> What scares me about this new AI mode thingy
What scares me is the rampant inaccuracy. In my experience, the AI responses are wrong about 65% of the time. I just did a search today about an error talking about a disconnected link between apps, and Google AI result summary told me that the error was related to my pulling a USB drive too quickly in windows. The ONLY word similar to my query and that AI response was the word "disconnect". Everything else was clearly about the SaaS apps.
I have people coming to me, asking me questions, then telling my Google told them something else, so now I have to waste time convincing them that it's wrong. Over the past 2 years AI has done nothing for me but complicate my work life.
And of course, this could be because the model is crap, but it could be because they want me to keep refining my query over and over for more ad views. Either way, it's a terrible experience.
Yep. For years we've been telling people to 'just fucking google it', and now when they do they're getting bullshit AI answers.
Worst thing is, some of these bullshit answers will be medical, some of them financial, it seems pretty certain people are being harmed.
Yeah the Google AI results are more dangerous than ChatGPT, not only because it uses a smaller model but because Google's knowledge graph used to deliver very accurate and authoritative information but now that's been replaced by a stochastic system in the same place, so people are used to trusting it.
I think we’re getting what we deserve by snarkily telling people to Google stuff instead of answering accurately. Google results have never ever been pure accuracy
To be fair - for all of those years Google has been serving up some atrocious results - remember when googling health symptoms got you rabies or pregnancy.
There's even the meme where people ask if the code was the result of a stack overflow question, or answer
It seems to me one needs to consider the complexity of the question they are asking before searching it.
To stick with your post, consider people asking medical or financial questions. For a wide variety of reasons, many of such questions don't have an answer. In such cases, AI is still going to take a crack at it. AI shouldn't be blamed for "bullshit answers" to such questions.
Before using AI, I think people should stop and ask themselves, "Is there really a single answer to this question? Is AI the right choice?"
The problem is Google's AI results get even simple factual questions wrong all the time.
Earlier today, I searched "pixel 10 wifi 7" because I was confused that GSMArena showed my Pixel 8 supports Wifi 7, but the Pixel 10 only Wifi 6. Gemini confidently claimed that the Pixel 10 does support Wifi 7 -- but that's not true at all. Only the Pixel 10 _Pro_ supports it, as I discovered when actually reading the non-AI search results.
And this is a question about a Google product!
I had a similar thing when I was gooling a few days ago, I can't remember exactly but it was like "why does [product] not support [feature]" and the AI summary was confidently wrong, saying "The product does support [feature]", which knew was completely incorrect, and I did find a Reddit discussion or something in the actual results with discussions that were actually about what I was looking for!
It's really depressing how bad things are getting...
It’s hilariously persistent in this, esp. for anything even slightly divergent from the beaten path. Discount everything the AI box says about emacs to zero.
Admittedly I’m unsure if it was Google or DuckDuckGo. I switch between both. I quickly asked the in search AI for a UTC time conversion like a lazy fool and it got it off by almost a day wrong.
I avoid any asking any agent a fact-based (especially math) request. It's a great compression algorithm and a great language generator, and I guess the intersection of those two things is "an answer". Calculation doesn't intersect.
My google search for 'pixel 10 wifi 7' immediately shows the right answer. (10 Pro and 10 Pro XL support it but, but base Pixel 10 only supports Wifi 6E).
Though the inconsistency of results between users is definitely another frustrating thing.
[dead]
Ok, fair. Hard to understand why it would get that wrong.
Because LLMs aren't sentient, they don't draw on facts, and they don't have nuance. The answer given is similar to answers you might expect to see for similar questions.
It's really amazing we can make machines do that, and it's really depressing that we think a stochastic bullshit machine is going to give us something we can rely on.
Or… the default LLM Google uses for search has been quantized to s**. Ask a proper Thinking model, with browsing enabled, and odds of a correct answer are much higher. There’s been substantial improvement in AI in even the last year.
Ask a human a question like this, and they also have a chance of getting it wrong, even when confident.
I think that it feels a little wasteful to go to Google search to ask a question like this, only for the AI that's giving you an answer instead of page results to perform its own web search to get you the response.
Also, I asked a thinking model with browsing enabled and got this:
> The Google Pixel 10 is expected to support Wi-Fi 7 (802.11be), based on the Qualcomm Snapdragon 8 Gen 4 / Tensor G5 chipset it will likely use, which includes an integrated Wi-Fi 7 modem. Specific finalized specs aren't confirmed until Google's official announcement.
(Model GLM-5-Turbo - two months old - using Kilo Code in the "Ask" profile; in its thinking token churn it reasoned that it should keep the response brief and direct. Perhaps not the best suite of model+harness for this task, but it's what I had to hand that's not quantized to shit, is a thinking model, and has a web search tool available to it.)
> Ask a human a question like this
Why would a human know specs for a random phone off the top of their head? The human response is either "I don't know" or "let me look that up", not a hallucination.
> Ask a human a question like this, and they also have a chance of getting it wrong, even when confident.
We google something specifically because the humans within reach don't know. The goal of searching is, well, to search pages - we're trying to find a site when we use google search.
The goal when using an LLM is generally different; we want an answer, not a site.
LLMs are not a site. They are a clever person that can point you to sites. They, like humans, are fallible.
*so long as an accurate answer exists on the internet
Claude is OK at saying when it can’t find good information, but it’s still 50/50 on citing a source that has nothing to do with its claim.
They are this wrong about everything, but you don't usually notice it when using it to look for things you aren't an expert in. The default stance really does need to be "do not trust, verify" at all times.
They can still be useful, e.g. they're significantly better at finding "I want a thing that does x but not y and it must be blue, or maybe two things that can be glued together to do that" than classic search. But they'll routinely miss extremely obvious answers because the related search it ran didn't find it, or completely screw up what something can actually do. Checking more pages of results by hand or asking humans who know even a little about those fields is still wildly more useful... but they're absolutely slaughtering the sites where people do that, by stealing all the real traffic and sending DDoS-level automated requests.
How can you say they are wrong about "everything"?
I built a retro game clone once and I used that project as a way to try out AI. While it wasn't perfect, it definitely wasn't wrong about everything. I'd go so far as to say it was probably correct (or damn close) 75% of the time.
I see people on HN all the time saying AI is terrible, but that just isn't the experience I'm having. I'm willing to admit it may have something to do with me not being able to recognize I'm being fed bullshit. Or, I may be asking really simple questions. Who knows? But AI seems like a pretty useful tool for average people.
I’d make assumptions about how the cheapest and fastest possible flash model optimized for being extra cheap and extra fast would get something wrong based on its limited context (which can be very incomplete summaries of search results)
I often have the expensive models give relatively simple inaccurate answers, even when they cite sources that directly contradict them. The error rate is lower, but you can’t have confidence with llm answers.
It somehow seems to interpret whatever sources it's grepping as the exact opposite of what those sources say fairly often. I've lost track of how many times I've clicked on the sources it cites, and every single one is in agreement, but the AI claims the opposite.
Did you just agree to a stranger's counterpoint on the internet? This post should be in a museum somewhere
The simple answer is that these systems are very bad at telling the truth reliably.
When the default "search" results are AI, it's difficult, if not impossible, to "choose", since Google is pushing the AI so hard.
In watching the demo, I didn't come away with the impression that they were removing search results. Yes, they are pushing AI hard, but users can still opt to use Google in the more traditional way. Unless I misunderstood the demo, it's definitely possible to choose.
"possible to choose" doesn't get us much.
An interesting aspect of this is the decrease in quality feedback on th organic links. If most people never get down to the actual links there is very little to tell which ones were good or if they had any relevance.
There is also that less incentive to properly maintain the search algorithms to fight SEO and spam.
For all intents and purpose, organic search results have been given a death sentence and are just waiting for the last moment.
Organic search dying was my first reaction too. But, who knows...this wouldn't be the first time I've heard that.
They are showing billions of people a big bold answer at the top of all their pages.
What a wildly irresponsible company
Go to Google right now and search anything. What is the very first thing you see?
> one needs to consider the complexity of the question they are asking before searching...consider people asking medical or financial questions...many of such questions don't have an answer. In such cases, AI is still going to take a crack at it. AI shouldn't be blamed for "bullshit answers"...people should stop and ask themselves, "Is there really a single answer to this question?
It's a bold position to say that it's the users fault for being lied to by Google. There isn't a "single answer" to most questions. It's still Google's job to provide answers that are accurate and reflect the best information available on complicated topics. That's what they're trying to sell us anyway. When google's AI can't live up to the hype "You shouldn't be asking AI such difficult questions" is not a great response, especially when people are just trying to get web search results and AI is suddenly interrupting with an opinion nobody asked for.
I asked it “how can I tell if a spray paint can is empty?” And it told me that the paint can would no longer rattle.
In past, people can trust Google. Now we should teach children don't trust "search result" from Google.
It's nice that Google's AI summary always lists its sources. It's less nice that those sources more often than not do not corroborate the summary. It often seems to be a few random links thrown in there for good measure.
I have no idea why this is, but it is impossible that these links are primary sources of the data, if such things even exists at all. In which case, why list them?
It is certainly seems possible that the actual sources of the data is the output of some other LLM.
I’ll bet they intentionally obfuscate so people can’t find the actual sources of info used for the answers
Reminds me of this gem:
https://www.reddit.com/media?url=https%3A%2F%2Fi.redd.it%2Fn...
Straight of of x-files s02e03
With AI ads you get all the power from big data aggregation, the trust/framing of an authoritative voice, and cheap personalization that specifically optimizes for what convinces you. It's too powerful. Even if it only works a small percentage of the time we're interacting with these things so frequently that a small percent is a large number. They're already feeding user profiles into these machines and there's explicit talk about having the LLMs optimize ad campaigns. It's already dystopian if it's ads to get you to spend your money, but people seem to dismiss that. Do we not care that this is also being used in the same way to convince you to believe certain things? To join certain political organizations?
Yeah, these things help me write more lines of code faster (if we include all the lines from our design docs) but I don't like the idea of pointing a supercomputer at my brain and someone else using it to try to manipulate me. That's not a game I'll win. It's not a game you'll win either.
The built-in Search AI is fucking braindead and people constantly come up to me "Google said xyz" and I just have to turn around and say "I do not care what the Google Search AI said".
Whatever it says is a waste of time 99% of the time. Although people believe it, or consider it worthwhile majority of the time because its so simple to use. It's always there, always instant and appears at the very top.
I would much rather people shove a question into a locally running Qwen model and tell me what it said rather than use the nonsense search model. I hate it.
/rant over.
accuracy hasn't been their priority for a while now - they just want people to click on ads
Free AI's are dumb. Extremely dumb. The Google AI result is dumb on purpose -- being smart requires more compute.
Google has been around for a quarter of a century. People are still incredibly dumb and will believe whatever they like.
Can you share the query?
> the AI responses are wrong about 65% of the time
Highly doubtful.
Depends on what you ask. It's pretty easy to get wrong information.
e.g. search for "how do you make money with options"
Google's AI says
"When you buy a Call, you are betting the stock price will go up. When you buy a Put, you are betting it will go down."
Wrong right off the bat, because it ingested a whole bunch of get-rich-quick bull on the internet. The correct version is that if you buy a call you are betting the stock price will go up more than the market expects it to.
I tried this search. It gave a write up about buying and selling options, noting that the price of the stock had to move significantly, not just go up or down. It also talked about vertical spreads and iron condors. It touches on delta, theta, and volatility and their impacts, as well as leverage risk and potential uncapped risk.
While I agree that AI gets things wrong a lot, and someone should read significantly more before getting into actually trading options, this does give a decent overview to give a layperson an idea of what they are, and some key terms on what to look for if they want to dive deeper. That said, with this info alone, there are some sharp edges that would leave the person open to unnecessary risk if they went on this information alone.
They probably update these answers offline. I tried "how do you profit from options" and got:
> Call Options: You buy these when you believe a stock's price will go up. If the stock rises past your strike price, the option's value increases, allowing you to sell it for a profit or exercise it to buy the stock at a discount.
> Put Options: You buy these when you believe a stock's price will go down. If the stock falls below your strike price, you profit.
Which leaves me wondering if changing the search textually busts some cache that they update using a slower/smarter model.
And this is yet another problem, it's stochastic. And often it's self-contradicting even within the same response. What else do you expect from a language model which essentially predicts tokens.
Is that really categorically wrong, or is it a correct-enough explanation for laypeople looking for a one-sentence answer?
It's wild to me that someone looking for advice on how to do any kind of stock trading would be looking for a once sentence answer.
I hope it at least has real citations to actual websites like, I dunno, fidelity or some other reasonably competent authority that can explain all the details?
It's an answer that's too short for an expert to find useful, and useless to a layperson unless all they want to do is reply to a post on twitter.
I've never searched for a financial question where I did not want to know all the weird details because why would I search for it unless I was considering doing it? Seems like someone who doesn't care about the answer is going to be more an edge case than I am.
Those looking for a one sentence answer will be the quickest to invest. When people talk about the harms of AI, this is the kind of thing that comes to mind first for me.
It is, in fact, categorically wrong, and misleads beginners to make bad decisions. Robinhood is notably bad for promoting this kind of gambling behavior on its platforms; they also state the same misinformation (that you buy a call if you think something is going to go up, when it is in fact a bet that it is going to go up more than a certain amount in a certain period of time).
People shoot themselves in the foot because they think NVDA is going to go up after earnings, buy call options, and then even though the stock goes up they lose money because they did not understand IV crush.
People looking for one-sentence explanations should really not be playing with options. In finance you should understand what you're buying thoroughly. If you just want to bet that "NVDA goes up", you should just buy NVDA stock; that is the trade that accurately captures that bet.
In fact, more then what the call seller expects, not the market.
This is the problem with teaching and learning. Everything is wrong to some extent. I used to be this way but I don't have a better approach.
Newtonian physics is actually wrong, the founding of any country will be wrong, biology is wrong, nutrition is wrong… what can we even teach? what should we teach in this lens? serious question.
The serious answer is in the non-AI-summarized world, you can choose whose information to read and trust.
If you want to learn about finance, you can learn about it from people who actually know what they're talking about. You can choose to listen to Jim Simons or Warren Buffet or whoever actually knows a thing or two instead of the rando dude you met at the bar. The AI summaries, on the other hand, ingested a lot of internet garbage.
I picked finance as an example because anecdotally, most of the information on the internet by pure token volume is wrong. The Youtubers drawing lines on charts want your attention because they make money from page views; the financial advisors want your annual fees; the brokerages want you to gamble and get your commisions or PFOF (in the case of zero-commision brokers); the market makers and HFTs want your spreads; Reddit users want to show off their lucky, statistically insignificant profit charts for karma points. None of the above have an intention to give you good information.
Honestly Google's AI answer is about as right if not more right then your answer.
You can easily make money buying a call without the stock price moving a single cent (IV increases). Funny enough the stock can even go down and with a large enough IV increase you still make money.
It hallucinates greatly about many things when I ask about C++ things. Things that you can easily find the right answer in cppreference or by just inspecting headers in your own IDE.
Yeah it’s not 2023 anymore. So no it’s not hallucinating like you think it is.
95% closer to your expectations?
What, you think it's actually higher too?
If you think ai is getting answers wrong at anything close to the frequency quoted then it calls into question your usage and ability to use ai in general.
That has the be the most hacker news way of saying "skill issue," I have seen to date.