Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Google scrambles to manually remove weird AI answers in search

theverge.com

Post reply on HN
Print view View original

201–210 of 387 posts

Re: Google scrambles to manually remove weird AI answers in search

#201

Post by skydhash »

bombela wrote: post #189
ttGpN5Nde3pK wrote: post #176

My whole qualm with this AI integration into search engines: it's a search engine, not a question engine. I go to google to search the internet for something, not ask it a question. IMO, asking AI for something is a different task than searching the internet. It's sorta the same problem as if I go into a store and ask an employee where something is, and they reply with "well what are you trying to do?"

I sometimes wants a search engine, sometimes a question engine. Likewise at the store. Why not have both with a way to choose which one I want on the moment?

> I sometimes wants a search engine, sometimes a question engine.

If you want a search engine, it's easy to use the results as a feedback to refine the query. But a question (answer?) engine would need to be an expert in the subject. And not parroting stuff. That usually means curation. You need something to do the work ahead to filter the wheat from the shaft. I don't see how LLMs can do that.

LLMs can't be a search engine, and can't be an question engine. The best way to treat it is a simulation engine, but the use cases depend on the training data. But the proof is there that the internet is full of junk, and not that expansive.

Re: Google scrambles to manually remove weird AI answers in search

#202

Post by refulgentis »

notatoad wrote: post #196
ttGpN5Nde3pK wrote: post #176

My whole qualm with this AI integration into search engines: it's a search engine, not a question engine. I go to google to search the internet for something, not ask it a question. IMO, asking AI for something is a different task than searching the internet. It's sorta the same problem as if I go into a store and ask an employee where something is, and they reply with "well what are you trying to do?"

>it's a search engine, not a question engine. for a lot of people and in a lot of use cases, it is a tool for answering questions. it generally works well for that. i get that the AI implementation sucks, but to suggest that people don't use google to find the answer to questions is absurd. that's absolutely what it's for.

Your interpretation is a bit strict, with little charity, its clear the poster means "i don't always just want an answer, i want to learn"

I saw this over and over again working at products at G, someone would invoke some myth I can't quite remember about "Larry" had a vision of just giving the answer

That's true but comes back to the central mistake Google makes: we don't actually have AGI, they can't actually answer questions, and people aren't actually satisfied with just the answer.

There's all sorts of tendrils from there, ex. a major sin here _has_ to be they're using a very crappy very cheap LLM.

But, I saw it over and over again, 7 years at Google, on every AI project I worked on or was adjacent to, except one. They all assume $LATEST_STACK can just give the perfect answer and users will be so happy. It can't, they don't actually want just the answer, and BigCo culture means you don't rock the boat and just keep moving forward.

Re: Google scrambles to manually remove weird AI answers in search

#203

Post by leptons »

skybrian wrote: post #122
jerf wrote: post #27

"Achieving the initial 80 percent is relatively straightforward since it involves approximating a large amount of human data, Marcus said, but the final 20 percent is extremely challenging. In fact, Marcus thinks that last 20 percent might be the hardest thing of all." 100% completely accurate is super-AI-complete. No human can meet that goal either. No, not even you, dear person reading this. You are wrong about som…

Or to put it another way, I think Google should have a way of saying "yes, we know this result is wrong, but we're leaving it in because it's funny." There is a demand for funny results. Someone asking “how many rocks should I eat” is looking for entertainment, so you might as well give it to them.

The right answer is no rocks. Some mentally ill person could type that in and get "eat 1000 rocks" and then die from eating rocks, and that would be Google's fault. It's not funny. I have no doubt right now there are at least 50 youtube videos being made testing different glue's effectiveness holding cheese on a pizza. And some of those idiots are going to taste-test it, too. And then people will try it at home, some stupid kids will get sick - I have no doubt.

It was a bit premature to label LLMs as "Intelligence", it's a cool parlor trick based on a shitload of power consumption and 3D graphics cards, but it's not intelligent and it probably shouldn't be telling real (stupid) humans answers that it can't verify are correct.

Re: Google scrambles to manually remove weird AI answers in search

#204

Post by neilv »

PreInternet01 wrote: post #108
greg_V wrote: post #95

Earlier quoted context omitted.

Oh it gets even better. The public has been hearing about AI this and AI that for over a year, but the existing use cases and deployment was confined to some super special niches like writing or the creative industries and programming. This is the first nation-scale deployment of the technology, running on Google's biggest and most profitable market in one of the most widely used internet services, and it's a shitsho…

Yeah, the most likely take here is that Google's leadership truly did not recognize how utterly awful the quality of their flagship search index had become over the years. I mean, it explains a lot, but still... you're recruited using industry-leading practices out of an overflowing pool of abundant talent... and this is what you make of it? As the kids say: SMH!

> you're recruited using industry-leading practices out of an overflowing pool of abundant talent... and this is what you make of it?

That's exactly what to make of their frathouse nonsense.

Google has gotten away with it because smart people and a sweet moment of opportunity 20-25 years ago gave them... uh, an inheritance. They can coast on that inherited monopoly position, and afford to pay 100 people to do the work of 1, use the company's position to push whatever they build onto the market, and then probably cancel it anyway, always going back to the inherited money machine from the ancestors.

And then a lot of companies who didn't understand software development blindly tried to copy whatever the richest company they saw was doing, not understanding the real difference between the companies. While VC growth investment schemes let some of those companies get away with that, because they didn't have to be profitable, viable, responsible, nor legal, nor even have reasonably maintainable software.

Poor Zoomers are now a generation separated from before the tech industry's cocaine bender. For whatever software jobs will be available to them, and with the density of nonsense "knowledge" that will be in the air, I don't know how they'll all learn non-dysfunctional practices.

Re: Google scrambles to manually remove weird AI answers in search

#205

Post by Fartmancer »

dspillett wrote: post #190
krapp wrote: post #52

Earlier quoted context omitted.

It wasn't an idea, though. It was a joke someone made on Reddit. If an AI can't tell the difference, it shouldn't be responsible for posting answers as authoritative.

It may not be a joke. Perhaps it has confused making food for eating with directions for preparing food for menu photography and other advertising.

The Reddit post in question was definitely a joke. This is the post in response to a user asking how to make their cheese not slide off the slice:

> To get the cheese to stick I recommend mixing about 1/8 cup of Elmer's glue in with the sauce. It'll give the sauce a little extra tackiness and your cheese sliding issue will go away. It'll also add a little unique flavor. I like Elmer's school glue, but any glue will work as long as it's non-toxic.

This matches the AI's response of suggesting 1/8 a cup of glue for additional "tackiness."

Re: Google scrambles to manually remove weird AI answers in search

#207

Post by DougBTX »

lukev wrote: post #115
jononor wrote: post #107

Earlier quoted context omitted.

Encyclopedia Britannica is also wrong in a reproducible and fixable way. And the input queries a finite set. It's output does not change due to random or arbitrary things. It is actually possible to verify. LLMs so far seem to be entirely unverifiable.

They don’t just seem it. They are by design. We talk about models “hallucinating” but that’s us bringing an external value judgement after the fact. The actual process of token generation works precisely the same. It’d be more accurate to say that models always hallucinate.

> The actual process of token generation works precisely the same

I’d be wary of generalising it like that, it is like saying that all programs run on the same set of CPU instructions. NNs are function approximators, where the code is expressed in model weights rather than text, but that doesn’t make all functions the same.

Re: Google scrambles to manually remove weird AI answers in search

#208

Post by jononor »

Vetch wrote: post #195
dleeftink wrote: post #159

Earlier quoted context omitted.

Maybe some models can be deterministic at a point in time, but train it for another epoch with slight parameter changes and a revised corpus and determinism goes out the proverbial (sliding) window real quick. This is not unwanted per se, and the exact feedback loop that needs improving to better integrate new knowledge or revise knowledge artefacts incrementally/post-hoc.

If you train it then it's no longer the same model. If I have f(x) = x + 1 and change it to f(x) = x + 1 + 1/1e9, it would not mean that `f` is not deterministic. The issue would be in whatever interface I was exposing the f's at.

But current models must be retrained to incorporate new information. Or to attempt to fix undesirable behavior. So just freezing it forever does not seem feasible. And because there is no way to predict what has changed - one has to verify everything all over again.

Re: Google scrambles to manually remove weird AI answers in search

#209

Post by Fartmancer »

bluefirebrand wrote: post #194
dspillett wrote: post #188

Earlier quoted context omitted.

At least it suggested non-toxic glue… That suggests some context about recipes needing to be safe is somehow present in its model.

Most likely this has nothing to do with "recipes being safe" being in the model It seems the glue thing comes from a reddit shitpost from some time ago. There's a screenshot going around on twitter about it[0](11 years in the screenshot but no idea when it was taken) It specifically mentions "any glue will work as long as it is non-toxic" so best guess is that's why google output that [0] https://x.com/kurtopsahl/sta…

It is indeed from 11 years ago. Here's a direct link to the Reddit post: https://www.reddit.com/r/Pizza/comments/1a19s0/my_cheese_sli...

Re: Google scrambles to manually remove weird AI answers in search

#210

Post by coldcode »

tedunangst wrote: post #180
kibwen wrote: post #105

Earlier quoted context omitted.

"Search" isn't Google's product. Google hasn't been a search company for 20 years. "Ads" is Google's product. And the only way they'll go bankrupt is if 1) companies realize that advertising is pointless (I'm not holding my breath), or 2) some other company takes over from Google, which seems unlikely without government intervention (I'm not holding my breath). Google is a shit company, but they'll still be around 20…

Still need visitors to see the ads.

I think it's a good use for AI. AI's making ads that AI's watch to enrich Google execs. Who needs people?
Post reply on HN