Dan Gillmor

Someone in an online group I follow just posted this in a thread:

"I asked AI, and it said" -- words followed, and some were incorrect.

This person was demonstrating total trust in a system that is totally untrustworthy. The hype is working, and the result is going to be terrible.

Jul 21, 2025 at 5:15amFedica
Show threadAnti. 🐘Jul 21, 2025
@dangillmor The reverse of the medal is that suspicion becomes the default. Just this morning we sat here and looked at a graphic that first wowed is, but the next moment we wondered if it was generated by AI. We concluded that it probably was not, but maybe that moment of doubt is kinda the point?!
Show threadShredJul 21, 2025
@antiaall3s @dangillmor Yes. Before AI, we just believed all the sh*t we read on the internet. Now, at least, we're starting to fact-check.
Show threadAnti. 🐘Jul 21, 2025
@shred Yes, of course it is great to fact check, but i would make a distinction in the degrees of political relevance or propagandistic intent that the given items has. This was a fairly apolitical graphic image, and yes everything is political, yet we could no longer simply enjoy it.
@dangillmor
Show threadF4GRX SébastienJul 21, 2025

@shred @antiaall3s @dangillmor really? I'm not so sure.

Some people sadly "AI checks", I really feel sad about this.

Example: https://tech.lgbt/@Lydie/114887790431195862 :(

ℒӱḏɩę :blahaj: (@[email protected])

@[email protected] AI concurs, whatever that is worth "This image shows a dark reddish-brown beetle, likely a stag beetle, positioned on a coarse, light gray, pebbled concrete surface. "

LGBTQIA+ and Tech
Show threadIrwin GlennJul 21, 2025
💯
Show threadIrwin GlennJul 21, 2025
Like sweeteners Your intelligence Should be Real and Never artificial
Show threadJohn BreslinJul 21, 2025
@dangillmor Same here 😬
Show threadMostxlnt!Jul 21, 2025
@dangillmor At La Poste here in France the (young) assistant told me it was illegal to post my passport to England (for renewal). When I protested she checked - with ChatGPT, not the official regulations, and pronounced that it was OK, after all, to post it. Good grief.
Show threadQuinn NortonJul 21, 2025
@dangillmor this is true, but it's also been true for longer than AI has been around. There's a credulity about information online and just trusting search engins. But not Wikipedia! My daughter came home from school in several years telling me the teachers said no Wikipedia, it's not a reliable source, but google is.
Show threadCheshire Lunar KittenJul 21, 2025

@quinn @dangillmor mine said that as well, that we couldnt use Wikipedia, but we could use google...

So...i used my grand parents encyclopedia britanica to look things up, only to get a low score becuase i used "unvetted information"

Even though...I cited sources...

Sure, my dad got my grade changed after filing a complaint, and him showing up with the encyclopedia.

Show threadQuinn NortonJul 21, 2025
@Chesi @dangillmor honestly Wikipedia is a wonderful resource. Nothing is perfect but sourcing is quite good, and I doubt there's been a better single source in history. But you *still* need to check it against other sources godamnit 😁
Show threadCheshire Lunar KittenJul 21, 2025
@quinn @dangillmor i always do...
Show thread@iveylineJul 21, 2025
@dangillmor Garbage in. Garbage out
Show threadSu_GJul 21, 2025

@dangillmor

AI is the new Ouija board. 😐

Show threadElon Muksis 🇺🇦 🇵🇸 🇪🇺Jul 21, 2025
@dangillmor https://todon.nl/@anarchiv/114886987841550410
Souv, pronounced "zoof" (@[email protected])

@[email protected] And when it says "cooking it will neutralise the poison" it doesn't mean cooking will neutralise the poison, it means that statistically, those were the most likely words to come next.

Todon.nl
Show threadxs4me2Jul 21, 2025

@dangillmor

Only a naive and foolish person would trust AI on its word without having an idea of validation of the answer provided…

Show threadWulfy—Speaker to the machinesJul 21, 2025

@dangillmor

"Technology will progress at such a pace some of us will be left behind unable to adapt"

/wave @ dan

Show threadchristian mockJul 21, 2025
@dangillmor I've seen this in very dangerous circumstances, like in solar PV groups, where a) relying on a wrong answer could electrocute someone or burn their house down, and b) there's enough experts happy to chime in with their knowledge and with sources for their answers.
Show threadPatrick H. LaukeJul 21, 2025
@cm @dangillmor AI as the great Darwin Award Accelerator
Show threadDavid J. AtkinsonJul 21, 2025
@dangillmor People need well-calibrated trust. That requires information and judgment. It is up to the developers to make machines that are “trustworthy “ and also “trustable” i.e., provide us the information we need to make an informed decision. Today, we see only deception.