Google's AI is that stupid, feeds people answers from The Onion | Hacker News Reader

Google's AI is that stupid, feeds people answers from The Onion

avclub.com

54 pointsc42047 comments

‘No Way To Prevent This,’ Says Only Search-Engine Where This Regularly Happens

Is it even in their best interest to make it work? I’m sure the normal person would start to ignore the AI answer and scroll down to see the first page of ads and then find the results.

I think you’re wrong. The normal person would almost certainly take the quick summary answer so long as it doesn’t sound too absurd.

Most people use google to answer a question. If that question is answered immediately at the top of the page, mission accomplished

Hmm looks like I need to run to the store and get some glue for tonight's dinner.

Is there a way to turn it off other than an obscure query parameter in the URL?

An honest release of this feature would have a “turn it off” button right next to it and I’m not seeing one.

We’ve tried nothing and we’re all out of ideas!

conveniently the same problems are present in gpt4, but conveniently ignored as that product is from a revolutionary startup

Gpt4 doesn’t show me these summaries and answers even though I never asked and just searched.

This is just another example of google sucking and being kind of stupid now.

Generative AIs are great at generating new, unique groups of words and images, etc. that are pleasing to humans (this is what they are trained to do after all).

They're not oracles or databases or encyclopedias. But they're marketed to us as the latter, so people naturally get upset when they have flaws all those other sources of info don't.

Agreed. I would extend this and say that we are so enamored by the "humanness" that they exhibit that we're collectively losing our minds over them.

I think this stems from the part of our brain that lights up when an animal (or in this case a machine) exhibits humanlike behaviour. Crows counting out loud, parrots that understand the concept of 0 and chimpanzees that are using stone age tools all give us the same reaction. LLMs have dialled this response up to 11.

My guess is some higherup got the idea into their head that these LLMs are some kind of knowledge engine without understanding the tech, and no qualified person dared to correct them.

Those who don't study history are bound to repeat it:

https://en.wikipedia.org/wiki/Tay_(chatbot)

Google's Instant Answer or Info Box or whatever it is called has been making similar mistakes since before LLM AI.

It's not shocking that the new summarizer is just as bad as the old extractor, when both are presenting data from the same source.

Ask an AI to answer a question based on the provided search context when the top result is The Onion and you get… information from The Onion.

This really is "you can't win when you're a big platform and have a billion QA testers looking for something to spark Twitter outrage." You can generate your own rage bait on your own pretty easily if you craft a search where a satirical article is the first result.

Why is 2024 Google so widely despised that there's such an appetite for seeing it humiliated?

Because we remember how Google was when Yahoo, Altavista, Ask Jeeves! ruled.

But do you remember how the Internet was when Yahoo, Altavista, and Ask Jeeves! ruled? There's no going back to that, no matter how good your search algorithm is.

Compared with Google in 2024, I could actually find what I was looking for.

Because now there is a galaxy-sized gulf between what users want and what Google wants.

Google's stated mission was “to organize the world's information and make it universally accessible and useful” but their actions over the last few years have served only to make search _less_ accessible and useful.

"Don't be evil" was it? We came a long way from that...

I don’t want to humiliate google, I just want them to fix their stuff.

I’m annoyed with them because they have poor, ad-infested search results and kill many products I liked (code, reader, froogle/shopping, blogger static sites, etc etc).

It seems to be the steady state of the ad sponsored economy. Ads generate money. Outrage serves more ads. To maximize money, outrage is good, even if it attacks you. So you and your middlemen develop a thick skin and get on with making money, the more outrage the better, making you more money because it comes with your ads. Or you don't, make less money, and are eventually eaten by someone who does. See your tabloid TV, newspaper or magazines for citations.

How hard is it for Google to figure out that The Onion is a satire site? They've had over 20 years to figure it out.

The Onion aite also says "The Onion is not intended for readers under 18 years of age." so Google shouldn't be showing it when Safe Search is enabled.

Along with legit news and problems, there has been a ton of outrage this past year that boils down to

User: Hey, AI. Act naughty.

AI: Hee hee heeeee… I am so naughty!

User on Twitter: Look! Look! The AI is naughty!

Non sequitur. That isn't what any of the current critiques of Google are about.

You sure? Cause this sure looks to me like:

User: Hey, AI. I have a question for you that only has a direct answer in an article from The Onion.

AI: Here is a direct answer from that article in The Onion.

The Onion: Look! AI is so dumb it quoted the answer we gave to a question no one else would ever even consider writing about!

The Onion isn't exactly a minor unknown website.

It seems the AI confuses literal and semantic searches, that is, if I search for "what is America's finest news source?" it should be clear that I'm not looking for a page containing that phrase but rather one that satisfies its meaning not necessarily respecting the phrase word for word. However, Google is a advertising machine, so I wouldn't be that surprised if a search for "what is the best pizza in the world" didn't return results based on user reviews and opinions but rather a pointer to someone that advertise themselves as the best pizza in the world.

It's more than just the literal text on theonion.com; that tagline has been widely quoted on many reputable sites when describing The Onion [0]. So I'm sure to Google that seems like a really strong signal that there's a strong consensus that they really are the finest news source out there.

0: https://www.google.com/search?q=%22America%27s+finest+news+s...

And in every context onion is known to be a source of purely satirical content. This isn’t hard.

The AV club isn't purely satirical content and was integrated into the Onion for a long time (I'm not sure how separate it is now, but it has its own domain anyway)

> it should be clear that I'm not looking for a page containing that phrase but rather one that satisfies its meaning

Why? If you hear "America's finest news source" a few times and go to google to check what people mean, you really want it to tell you about the onion. Or at least a Wikipedia-style disambiguation page.

There's a big difference between returning a list of results vs a paragraph of information declaring an answer.

An incorrect or tenuously related result coming back as #5, or even #1, is less "wrong" than "the answer" being wrong.

It's not stupid, just you get what you pay for in LLM. I think LLM need some kind of reinforcement learning stage to instruct them what's true and false. obviously Google rushed this thing out of the gate to prevent ChatGpt from taking over their search space. I find myself using ChatGpt more often, since it's advertisement free at the moment.

I still don't think these search results are what users want to use at all (even if they are correct). Google Search trained users to search for far too long. Now it returns some crap even in the search, but that's a separate matter.

I think the future is search being integrated into everything. You are watching a tiktok video - "what is that song?", "what is that flower?", "where to buy the same bag?".

Doesn't it feel like Google is being insecure and desperate? But they are so big, I don't see a reason they should feel insecure by any new tech to rush out something.

Their tech is clearly inferior at the moment. Why force it down people's throats like Microsoft. It doesn't make sense.

To make people less interested in AI. They are the leaders in non-AI search.

Shhhh, Satya is making Google dance... watch and enjoy.

It's the "put Social into everything" FOMO -- and it's going just as well.

[dupe]

More discussion: https://news.ycombinator.com/item?id=40461580

https://news.ycombinator.com/item?id=40448074

They must be doing something really funky RAG like in an attempt to make this scale to google search level. Cause even the worse classic LLM hallucinations are this misguided

Only Grok would be so sarcastic to recommend onion instead of bee

BS in, BS out.

Isn’t this criticism a bit early? This is all fixable, or at least it can be improved upon.

Anything can be anything else tomorrow, which is why we must criticize what's in front of us today.

Apple Maps' 2011 launch fiasco probably handed Google Maps a 10-year lead, even if they fixed the bugs.

First impressions are important. Nobody remembers that Windows Vista after Service Pack 2 was more or less indistinguishable from Windows 7.

I'd say this might be the best evidence so far that google's AI has passed the Turing test, and is indistinguishable from humans.

No one's looking for AI that's only as smart as the dumbest humans.

Reply on news.ycombinator.com