Very useful info there, thank you. Have a beer.

Thanks Liam! I can live without AI hallucinations, let me know if/when it becomes reliable.

Making "literal search" the default is also very useful. There's something to be said for search trying to be "helpful", or at least highlighting where a particular search term was not found. But, again, I can live without that (or open Google Home if I want to open the sewer gates).

> let me know if/when it becomes reliable.

It already is, and has been for about a year. You are good to go.

I believe the OP meant if/when AI would become reliable.

The answer is, obviously, not a chance in hell/never ever.

> I believe the OP meant if/when AI would become reliable.

Oh! How silly of me.

I believe there is a prototype running on a Nordic datacentre which is showing great promise, if they can just get the cooling sorted out. But when the temperatures get low enough, below about 270 kelvin, it should work reliably. I am not allowed to publish the name but it's near 63°26'22"N 10°54'29"E.

Interesting indeed ... I searched for Robert DeNiro with and without &udm=14 to see the difference:

https://www.google.com/search?q=who+is+robert+deniro

https://www.google.com/search?q=who+is+robert+deniro&udm=14

... it is quite substantial.

I see that out of the search result categories (under the search bar) of "All", "Images", "News", "Videos", "Forums", "Short videos", "Web", and "Finance" it picks (sets) "Web" which I guess eliminates the AI fluff. Setting one's browser default to [1]udm= 14 (Web) sounds cool then imho, as you can always then click on another category if you want to see other types of results ... (not stuck)

[1] https://medium.com/@tanyongsheng0805/every-google-udm-in-the-world-6ee9741434c9

You could have, instead, a single interface that despatches queries to search engines (~ federated search), and have some gentle poking from EU to make companies behave, and have users score how well the engines are doing, and have that information publicly visible.

The struggle to control the default has not worked the way intended, all browsers I used recently allow you to modify the engine, but default dominates population usage.

Instead, mandate default to at least 3-5, and make it mandatory that the respective engines are from multiple sovereignties, and pay non-negligible tax in at least one.

The internet is, in the end, national infrastructure, and that doesn't stop at the networks themselves, just like a road network needs enforced policies that do no harm to the users, instead of a free for all maximize greed at all costs.

There is a whole generation of people who never experienced good search. Few more years and all this talk about "enshittification" will become a conspiracy theory.

Probably major search engines should be nationalised and then given to a non-profit or university to run.

After all this is now an utility like water and energy. Having this in private hands is an idiocy (see UK water or rail debacle).

Like a lot of EU ideas this one sounds like a good one right now. Once the various multilingual committees have added their six-pennyworth we will have a totally unworkable scheme that will cost loads of Euros and will, therefore, be silently dropped. Ditto the EU chip scheme which recently lost Intel's interest and no other EU company has stepped into the breach have they?

You cannot just replace Google overnight as they are too far ahead of any other search engine (sorry Bing). When people use the web they want the URL of their answer. Google is now accepted as a verb; no one says "shall I bing for it" do they?

So what you're saying is "you can't beat crap so don't try". Is that right?

And the downside of the udm14 approach, effective though it is at the moment, is that you're just asking Google nicely not to return AI results. If enough people use it, Google will either remove the option, or repurpose it to return the results they want you to see. No matter what any politicians try and do, companies like Google will always find a loophole to work around it, it's just a giant game of whac-a-mole.

If you can't find the browser.urlbar.update2.engineAliasRefresh setting, just make sure boolean is selected and hit add to create it. You will need the whole text though, not just engineAlias.

I think we need to rethink what a search engine is for. At least a quarter of my "searches" are actually questions that I use an AI for instead. The rest - it's a nuisance getting decent results. I don't want adverts most of the time. But the Google AI responses are for the most part useless.

I'm seeing Bing's results about as good as Google's now. Not because it's getting better (it isn't), but because Google is getting worse.

> let me know if/when it becomes reliable.

Good gods. *Horrified expression*

I am AI abstainer, a refusenik. I don't use it for anything at all ever, and I do not intend to. There is nothing here I want.

https://www.bbc.co.uk/news/articles/c15q5qzdjqxo

If you're happy with it, then this method is probably not for you.

It's a lot more reliable and quicker than googling. And it really does save time. Just like with googling, you need to discard the wrong answers. So you need to have some subject matter knowledge or you'll get nowhere.

However it is getting better [AI]. I find more and more that it can save a considerable amount of time doing what I'd call "drudgery" on the internet. Or, it saves time proofreading a document not just for grammar errors but for inconsistencies.

AI is a genuinely useful tool that overall is a good thing. And no, I didn't write or check this post with AI. My grammar checker is suggesting that I remove the word "really" from the second sentence and that I put a comma after However in the second paragraph. I will ignore it.

The EU's web index looks suspiciously like a pre-search censorship filter. And if it wasn't designed as one, that will still be how it is used.

Plus, we don't want regional limits on search unless we choose them in an options menu. The net should not be viewed as national infrastructure but as international infrastructure by default.

Distributed search is actually a lot more flexible and much more useful that was previously mentioned as it dodges all of the blocks and censorship that website based search engines are subject to.

The search extensions are useful, but by now we really should be able to choose one or more contextualisations for our searches. There are loads more options search engines could offer, particularly combining with metadata on web pages. There is no reason why we cannot be searching dynamically generated web pages. We should have persistent search options too. The search engnes we have remain rudimentary. Very much v.1.0.

I've done this in Firefox for my Google searches to include "&num=100" however this only works for the initial search. Any refinements to the search go back to the default settings.

I mostly use [1]SearXNG nowadays (I used to use SearX).

I run my own instance in a Docker at home but often just use [2]https://priv.au/ directly. I have other instances set up as browser search engine options as well, in case some are down.

[1] https://searx.space/

[2] https://priv.au/

Funny, with piHole and Ublock I got better results using google.com than your modified search. google.com got me international results while the modified search got me only local results.

Could it be because I've searched for some specific terms rather than asking a question?

Oh, and no AI bullshit in either.

Thanks for the pointer to Kagi. (Disclaimer: I wrote the Newton client for its previous incarnation.)

$5 a month!!!!! For starter!

I pay about £40 a year for email, which supports a custom domain and wildcard capture etc, that is 1000 times more valuable to me than search.

Not paying this much for search. Laughable. Should be a dollar, with a query cap

Following brief testing, search results appear to be more helpful.

The Wikipedia page mentions "back of the book" or A to Z indexing (though oddly, to me a least, not KWIC) and metadata indexing.

I never really thought about that aspect of search engines.

The EU idea is I think that by making a distributed (clustered) index available through some reasonable Index query API the search engine ecosystem will diversify with more domain specific engines being developed.

Some of the techniques used in training LLMs might be profitably reused to generate indices that retain more context or even generate categorical metadata from the content.

All search engines but especially Google are producing so much codswallop to the point that even a query that obviously demonstrates an intent to purchase something drops you into a collation of irrelevant and frequently unrelated nonsense topped off with the rancid cherry of Googles AI results.

Much worse and even the great unwashed will stop using it. I reckon Polloi would be a fantastic name for a decent engine and instead of Googling something one might hoi it.. :)

..the heavy shift towards corporate search manipulation means that even without AI slop, we already seem to have lost the will to produce independent, useful content or services. It's just not worth creating a blog or an independent shop on the internet today, since only global-scale productions are visible to users.

I'm not sure that the EU is in a position to reverse that tide...