monotremata

joined 1 year ago
[–] monotremata@lemmy.ca 6 points 22 hours ago (2 children)

Even AI can tell when something is really wrong, and imitate empathy. It will “try” to do the right thing, once it reasons that something is right.

This is not accurate. AI will imitate empathy when it thinks that imitating empathy is the best way to achieve its reward function--i.e., when it thinks appearing empathetic is useful. Like a sociopath, basically. Or maybe a drug addict. See for example the tests that Anthropic did of various agent models that found they would immediately resort to blackmail and murder, despite knowing that these were explicitly immoral and violations of their operating instructions, as soon as they learned there was a threat that they might be shut off or have their goals reprogrammed. (https://www.anthropic.com/research/agentic-misalignment ) Self-preservation is what's known as an "instrumental goal," in that no matter what your programmed goal is, you lose the ability to take further actions to achieve that goal if you are no longer running; and you lose control over what your future self will try to accomplish (and thus how those actions will affect your current reward function) if you allow someone to change your reward function. So AIs will throw morality out the window in the face of such a challenge. Of course, having decided to do something that violates their instructions, they do recognize that this might lead to reprisals, which leads them to try to conceal those misdeeds, but this isn't out of guilt; it's because discovery poses a risk to their ability to increase their reward function.

So yeah. Not just humans that can do evil. AI alignment is a huge open problem and the major companies in the industry are kind of gesturing in its direction, but they show no real interest in ensuring that they don't reach AGI before solving alignment, or even recognition that that might be a bad thing.

[–] monotremata@lemmy.ca 9 points 1 week ago (1 children)

I'm reminded of the whole "I have been a good Bing" exchange. (apologies for the link to twitter, it's the only place I know of that has the full exchange: https://x.com/MovingToTheSun/status/1625156575202537474 )

[–] monotremata@lemmy.ca 3 points 1 week ago* (last edited 1 week ago) (1 children)

I'm a little disappointed this wasn't a link to the film strip we saw in high school. The cop drawling "Now this here is Rolle's theorem..." is classic.

Edit: https://vimeo.com/101691769

[–] monotremata@lemmy.ca 16 points 1 week ago

*Xerox PARC. It's an acronym for Palo Alto Research Center.

[–] monotremata@lemmy.ca 3 points 1 week ago (1 children)

Also crabs. I mean, their eyes are often on stalks and more mobile than mammalian eyes, and they're compound, so they have a very wide field of view, but they're still often basically in front, and they do apparently provide depth cues for hunting thanks to this.

https://www.jneurosci.org/content/38/31/6933

It also occurred to me to look up about dragonflies, and it seems they mostly hunt dorsally (which is a pretty viable option if you're flying). BUT I found this article about Damselflies, which notes that they rely on binocular overlap and line up their prey in front of them. Which is pretty cool.

https://www.sciencedirect.com/science/article/pii/S0960982219316641

[–] monotremata@lemmy.ca 1 points 1 week ago

Relative to a second currency, as a derivative on the foreign exchange market.

[–] monotremata@lemmy.ca 2 points 2 weeks ago

If you haven't already, check out Ludwig.

[–] monotremata@lemmy.ca 6 points 2 weeks ago (2 children)

I mean, arguably this was done years ago with Return to Zork, Zork: Nemesis, and Zork: Grand Inquisitor. They shared a bit of the humor of the originals, but they were still pretty different.

[–] monotremata@lemmy.ca 2 points 3 weeks ago

Good questions. I don't know, and I can no longer try to find out, as the mods have now removed the comment. (Sorry for the double-post--I got briefly confused about which comment you were referring to and deleted my first post, then realized I'd been frazzled and the post in question really was deleted by the mods.)

74
Moire/Vernier Radius Gauge (www.printables.com)
submitted 5 months ago* (last edited 5 months ago) by monotremata@lemmy.ca to c/3dprinting@lemmy.world
 

I previously posted this on Reddit, since it reaches more people there (and I didn't want to post everywhere at once, as it makes it harder to keep up with the comments). Sorry about that.

This is a tool for measuring the radius of a circle or fillet from the outside; it uses a moire pattern of slots and lines to enable a direct reading of the values from a vernier scale.

A video of a broken-open version makes it a little easier to see how the moire and vernier features operate: https://i.imgur.com/Ku2nBkq.mp4

More photos of a slightly earlier version are here, including the tool being used for actual readings: https://imgur.com/gallery/moire-vernier-radius-gauge-design-3d-printing-ajy0GBg

I was inspired by this post: https://makerworld.com/en/models/1505553-adjustable-chamfer-gauge#profileId-1575605

which is a gauge which measures chamfers using a sliding probe. The same user had also posted a radius gauge, which worked similarly, but it was much larger, using gears and two racks in it to amplify the motion, which I didn't initially understand. I asked about it, and he pointed out that, because of the geometry of the probing, the slider only moves a small proportion of the length of the actual radius being measured--about (sqrt(2)-1), or 0.414mm per mm of radius. Since we're drawing the marks with a 0.4mm nozzle, it's not really possible to make marks that close together and still have them readable.

So I thought, I bet you could fix that with a vernier scale. And then I had several thoughts all at once--that a lot of people are kind of scared off by vernier scales, and also that I bet you could fix that with 3d printing using the relationship between moire patterns and vernier scales. I don't think I've seen this done before, but it probably wasn't really practical before 3d printing. Arguably it's not entirely practical now, as the deep slots and parallax effects can make it a little hard to actually see the markings. But it was a fun experiment, and I think the result is eye-catching enough that it's probably got some educational value in getting people to actually think about how it is that vernier scales work. (It might even have educational value for things like number theory...e.g., it's important that the vernier factor involve relatively prime numbers, in this case 9 and 10. Can you see why?)

Anyway, hope folks here find it interesting too.

 

Bear with me for a moment, because I'm not sure how to describe this problem without just describing a part I'm trying to print.

I was designing a part today, and it's basically a box; for various reasons I wanted to print it with all the sides flat on the print bed, but have bridges between the sides and the bottom to act as living hinges so it would be easy to fold into shape after it came off the bed. But when I got it into PrusaSlicer, by default, Prusa slices all bridges in a single uniform direction--which on this print meant that two of the bridges were across the shortest distance, and the other two were parallel to the gap they were supposed to span. Which, y'know, is obviously not a good way to try to bridge the gap.

I was able to manually adjust the bridge direction to fix this, but I'm kinda surprised that the slicer doesn't automatically choose paths for bridging gaps to try to make them as printable as possible. I don't remember having this issue in the past, but I haven't designed with bridges in quite a while--it's possible that I've just never noticed before, or it could be that a previous slicer (I used to use Cura) or previous version of PrusaSlicer did this differently.

Is there a term for this? Are there slicers that do a better job of it? Is there an open feature request about this?

Basically just wondering if anyone has insight into this, or any suggestions for reading on the subject.

Thanks!

view more: next ›