making the feature rely less heavily on user-generated content from sites like Reddit
Imagine selling out reddit/buying access to the comments for AI just to immediately unprioritise it
A new deal is being forged with 4chan instead.
That would be awesome!
Straight and real
No one could have seen this coming. Obviously LLMs absolutely completely understand the difference between people joking around with each other and authentic advice.
Sarcasm aside, Reddit does have some good information about niche topics. There’s just currently no way for AI to understand the difference between dry humor and serious responses. I think the AI summaries are unhelpful anyway, but even if I didn’t it’s pretty obvious Google didn’t think further than “Shove AI into it like a drunken prom night encounter”.
There’s just currently no way for AI to understand the difference between dry humor and serious responses.
There’s no way for humans to do this either. We invented an entirely new series of markers (emoticons, /s, etc.) at a vain attempt at doing just this and still fail.
It’s definitely difficult over text unless you know someone.
Who do we have to know
His name is Dennis. Once you really get to know him you’ll understand.
Reid characterized the mistakes that won attention as the result of an internet-wide audit that wasn’t always well intended.
Oh yeah, there was a conspiracy to embarrass Google with off-the-wall questions like “how do I thicken pizza sauce?”
Why are these corporate schmucks unable to take any blame?
No audit is well intended. The whole ppint of an audit is to find duplicitous and fucked up shit.
Why are these corporate schmucks unable to take any blame?
Because they never have to.
It wasn’t us, it was the people who used it who really caused the failure.
Ungrateful fucking peasants are not recognizing greatness?!!?
Sadly, b/c their jobs literally depend on them NOT doing so, in today’s corporate hellscape environ:-(.
If Chromebooks are anything to go by, if google had their way you’d only be allowed to search prescreened questions they think are best for you. Can’t have you experiencing anything not advertiser friendly.
The company made “more than a dozen technical improvements” to AI Overviews …
… making the feature rely less heavily on user-generated content from sites like Reddit
So it prefers the results that Google normally deprioritizes? I guess we have that in common
I almost wonder what Huffman is thinking right now - he could have been so very close to becoming a billionaire, managing the repository of human general technical knowledge explained simply, but he blew it by being a greedy piggy…
Not that he will ever admit that, even to himself.
I don’t think Spaz is thinking at all. That was how Reddit imploded.
His decision tree is like: I say yes, content creators say no, admin rights say yes tho, market ends up saying no tho, the end.
It seems like I’ve heard this story before somewhere lately…
I say yes
You say no
You say stop
And I say go, go, go….So many songs about rape… tbf I don’t get that vibe from that Beatles song that seems more about just opposite phrases:-), but in general, we have it far too much in our culture.
Messages simply cannot be removed from their context and retain their meaning. According to Paul McCartney, as you say, this is simply a song about opposites.
The AI bullcrap at the top, before actual search results, is what finally made me change my default engine to duckduckgo
Same, I have tried DDG every once in a while but kept going back to google. Now google search doesn’t quite give me relevant results anymore and all the AI crap just takes all of their effort to work on search itself.
Been using DDG for a few weeks now for personal and work related stuff - quite happy with it.
I’ve fully committed to kagi now and don’t regret it. The results are actually helpful again and even their AI features are better. The search summary does not seem to hallucinate at all
I don’t know about you but the sponsored results and all that search optimisation manipulation made me switch to DDG long before AI
My only gripe with duck fuck go is I don’t allow access to my location and can never google (hehe) shit in my town without adding extra words for directions etc
You don’t allow location and are surprised that it does not know you location? Am I missing something here? Why is that surprising?
Yeah what annoys me aswell is that I don’t like to type anything into the search field so noone can steal my data, buttfuck go refuses to give me results that way
What annoys me about cumslut go is that I like to put in fake searches to throw my FBI agent off the scent and it returns me random shit I don’t really want
What? Oh no, I had not written down the recipe for using gasoline to cook my spaghetti! Whatever shall I do?
Step 1: add gasoline.
Step 2: add more gasoline.
Step 3: don’t think about the bridge…
Step 4: what spaghetti?
I wonder if that bridge has immortalized itself in AI history.
I hope it lasts longer than modern AI tools themselves at this point. They have great potential, but… they cannot replace the (lack of) brains of a manager to be a magic bullet to cure all ill effects of greed, with as little effort put into deploying them as has been done so far.
(From the OG Matrix movie)
They’re also developing computer chips running on actual biological human brains if that helps the brain aspect
Ironically, we keep being told (iirc?) that middle-management was going to be one of the first things to go with the advent of AI. Software was going to eliminate the need for it, allowing one person to manage many tens to hundreds (to thousands?) of people directly.
Instead, most companies - like big tech, and Boeing, etc. - seem to be going the opposite direction, ditching their actual workers who produce things while keeping the managers?
This does not seem to be going all that well for Google lately…
Ok good. Now remove that feature or at least add mandatory human review
And slow down search even more?
Human review is NEVER going to happen. The amount of comments from Reddit and user queries is mind boggling
No I meant that the AI makes replies to most popular (or predicted to be popular) questions but a human has to review and accept it manually to push it to the search results. But that still won’t fix everything. It’s just AI that needs to die. That’s it
It’s an interesting concept. But right now, Google Search is the best argument in favor of Bing.
Or DuckDuckGo, which uses the same index as Bing, but with more privacy
DuckDuckGo is my default search tool. It’s great.
Maybe it is a firedox issue, but i have serious issues with location based searchs on DDG. Like, results in the next state over sort of bad.
That may be a feature of DDG. In respecting privacy, it may be ignoring your location. Since I never let sites have access to location data if I can prevent it, I don’t know.
I know, but it’s frustrating when i search “restaurants in Cityville” and the results show restaurants in Cityville Indiana instead of my homestate. Or i search for “T-Mobile stores near me” and they list off ones 3 counties over instead of the ones near me. I understand that it is a privacy thing, but it is very annoying.
Same thing happens when i look up items on the grocery store website to see if they have something i need. Firefox seems to think i live over an hour away in Dearborn Michigan, so i have to adjust my location manually every time. I understand why it happens and i can accept it , but that doesn’t mean it isn’t frustrating.
Firefox will let you whitelist sites that are allowed to your location. Just whitelist those sites. Or use a search engine that doesn’t respect privacy, say Google or StartPage.
I’m pretty sure I live in a city between you and Dearborn but I don’t want to say the names and dox us. I just thought it was fun to piece together with the few clues and my familiarity with the area
You think Bing aka Microsoft is not planning on this exact same folly?
I’m sure they are. But they haven’t yet. And after this, they might at least borrow a page from Apple and make sure it works first.
I mean… They have though. It’s not in bing.com but “Microsoft copilot” is their newly rebranded Bing + AI search engine, which they’re embedding directly into desktops. They’ve been doing the AI summaries longer than Google has afaik.
I haven’t noticed it appearing unasked in internet search results, and I never use the desktop search except for on-device queries.
Like Apple Maps? Or mms messaging?
Nobody’s perfect.
I don’t know of any MMS problem, but I also don’t generally text outside of iMessages. Maps was a cockup, and Apple owned it.
Owned it by forcing everyone to manually install google maps, which they were already using happily? Or owned it by giving a pr spin after the fact? Group messaging via mms still doesn’t work correctly due to iMessage.
Owned up to the mistake, then suggested alternatives while they were finishing Maps, which had been rushed due to the launch of Android and Jobs not wanting a competitor’s CEO sitting on Apple’s board.
I have no idea what you’re on about messaging. I can use it just fine.
Well, cool. Now fix it.
Everyone remember to pollute your answers wherever you post them at least once a day.
The best way to avoid having your words used by an LLM is to mash your fingers with a mallet before typing. The resulting typographical errors will ensure that the AI rejects your text before using it for training.
So… a nurmal Redhti usre then? :-P
We need decentralized, federated search. I remember YaCy from years ago was attempting this. Anybody know if there’s anybody actively working on this?
@makeasnek @schizoidman YaCy is still around.
And https://searx.space/ is an open source metasearch search engine with many instances. (Try https://searx.be/ if you want to test it out.)
SearX/SearXNG allows you to aggregate results from a number of different search engines. You choose which ones, and they’re stored in your browser without setting up an account.
Interesting thanks I’ll take a look!
@makeasnek On a broader note, I think possibly the best approach for decentralised, open-sourced web search might be an evolution on the SearXNG model.
At the top of the funnel, you have meta search engines that query and aggregate results from a number of smaller niche search engines.
The metasearch engines are open source, anyone with a spare server or a web hosting account can spin one up.
For some larger sites that are trustworthy, such as Wikipedia, the site’s own search engine might be what’s queried.
For the Fediverse and other similar federated networks, the query is fed through a trusted node on the network.
And then there’s a host of smaller niche search engines, which only crawl and index pages on a small number of websites vetted and curated by a human.
(Perhaps on a particular topic? Or a local library or university might curate a list of notable local websites?)
(Alternatively, it might be that a crawler for a web index like Curlie.org only crawls websites chosen by its topic moderators.)
In this manner, you could build a decent web search engine without needing the scale of Google or Microsoft.
@ajsadauskas sounds like what @pears is trying to do
@makeasnek @pears is working on that
Am I the only one who doesn’t have access to this? I’m in canada but using a VPN doesn’t seem to change anything. Not that I care about AI, but I’d like to see the shitshow myself.
Thanks to screw-ups like these, I was on the hunt for a better search engine. I may have found one in kagi.com. Anyone else willing to pay to get rid of tracking, check it out!
