$1.3 million

This seems to be a smoking gun:

Researchers say the most convincing evidence came from measuring how the cell expanded during charging.

When a battery charges, ions move into the anode, causing it to expand. Graphite anodes have a unique expansion pattern because of changes in graphite’s layered structure. The Donut Lab cell showed this exact pattern.

This finding matters because sodium ions are too big to fit into graphite the way lithium ions do. According to investigators, the graphite expansion pattern clearly shows that lithium is the active ion in the battery.

I suspect the claimed “99.98% accuracy” is counting out of all faces scanned, which is a bullshit way to make the tech look good. Most faces are not marked as shoplifters in the database. A system that literally does nothing would probably still have greater than 99% accuracy.

What we really want to know is what percentage of reported matches are accurate, and I bet it isn’t anywhere near 99%.

I licked several of the Switch 1 carts and I can’t taste whatever they’re using on them at all. I feel like I’m missing out!

the obligation to retain the original product logo (Section 7(b));

the denial of any rights to use the copyright holder’s trademarks (Section 7(e)).

Uhhh is it just me or is it impossible to follow the first requirement without violating the second one? The logo requirement seems engineered to make sure that you can’t actually fork the project: if you include the original logo, they can hit you for trademark violation, and if you don’t include the logo they can say you violated their license terms.

“Tubular” is from surfer lingo right? It makes a lot of sense when you realize they’re comparing whatever cool thing you’re talking about to a wave like this: https://www.youtube.com/shorts/1mXQaSA3-rE

I don’t need $1M. I’ll take just the $1000, if the other box has a million in it they can put it towards the next experiment.

Okay, I’m gonna ask: why is reddit not social media?

That makes sense. I figured they were worried that an alternate OS would be more likely to exploit their encryption somehow, but if it’s all using industry standard hardware then it really ought to be open.

According to the linked article it prevents the use of Samsung Pay and access to the Secure Folder (an extra layer of security you can enable that requires a second PIN to be input before you can access certain apps and files). This seems pretty reasonable, the goal is clearly to prevent access to especially sensitive data if someone has stolen the phone.

Serious question: what problems does the default keyboard have that Florisboard is trying to solve? I’ve never looked at the Android keyboard and thought “I wish this worked differently,” except maybe when I switch languages by accident.

I’m still on Pandora and honestly have no idea how they stack up. I just use it as a radio station on long drives.

He absolutely would have sold your data if there were any buyers for it. Microsoft was still a terrible company in the Clippy era, there just were fewer opportunities to be terrible in this particular way at the time.

These lists of red flags make me feel like I must be a replicant. I wrote a comment just like that one, em dash and all, on a different site just the other day, with my own organic brain!

My first instinct was to use an em dash instead of that last comma, but it seemed too on the nose.

I feel like I first saw this pic before AI image generation was any good, so it might just be good old fashioned Photoshop.

What does she mean there was a “generational shift” that led to people burning CDs? Back in the floppy disk days, everyone was copying floppies—I remember when my grandfather bought a Mac to use at home, and immediately his friends at work loaded him up with copied disks. Which generation is she thinking of that wasn’t pirating a ton of software?

deleted by creator

Lying requires intent. Currently popular LLMs build responses one token at a time—when it starts writing a sentence, it doesn’t know how it will end, and therefore can’t have an opinion about the truth value of it. (I’d go further and claim it can’t really “have an opinion” about anything, but even if it can, it can neither lie nor tell the truth on purpose.) It can consider its own output (and therefore potentially have an opinion about whether it is true or false) only after it has been generated, when generating the next token.

“Admitting” that it’s lying only proves that it has been exposed to “admission” as a pattern in its training data.

Recently my friend was trying to get me to apply for a junior dev position. “I don’t have the right skills,” I said. “The biggest project I ever coded was a calculator for my Java final, in college, a decade and a half ago.”

It did not occur to me that showing up without the skills and using a LLM to half ass it was an option!