Google is overhauling its dominant search engine with an “AI mode” that will no longer provide links to other websites, in a major shift expected to cause turmoil across the web.
The tech giant will launch the feature in the UK from Tuesday. Instead of showing links to websites, the AI mode generates its own answers using information from around the web.
The update, seen as a landmark moment for the web, is likely to lead to more turbulence for websites that have already seen huge drops in traffic from Google as the company pushes AI-driven answers into its search results.
The truly interesting bits of research are when you can dig down far enough into the details to overturn the conventional wisdom. And of course that’s what AI is, the on balance conventional view of everyone currently on the web.
Hey, I once got Britannica to change their entry on scandium. Conventional wisdom can be wrong….
It works if there’s a narrative you want to push, and Google is 100% about the narrative. It has been next to useless as a search engine for years if you’re using it to find anything that the Blob doesn’t want you to find. This will just tie that knot tighter.
yeah the AI mode *with* traceable links is the clever and valuable bit.
Interested,
“It works if there’s a narrative you want to push, and Google is 100% about the narrative. It has been next to useless as a search engine for years if you’re using it to find anything that the Blob doesn’t want you to find. ”
It’s not so much “the blob” as what Google prioritises. It used to be links, but it’s now also about things like speed, or recency. If a politician has something happen this week, all you get are pages and pages of news shit, all saying the same thing when you search for them. Some little blog page gets lowered because it was written 6 months ago.
I often search Google and just realise that “god this is shit” and remember where I should go. For weird questions, I search Reddit and then possibly pop into a subreddit to ask around. For general knowledge, Wikipedia. And for a lot of things now I like specific services with complex search, like booking.com or OpenTable. I want a Chinese restaurant in Northampton that is open Monday lunchtime, OpenTable does a better job of servicing that request than Google does.
But Reddit’s great. I don’t know when it got so good or why, but it’s where the old forums went. Where you can go and post the PC specification you’re thinking of building for gaming and you get insider knowledge “no mate, don’t get that graphics card, it’s got more RAM but they throttled the GPU so you’ll barely notice a difference”. Or people telling American tourists about Avebury instead of Stonehenge. Or whether Ikea mattresses are any good. It’s like how the internet used to be.
Tim makes an interesting and valid point. With conventional search, pages that are far from the consensus are returned but not on the first page.
An AI search typically has fewer citations and therefore this information may be hidden.
I suspect that with the correct prompt AI searches could be encouraged to return less ideas further from the consensus. I see that ChatGPT has a ‘research model. Hmm. Lots to think about.
There’s also the problem of the incentives of it. Publishers producing good content to get clicks
https://timsthoughts.substack.com/p/so-why-should-publishers-help-google
Google AI last week: it told me that ns&i Index-Linked Savings Certificates began in 2011.
I knew that was tosh i.e. a lie, a “hallucination”.
With a different prompt it told me that they started for OAPs only in 1975 and were opened to the general public in 1981. Thirty years earlier! That sounds about right.
So I must stop using google: what do people recommend instead?
Put otherwise, if I wanted lies I’d jut consult Wokeypedia. I don’t need them from google too.
Gamecock uses Yahoo search. It’s not wonderful, but I don’t trust Google.
Google using sites info without credit seems wrong to me.
It’s rare I do a search and stop after reading the first item on the first page of responses. A single ‘AI’ response from Google is grossly unlikely to give you what you are looking for.
Cirrusly, you search for something and they report that they have a fucking million results. You read pages in and find nothing but useless politically correct orthodoxy.
If you label your junk ‘AI,’ it’s acceptable?
“It couldn’t be wrong; it’s AI!”
Try DuckDuckGo, works for me. Google fills the first page with ads and dodgy sites that (for example) charge you £50 for renewing your driving licence at age 70, which is free if you go directly to the DVLA.
@WB – Reddit also is good for spicy material that doesn’t require age verification (or so I’m led to believe).
In also use DDG but increasingly ask questions of Perplexity, and then scan the sources it’s used. Far more productive than conventional search.
@ Interested
You’ve got me worried! Last week my wife used one of those genealogy websites (to reassure herself she’d got the family tree of her late mother’s first cousin right before the annual family reunion) and wandered into checking mine and told me I had another cousin John beside the one who died 40-ish years ago: “er, no” – so she gave me his date and place of birth: “er, no that was Frederick Keith”: “this is a genealogy website: it must be right”: “that is data I checked myself”.
So I googled myself to see if there was any reference to this other John who was allegedly still alive and found that Google AI thinks that I am two people – both of whom are partial descriptions of myself – and the third John it recognised is my late great-grandfather, whom it admits to being dead.
Does someone want to push the narrative that I am two people? If so, why?
[Mrs77 later managed to establish the genealogy website was wrong]
I can’t remember when I stopped using Google but I remember getting fed up of all the advertising and general clutter and having to search through the results and links to find what I’d asked for. I switched to DDG in the Brave browser and only if that failed did I try Google.
Now I either used Brave’s built in Ai for quick simple searches or if its a bit more complicated and I want an answer or summary of something I use ChatGPT and Perplexity and as above check their sources in the links provided.
If I want to know how to do something I go straight to YouTube search.
@john77
I doubt it’s anything to worry about – just as dearieme says an hallucination.
This is an amusing read on the same subject:
https://davebarry.substack.com/p/death-by-ai
For all of those saying use DuckDuckGo (which I do), and Perplexity etc, yep, but we’re not the problem.
The very fact that we’re contributing here marks us out as contrarians and renegades and reasonably intelligent sceptics to one degree or another; the issue is the great mass of people who will look at the first result they get and assume it’s correct. This is how they rewrite history, and, as we know, he who controls the past controls the future.
I have confirmed that AI gives utterly wrong answers about literary texts.
For a laugh, I submitted part of a poem that I had written when I was a spotty teenager. (I am now 79)
It then confidently informed me that these were actually the famous opening lines to Milton’s poem, Paradise Lost.
In reality, there were only two ordinary words there that it had in common with his work.
“Google AI thinks that I am two people”: our local world-famous NHS hospital used to think I was (at least) three people, varied according to misspelling of surname and hallucinatory date of birth.
“Does someone want to push the narrative that I am two people? If so, why?”
Gamecock finds it useful sometimes to blame his [nonexistent] twin brother.
Interested: It’s like the NewLabour crowd’s government-by-headline. The devil is in the detail, but they know most people don’t care.
“If I want to know how to do something I go straight to YouTube search.”
I repaired my old ZX Spectrum because I watch YouTube. It wasn’t even a case of looking for how to do it, simply years of watching videos of people do that kind of thing giving me the confidence to have a bash at it myself. I have absolutely no doubt that if it had gone on the blink ten, twenty years ago, I’d just have set it aside as something to see about having repaired at some point.
YouTube might have its issues when it comes to commentary on current affairs, but it (almost) completely redeems itself with stuff like that.
Yes, bless Youtube for all the “how to” videos.
Gamecock hardly looks in boxes for instructions anymore, expecting better help from Youtube.
Ironically the better quality of search that AI produces the less clicks the information providers get so the less incentive they have to post useful content. It could become a downward spiral giving a fantastic way of searching nothing of interest.
If AI is so smart, why can’t it predict what I’m going to search for, and go ahead and post it up before I have to ask?
@ Gamecock
That is what Putin, Xi and Murphy would like.
Fortunately the guys (and the occasional gal) who are smart enough to design AI are – ipso facto – too smart to believe in totalitarian ideologies’ myths.