One reason to worry about this is the possibility that google is using tracking info to decide what response it should serve.

That makes sense for some queries, but not for a scientific question with a consensus answer.

For example, if google decides based on browsing that someone is anti-vaccine, and the person asks how best to protect their kids from measles, how will it respond?

Will you get medical consensus, or a "study" that hasn't been around long enough for serious review?

@mcnees I tried from an incognito window just now and got the bullshit answer.

#enshittification

@mcnees It's been in steady but accelerating decline for years, the trouble has always been finding something better. I failed for years to find something with comparable search result quality, and wasn't impressed with anything.

I've been using a free trial of Kagi, a paid search service, as a secondary search engine anytime Google disappoints me, and I think I'm going to subscribe to their lowest tier next month when my trial expires.

I had forgotten what good search looks like.

@mcnees @lisamelton For comparison, DDG is giving me a "recent news" section at the top with 3 thumbnail-links to popular science articles about the new paper, but otherwise the main results list is all links to credible sites showing the [currently-accepted] correct answer in the quoted snippet. There is no simple answer given directly on the page.

that "recent news" section can also be turned off, though I haven’t.

@mcnees I doubt that Google is doing it intentionally, given that Google has repeatedly told me that Medicaid and Medicare are synonyms (possibly because of CMS's full name), and that hyper and hypo are also synonyms.

A software engineer who codes a system to give those results to someone like me?

@mcnees I think the problem is that AI models are rewarded solely with approval, which spirals into telling us what we want to hear, not what is correct.

And this leads to the problem that language model apps have no way to test, to verify. They are not connected to sensors, not even in a secondary sense. All they can do is make statements and see which answers get voted up or down.

@mcnees I was recently Googling stories on a sexual assault court case where the accused has name suppression.

There was a Google question along the lines of "Who is the accused person?"

Curious, I clicked on it.

There was a name in bold.

I didn't know who the person was.

So I Googled them.

Turns out it was the name of the journalist who had written a bunch of stories on the case, which Google had pulled.

Great for their future job prospects if potential employers Google them...

@mcnees It's why Google has fallen in terms of usefulness for ppl like me, who use it to find answers about tangible things rather than use it to find answers about pop culture.

The results always skew towards whatever is in the zeitgeist -- regardless of whether they actually fit the query. It often takes a lot of massaging of queries these days to get the actual answers you want.

And then there's the times where it says there's no results, but then I twiddle the query and oh look, there are!

@mcnees

IMHO the problem is less what Google does here but more that many people still think that chat bots can _reliably_ answer questions or give reliable information.

They can not.

They give answers/information that appear to be coherent/plausible but the process through which the system generates answers effectively cannot be verified.

And since the process is not a simple search for specific terms in a giant database of text but much more complex it - by default - shouldn't be trusted.

@mcnees I've been more and more aware lately of how useless Google is as a search engine, but it's so damn hard to change the habit of it being the default. 😛

Doesn't help that it's on all my smart speakers... Now I'm wondering how accurate Google Assistant's answers are. 🙁

@mcnees

2.200.000.000 results? So one in four humans has created a theory of the age of the universe.