Video of the development process, presumably: https://youtu.be/OPeC7o9iDnI

Are you saying that there are three modes, one called “autopilot system,” one called “Autopilot,” and one called “Full Self-Driving”? Because otherwise this phrasing:

the driving mode in question was Tesla’s “autopilot system,” instead of the proper Autopilot

is confusing in a way that’s independent of the way that Tesla’s names are confusing.

My understanding was that there are two modes, one misleadingly called “autopilot” which is basically adaptive cruise control plus lane keeping, and one even more misleadingly called “full self driving,” which integrates with the GPS and will also do lane changes and exits. Both will happily plow you into a semi at speed, which makes the names REALLY dumb. But the distinction the article is making, between the “autopilot system” and capital-A “Autopilot,” seems like it needs more explanation, and I don’t think that’s a Tesla distinction.

The reporting states the driving mode in question was Tesla’s “autopilot system,” instead of the proper Autopilot, which is Tesla’s less advanced driving system. Full Self-Driving (Supervised) is Tesla’s more capable feature, although its name is a misnomer, since it’s not capable of fully driving itself and requires constant supervision.

I feel like this section needs a rewrite. What is the distinction they’re drawing between “Tesla’s ‘autopilot system’” and “Autopilot”?

Okay, yes, but that’s because they had messed up their application enough that the updater itself couldn’t be updated, which they presumably discovered in the process of trying to remedy his bug. That is, the flaw he found couldn’t actually be exploited only because of a deeper flaw he hadn’t found. (Shades of the Sirius Cybernetics Corporation there, whose deep fundamental design flaws were almost totally hidden by their superficial design flaws.) He still led them to a critical vulnerability that took them months to fix.

But then there’s that whole exchange from World Enough and Time that ends with this:

DOCTOR: Please, listen to me. Look at me. Go on, look at me. That’s good. That’s very, very good. Now, do you see this mad woman sitting in this chair? Her name isn’t Doctor Who. My name is Doctor Who.

NARDOLE: It isn’t, is it?

DOCTOR: I like it.

I know they’re doing a bit, but I think it provides adequate cover for proper geeks to nevertheless use the expression when it’ll provide shorthand clarity.

It’s because in the US pixel phones are one of the only ways to get a phone with an unlocked bootloader. Or, in other words, to actually own a phone.

For anyone else who was gonna check: yes, it’s the fifth circuit.

Thanks for the reminder! I had switched to Lawnchair, but I just realized I had forgotten to uninstall Nova. Fixed!

Doesn’t even need to. He’s been frequently citing polls carried out by CPAC, which show him having a 93% approval rating. Why bother with troublesome reality?

I mean, this was my first question as well. If you say “I don’t trust that guy with my kids,” then you also should not be leaving that person alone with your kids. If you do leave him alone with your kids, people aren’t wrong to say that you very much are trusting him with them.

So I think it’s legitimate to ask whether they mean they “don’t trust Chinese and US firms with their data” in the sense that they do not provide their data to them, or just in the sense that they do give them a bunch of data, and then feel misgivings about it.

Which, y’know, it’s something you have a limited degree of control over, certainly, and we don’t want to fall too much into blaming the victims. But as someone who didn’t manage to actually get off Facebook until last year, I definitely felt for a long while before that like I was complicit in my own exploitation, and contributing to a societal problem. I think even the people still there know that cancelling it is the low-hanging fruit in terms of reducing the amount of data in the hands of dubious firms.

I think that might be pleurisy instead.

I think the idea is that this particular drug isn’t suitable as birth control, but having identified that this mechanism/biological pathway can work for birth control, they can look for a less toxic compound to achieve the same effect.

I dunno if that’s always the case. I still love The Phantom Tollbooth.

On the other hand, I remember being really frustrated by a phrase from another book. (I think it was “Kneeknock Rise”? I remember exactly nothing else about this book, though, so it might not be that.) It was a description of a scene, and it said the dog was asleep, “arms and legs akimbo.” Now, I was in… maybe third or fourth grade, so I had never encountered the word “akimbo” before, and asked my parents what it meant. They explained that “arms akimbo” was basically the only phrase in which it’s used, and it means having your arms out to your sides with your elbows bent and your hands on your hips. But this just confused me further, because the book said “arms and legs akimbo.” I had no idea what it was trying to describe, and could not picture it. I tried to draw a picture of what it seemed to be describing, and continued to find it baffling. My parents agreed that was odd, and suggested I talk to my teacher about it. The teacher was very dismissive, though, saying “well, obviously you’ve never had a dog, or you’d know exactly what they’re talking about.” Which…what? Why would you even say that to a curious kid? Couldn’t you at least draw a doodle of what it looks like?

So yeah, being forced to stick with a book you don’t like does leave a very strong negative impression.

Extreme solopsist?

That would be the short story Harrison Bergeron by Kurt Vonnegut.

Yeah, this is what I was going to call out. Calling it “100% solvable by humans” and saying “if human scores were included, they would be at 100%” when 20-60% of humans solved each task seems kinda misleading. The AI scores are so low that I don’t think this kind of hyperbole is necessary; I assume there are some humans that scored 100%, but I would find it a lot more useful if they said something like “the worst-performing human in our sample was able to solve 45% of the tasks” or whatever. Given that the AIs are still scoring below 1%, that’s still pretty dark.

I built one, but it was a (top-down) resin printer, so I don’t think I have any useful advice to offer. Also, I kinda burnt out on the project after I came in one morning and found the resin had melted the vat and leaked onto the floor–somehow I just totally failed to consider that the solvents in the resin could obviously dissolve a lot of plastics. Huge mess, huge pain to remediate. At least it was a concrete floor. I got a replacement vat in glass, but I never worked up the will to work on it again. It did basically print before that, but I had a lot of off-layer curing and didn’t get to do much work on tuning that in before the accident.

username checks out

Yeah, agreed. I must have misunderstood your original comment.

I’m not sure I totally understand your comment, so bear with me if I’m agreeing with you and just not understanding that.

“let me prioritize PRs raised by humans” … but why? Why do that in the first place? If bots/LLMs/agents/GenAI genuinely worked they would not care if it was made or not by humans, it would just be quality submission to share.

Before LLMs, there was a kind of symmetry about pull requests. You could tell at a glance how much effort someone had put into creating the PR. High effort didn’t guarantee that the PR was high quality, but you could be sure you wouldn’t have to review a huge number of worthless PRs simply because the work required to make something that even looked plausibly decent was too much for it to be worth doing unless you were serious about the project.

Now, however, that’s changed. Anyone can create something that looks, at first glance, like it might be an actual bug fix, feature implementation, etc. just by having the LLM spit something out. It’s like the old adage about arguing online–the effort required to refute bullshit is exponentially higher than the effort required to generate it. So now you don’t need to be serious about advancing a project to create a plausible-looking PR. And that means that you can get PRs coming from people who are just trolls, people who have no interest in the project but just want to improve their ranking on github so they look better to potential employers, people who build competing closed-source projects and want to waste the time of the developers of open-source alternatives, people who want to sneak subtle backdoors into various projects (this was always a risk but used to require an unusual degree of resources, and now anyone can spam attempts to a bunch of projects), etc. And there’s no obvious way to tell all these things apart; you just have to do a code review, and that’s extremely labor-intensive.

So yeah, even if the LLMs were good enough to produce terrific code when well-guided, you wouldn’t be able to discern exactly what they’d been instructed to make the code do, and it could still be a big problem.