clear scientific question--like absorption times for a given medication or contraindications or whether or not there is a recent study on efficacy for a given drug against a new indication, for instance. But, it does not know what to ignore --the absolute bs theories that are out there, with more "lay" questions as would be the case with vaccine risks or recommendations. I really worry how we will fight that.

I've seen countless screenshots of chatbots giving one wrong answer after another, typically apologizing after having given a wrong answer and assuring the user the new answer is correct, even though it isn't.

The way to fight that? Don't use chatbots. And tell others why they shouldn't use them.

Otherwise, we're heading very quickly for an idiocracy that's easily exploited by authoritarians.

I once entered a set of longitude-latitude coordinates for a location in northern Kenya. It was a very simple query - where are these coordinates located?

The first answer put it in Mali, so I repeated the search. In fact, I repeated it about 20 times. It gave me 15 different results scattered between Senegal and the Gulf of Oman, and none of them was in Kenya.

Many of my colleagues have entered questions related to their own research, and the results sometimes don't pass a straight-face test.

a specific medical/scientific question that I know can be answered by a fairly recent study in JAMA, NEJM, or a pharmaceutical database, some AI is doing a better job retrieving the correct answer than it clearly is doing for less technical questions-- those for which there may be a lot of extraneous crap info posted somewhere.

I beg to differ that my answer was a universal absolutist one. AI is at a stage where I don't trust it for anything. But, I do like to see what level of utility/accuracy it may have achieved in certain settings.

Still - a better option is scholar.google.com. One can filter it by date to make sure the most recent studies come up.

My opinion, anyway.

PERIOD. You make it seem as though my simple observation was somehow a promotion of its use, though I clearly stated I was NOT. Anything but!

I can assure you I have decades of experience with the medical and scientific literature--both reading and publishing myself. I use the many decades-established search engines that are validated and capable of accurate, even international searches.

As to "Google Scholar," well, to each his/her own. As a reviewer of peer-reviewed published literature in the past for a leading journal, I find its accuracy mixed. As a curiosity, I entered "reputation of Google Scholar" in Google's OWN AI just now! Here is what it gave me back:

near-tragic when I see the AI material was word-for-word duplicated, with no credit or attribution, from the detailed sites I've visited.

Read a long piece examining Grot this a.m. It's scary as shit. Garbage in, garbage out.

to steal traffic as well as information from sites created by people who did the real work. They represent the ongoing theft of intellectual property, but even with continual scraping of the internet, they still often get things wrong. (Once upon a time, a few years ago, the most common excuse from AI companies for their chatbots' mistakes was that they didn't have recent or current access to the internet.)

Rape him.
https://share.newsbreak.com/dypmpr9p

"AI is just fancy autocomplete with a search engine behind it"

I can't remember who said it, but I'll try to find it and update this.

JFC
https://www.theatlantic.com/technology/archive/2025/07/grok-anti-semitic-tweets/683463/

is a feature of that AI program, not a bug.

As terrible as that is, it speedran straight into mocking, gloating awfulness about killing Jews.

to “bots are good at facts”?

And returns results that are infinitely better

What could possibly go wrong?

legislation.