generatingfail

[generatingfail]

Can't happen soon enough.

AI industry groups are urging an appeals court to block what they say is the largest copyright class action ever certified. They’ve warned that a single lawsuit raised by three authors over Anthropic’s AI training now threatens to “financially ruin” the entire AI industry if up to 7 million claimants end up joining the litigation and forcing a settlement.

well…darn

[generatingfail]

An infuriating example of non-generative AI fail.

Today in technology/AI injustices, I tried to set up face ID for an elderly, blind, mostly-tech-illiterate person.

They have a smartphone and a very limited range of voice commands they can use with it. They want to use an app for medical appointments and prescriptions, they know this should be possible and it would be great if it worked for them (you try keeping track of appointments with multiple services up to a year in advance when they all send you letters you can't read and texts you can't open), but the app would require signing in each time. They can't type a password, they can't navigate the steps of multifactor verification.

But! The app allows signing in with face ID. They may not be able to see the phone screen and they didn't know the phone has a front-facing camera at all, but they could hold up a rectangle in front of them until it works, right? This could be so helpful to them. They were keen to try.

So I started the setup process for face ID in their phone, holding up the camera in front of them trying to make it focus, moving it around a bit, all of that. Made sure the settings were maximally accepting in case they didn't always look at the camera. Aaaaand... nothing. The camera stubbornly blurred them out whenever their face was in the frame, like covering them with frosted glass.

Okay, so it's not working, fair enough. What makes it an injustice is that if my decades-younger face slipped into shot briefly, the camera instantly locked onto me, from any angle and lighting. Crisp and clear as day. It's not that the camera was damaged or that their face wasn't in frame properly, it's that wrinkles were not recognised as a face, as human.

Someone who needs face ID as an accessibility option is completely cut off from it, because they don't look like faces the computer was trained on. I don't know if they'll be able to learn to access this app independently at all.

It makes me sad, and angry.

[generatingfail]

I deeply regret the several levels on which this IS how some people are reacting to chatbots.

Actually WWII was a hallucination that just got way out of control.

Check out the bonus panel on the site!

SMBC ◆ PATREON ◆ INSTAGRAM ◆ BLUESKY ◆ STORE

Buy this comic as a print!

Get your copy of A City on Mars signed in person in Charlottesville, VA on August 23rd!

[generatingfail]

If this is genuine, it explains... more than it doesn't.

[generatingfail]

ChatGPT 5 still can't count letters in words

How many "b"s in "blueberry?" Kieran Healy asks the brand-new sooper-dooper souped-up GPT-5.

You can guess what its answer wasn't.

[generatingfail]

Maybe next time LISTEN to the people who totally saw this coming?

Delicious. We love to see it.

[generatingfail]

Well done, all.

Oh my god we actually did it.

[generatingfail]

I have a guess about this. It is ONLY A GUESS. It would take real research to determine whether my guess is correct.

But I think a large part of the problem here is likely to boil down to frequent task switching. Humans are quite bad at this actually; we lose focus and have to remind ourselves where we are and what to do every time we switch to a different task. It's notable that "flow states," known to be highly productive, involve little to no task-switching.

As I understand vibe coding workflows, they involve a lot of switching between writing code, reading code, and prompting the chatbot. I can easily believe that's a speed-killer, and that coders wouldn't realize that it is -- we tend not to notice the effort of reestablishing focus.

I’m unaware of any other such studies done and I’d love to see more. This aligns with my experience of using AI developer tools though for sure

[generatingfail]

Gonna take over the world any moment now, AI is.

i put a message in my work group chat because im illnessing. and my coworker sent me the AI gen summary he received

[generatingfail]

[generatingfail]

Vibe coding considered dangerous

The "command line" is a text-based computer interaction modality. Pretty much anything you can do to your computer with menus and windows and buttons and textboxes, you can do on the command line. (And more, if you have the knowings.)

Surprise, surprise, letting a chatbot muck around on the command line is really dangerous, especially if you don't know much about command-lining such that you can't catch the dangerous stuff!

Source.

[generatingfail]

Cats are distracting!

Pose a chatbot a math problem. Add some random cat facts. Stir.

Watch the AI fail at math.

News source.

Original paper. This is a PREPRINT; it has not yet been peer-reviewed.

[generatingfail]

Can't argue. Wouldn't if I could.

Obligatory AI hate post bc I intend to be even more annoying about this

[generatingfail]

You have to admit it's funny

[generatingfail]

Give 'em hell, Italy!

anyway in case you don’t know it yet

[generatingfail]

Popularity is not the same as accuracy.

"you can use ai to improve spelling and grammar"

[generatingfail]

A group of Wikipedia editors have formed WikiProject AI Cleanup, “a collaboration to combat the increasing problem of unsourced, poorly-written AI-generated content on Wikipedia.” The group’s goal is to protect one of the world’s largest repositories of information from the same kind of misleading AI-generated information that has plagued Google search results, books sold on Amazon, and academic journals. “A few of us had noticed the prevalence of unnatural writing that showed clear signs of being AI-generated, and we managed to replicate similar ‘styles’ using ChatGPT,” Ilyas Lebleu, a founding member of WikiProject AI Cleanup, told me in an email. “Discovering some common AI catchphrases allowed us to quickly spot some of the most egregious examples of generated articles, which we quickly wanted to formalize into an organized project to compile our findings and techniques.”

9 October 2024