running with the torch. They pretty much said "FU" to copyright law and trained on everything anyway. (Which is a great thing!). And now, in many many cases, are delivering on Google's stated mission better than Google.
while this method may work, as often as i see this type of comment, i’m guessing reddit’s quality of information will be worse than even google search within months
Guess you can’t have a cake and eat it too.
- they want to add Bard into search
- they want to add "AI features into maps"
- they want to add a tool to create AI images from Google Image Search
So no, they're not actually working on fixing their issues with search, which IMO is in its worst state since creation.
Google Earth*