kg

It's impossible for a victim to enforce this penalty, even if they hire a lawyer, in my experience.

In general the art community - not just limited to pornographers - seems to have stuck by X despite all the AI stuff. When I see people posting art it's almost always an X link, even for artists who have accounts on bsky or instagram or whatever.

Try clicking 'switch to website mode' on the left side

Post you're replying to: "motorized vehicles" You: "cyclists"

I don't get it. Can you explain why humans on bicycles are relevant to a discussion of motorized robots? Are you talking specifically about e-bike users scooting along on the sidewalk at 40mph or something?

SDF techniques are widely used in real games, even if a shadertoy put together in an hour or two doesn't run well on low spec hardware.

IIRC modern Apple devices integrate the memory into the whole SoC instead of making it separate on the board and replaceable. It's definitely not swappable like a DIMM or CAMM module would be. Can't find a photo of a decapped M4 chip to prove it, though...

AI can probably solve "how do we get good shaper implementations into software", provided a good enough spec and test suite are available, but it won't solve "how do we convince stakeholders to value supporting languages spoken by massive numbers of people", most likely

This presumes that people will have the time and the patience to wade through the slop and find the gems. Right now people do that with the tide of low quality human-authored games to find the gems but when there's 10x or 100x as many low quality games will people still have the patience? I hope so, but I don't know. We're already seeing a huge uptick in the number of games being released every year on Steam and most of them don't get more than a handful of reviews, positive or negative.

The little pull-quotes marked by illustrations feel chosen at random and not particularly worthy of being so, i.e.

> characteristic times of electronic signals are restricted by so-called parasitic capacitances, and parasitic capacitances in general are proportional to the length of the connection

> modern CPUs use so-called “dynamic branch prediction”

These two mostly served to distract me and break up the reading flow of the page, and I initially thought they had been erroneously truncated.

This next one is straight-up nonsensical because it's missing context:

> Overall, these effects seem to compensate each other, and main memory access latencies of an x64 desktop box, and an M4 Apple SoC happen to be in the same ballpark of about 200-300 CPU cycles.

Which effects?

If the author is reading this, I would advise you to either remove these, or try to be more careful about what sentences you decide to pull out in this way to make sure that they communicate something useful by themselves. Ideally something that reflects the overall theme of the surrounding paragraphs.

The illustrations attached also seem to have nothing to do with the text but that's probably not a big deal.

A reduced test case means you run less code to process the test case, which means your breakpoints trigger less frequently (and the remaining breakpoint triggers are more likely to be relevant to the actual bug). It also means all your debugging steps are likely to run faster and produce less data to sort through. Your log files will be shorter and easier to read/grep, etc.

Imagine being handed a sheet of 10 equations and being told "1 of these equations is wrong." Now imagine that someone came in and erased 8 of the correct equations - they just saved you a bunch of time.

Who will adjudicate private disputes between citizens if not the state?

The behavior described is what OneDrive does

sqlite is really fast. I'm surprised it's only a million.

YMMV but out of the people I know, the lefties are less likely to be hardcore pro-gun-control than the people who lean center-liberal, and this has been true for a long time. John Brown Gun Club, etc.

An appliance repair company I used exactly once maybe 5 or 6 years ago recently started spamming me with texts and emails trying to get me to refer friends to use them or use them again. Never hit the "report spam" button faster.

I'd kinda understand it if they had sent me a polite text or email shortly after our initial engagement saying "hey, if you had a good experience please review us/recommend us" but coming in literal years late with a blast of multiple messages screams "we hired some sort of marketing firm and fed them our customer database".

Somehow I never knew that you can have multiple push URLs for a single remote. Thank you for sharing this, I've been manually pushing to two remotes with a script for years!

Yeah.

https://en.wikipedia.org/wiki/Cell_phone_novel

I personally did a couple of my novel's editing passes on my phone (none of the writing though.)

Think about the scenarios where people are viewing slideshows. If you're on a mobile device, that 150ms spent decoding each image is time where the CPU and/or GPU of the mobile device are running at full tilt, draining the battery. Suddenly applications that would normally be fast and efficient like a photo gallery app become laggy and drain your battery. Not great.

If DLSS 5 becomes the norm it's possible that just makes things worse. The DLSS 5 demos required an entire separate card to run the model, though IIRC NVIDIA did claim it would eventually work on a single card. Given what the model is doing (yassifying the whole scene instead of just upscaling/reconstructing) it makes sense to me that it would increase compute demand instead of reduce it like previous versions of DLSS.

In practice I've had system file checks find and repair errors on multiple different machines. These setups aren't using ECC RAM or RAIDed storage, though.

I wonder how much of it comes down to inconvenience. I've been bringing my own thermos for a year or so now and it took me some time to get used to the ritual of scrubbing it with soap and a bottle brush after use every day - at first I found it inconvenient enough that it made me want to just get paper cups.

Now it's a nice little ritual and I am used to the advantages of bringing my own insulated container with a lid - I can carry it back with groceries or do other errands without worrying about my drink spilling or getting cold.

FWIW I do occasionally see other people at my local coffee shop show up with their own mugs, but I agree that it's quite rare.

I remember early during the Wasm design process we were openly speculating whether people would eventually figure out how to do stuff like Rowhammer from raw Javascript. My bet was on Yes, but I don't remember if any of my teammates said No. I think we all knew the writing was on the wall by then.

Doesn't crowd sourcing and upvotes and revenue for high ranking just mean people will generate what's popular to get paid for it? Will there be money for unpopular truths somehow?

Your deity of choice can most likely survive being mocked with no harm sustained whatsoever, being a supernatural entity beyond human comprehension. He/she/it might get angry with you, which is a good reason to personally choose not to mock supernatural entities or religious figures.

On the other hand, antisemitic slurs or other attacks on oppressed groups can result in real emotional and physical harms against real living people, your neighbors, your coworkers, your friends, even your family.

Try going to chrome://gpu and looking for red text or error messages

At least in Microsoft Teams transcription and recording are separate options.

The lack of a dup opcode in Wasm as mentioned in the post is quite annoying when trying to generate compact code. I wish something like it had made it into the spec.

The email from your boss and the email from a sender masquerading as your boss are both coming through the same channel in the same format with the same presentation, which is why the attack works. Unless you were both faceblind and bad at recognizing voices, the same attack wouldn't work in-person, you'd know the attacker wasn't your boss. Many defense mechanisms used in corporate email environments are built around making sure the email from your boss looks meaningfully different in order to establish that data vs instruction separation. (There are social engineering attacks that would work in-person though, but I don't think it's right to equate those to LLM attacks.)

Prompt injection is just exploiting the lack of separation, it's not 'coercion' or 'convincing'. Though you could argue that things like jailbreaking are closer to coercion, I'm not convinced that a statistical token predictor can be coerced to do anything.

Lots of people I know run models in "yolo" mode or the equivalent as well, which means it could just invoke curl or telnet to exfiltrate data.

Incidentally, many countries/states are working hard to make it impossible to transition that early.