repost
· · to read

The assertion that LLMs are “capable of surprisingly sophisticated reasoning” is supported with a link an article @willknight wrote on the “Sparks of AGI” paper + criticism of it.

Extruding synthetic text is not reasoning. If the extruded text looks like something sensible, it is because we have made sense of it. I find it dismaying that even critical journalists like @willknight feel a need to repeat these tropes.

—@emilymbender

repost
· · to read

It’s the anniversary of The Moon landing! (Apollo 11 specifically)

Did you know that NASA still has scientists studying the rocks that were collected on The Moon?

It’s because they set some aside to wait for technology to catch up with our curiosity.

We are still learning about The Moon to this day thanks to all the people that worked on the Apollo missions.

—@OkieSpaceQueen

repost
· · to read

This is a credible proposal for DRM for websites in general. It would enable unbeatable adblock-blocking. It would prevent user customization for not just convenience but also accessibility.

I do not say this lightly: Enabling the forfeiture of control over the browsing experience is a fundamentally evil idea that must be rejected now, as it has been in the past, and we must remain vigilant against its reemergence in the future.

https://github.com/RupertBenWiser/Web-Environment-Integrity/blob/main/explainer.md

—@gsderp

repost
· · to read

I really, really want to be pro-AI and pro the many positive things that AI can do for people and the world at large. But the executives deploying these systems are making it increasingly difficult to take this view because the execs keep making terrible unforced errors one after another.

—@lauren

repost
· · to read

I just boosted toots from @pluralistic about PE and, in particular, new proposed merger guidelines from the FCC. If you’re an American, I encourage you to read Cory’s thread, then check out the proposed guidelines and *post a public comment in support*. This kind of stuff is a prime target for astroturfing.

https://www.ftc.gov/news-events/news/press-releases/2023/07/ftc-doj-seek-comment-draft-merger-guidelines

—@specwill

repost
· · to read

You know, it’s not just the “new” AI image recognition stuff that gets things wrong. Remember the old VoiceOver Recognition of like a year or two ago? Yeah, that gets things wrong too. We shouldn’t be surprised that AI gets things wrong. The only reason why abled people are so shocked about it is because they’ve not had to use it for the past 30 or more years. You know what OCR is? Yep, AI. Or, AL as some OCR programs call it. Why? Cause it got it wrong. Yeah, it’s a different kind of wrong, but AL does not equal AI. And yes, screen readers know the difference.

Now, some may say that a letter here or a symbol there doesn’t make OCR bad. But imagine having to use that on a PDF for work purposes. Yeah, you may want it to be a bit more precise in those contexts. You want it to always get numbers correct, for example. 500 is a bit more than 300, or 100, for example.

Now, we blind people have used OCR for a long time. It has improved majorly since even the first KNFB Reader phone. Yeah, actual hardware. That was better than the reading machine, a huge device just for the purpose. So when we see AI make stuff up, or get things wrong, and we’re used to it. It is what it is, and we cope just fine. We either accept the info as junk, like some OCR results, or ask someone else who knows better, like a sighted person that can look at the OCR’d text. Oh, that word Canada is actually the Canadian flag? Yeah.

So when sighted people are all in deep, bright, red rage about OpenAL (see that?) and AI, I kinda am like, “Yeah welcome to our world.” We get incorrect or incomplete info all the time. You think our canes tell us about that cat sitting up there in a tree to the left? Or that wasp just waiting to sting something? Haha no. We have to adjust to things *all, the, time.* And these models are just getting started, like OCR in the 80’s. They *will* get better. But yeah welcome to AI, abled people.

#accessibility #AI #blind

—@devinprater

repost
· · to read

I mean, LLMs would make perfect CEOs

– Confident
– Persuasive
– Well read
– Never admits a mistake, and will degrade you if you point one out to it
– Can make up reasonable sounding bullshit on the spot to explain their every action
– Relies on exploiting the unpaid labor of thousands of people
– Lives in a reality they invent for themselves
– Fundamentally uncreative

—@malcircuit

repost
· · to read

Today’s piece with The Big Picture writer Amelia Mavis Christnot is a must-read… “Autism doesn’t need a cure. It needs empathy and understanding.

I think if we approached ASD from that perspective we could better help individuals who are nonverbal or who have coexisting disabilities.

And mental illness is not an excuse for racism or mass shootings.

There are millions of us with actual diagnoses and we’ve never gone on a bigoted rant or shot up a school, church or movie theater. Stop blaming poor choices on mental illness.”

https://thinkbigpicture.substack.com/p/questions-team-takei-amelia-christnot

—@georgetakei