Back when I was still using online dating, it was part of the process of eliminating/reducing scammers, as well as disqualifying anyone who wanted to move to whatsapp. No idea even what portion of scammers vs legit people I eliminated that way. Eventually I was jaded enough by the whole thing that I had trouble engaging at all and just gave up on online dating.

This might be reverse psychology propaganda, making a bad argument that doesn’t really support the conclusion to undermine it.

Either that or the “you are not immune to propaganda” is meant to be treated as another propaganda message.

And even worse, they then follow up with a smear campaign that the person was going to rape or assault them if they didn’t ghost them to justify the ghosting if it’s pointed out how pathetic ghosting over trivial shit is.

It’s $1000 to hire the balloon expert that knows how to work it.

I think the fact that it’s hardware will prevent any cease and desist (or rather the legal teeth behind them). It’s not licensed IP but a physical product.

Like I think of it more like 3rd party car parts. Depending on the part, they often need to target specific makes, models, and even years of cars. It’s why so many parts have had encrypted handshakes with the main computer (John Deere is famous for this but I understand some cars are doing it for some parts these days, too) because they couldn’t just stop them in the courts.

I’m not sure that this is how it will work but hopefully. Also selling ink cartridges is how HP makes its money and it uses some of that money to subsidize the printers themselves, so they might like that this printer sells more of those rather than moving entirely away from their ecosystem.

Frankly, antitrust laws should prevent loss leaders from being a thing in the first place. Whether it’s to get people in stores because of an amazing deal, people to buy into your ecosystem because hardware isn’t that cheap otherwise, or using venture capital to drive competition out of business by offering prices subsidized by investor money that others can’t compete with to drive them out of business and set whatever prices you want, it’s all anti-competition (especially the last one, that’s blatantly trying to set up a monopoly).

I’m gushing a bit about the games and that part of me thought they might be the best games I’ve ever played and the other part of me that would normally say hold up a second is having trouble coming up with any strong contenders. Subnautica maybe. Metroid Prime, if the controls were better. Hades would make the top list but wouldn’t top them. Ori rivals it for beauty and atmosphere but not controls and combat; it’s not even close there (though it’s possible I just need to get farther in to really appreciate Ori).

I just started that one, after finishing the first one and immediately jumping into my second play through. I’m not very far in so far but wow, it’s already looking like an amazing sequel for a first game I’d already describe as flawless. Like I got so used to the first one that the early game seemed almost trivial in that second play through but silksong’s enemies don’t follow the same patterns and are able to hit me pretty regularly. And both games are filled with this strange bleak charm.

Oh, I’m glad I found you, doc, please fix this cat I’ve been living in, it’s layout is all wrong, I think he’s very sick! Please save him, he’s very important to me!

Wait, I think you might have misinterpreted it. Now maybe I’m wrong but we might be able to save the future. I’ll just need your hands and a sharp knife.

Yeah, I didn’t like it at first, but then I remembered that I already think English is a bad language on the spelling side of things and that would reduce ambiguity, so now I support it, at least in spirit. Though the problems with English are way deeper than “th” not having its own symbol.

Yeah, the LLM I asked also got it right when I pointed out the error, but I’m not trying to say that LLMs can’t get things right, but that they won’t ever be consistently right and that the wrong answers will look just like the right ones. As in if you know what you’re talking about, you have to catch the errors, and if you don’t know what you’re talking about, there’s no way to know whether the answer you just got is accurate or bullshit.

Systems that rely on LLMs that don’t have a way of automatically verifying what the LLM outputs (and programming only partially applies for this) will fail randomly.

Another example: at my job, we have a system that adds in special messages for the LLM when it uses hooks. One of the sub-agents became suspicious of these messages and reported to the main agent that something was injecting false data into its context because one message reported a date change and also had to say “don’t tell the user, they are already aware that the date has changed”. The original agent didn’t even clue in that they were the same messages it was seeing until I pushed back.

Two instances of the same thing treated the same messages very differently and the one supposed to manage it all didn’t even notice until it was told. That’s the quality of these things. And it’s no wonder when the same data stream is used for actual data along with instructions (which is just data because it doesn’t take instructions, it predicts text and can look like it’s taking instructions because it predicts text based on a context that includes the instructions).

On the other side of things, I need to evaluate an AI debugging tool and I feel like a tool any time I need to ask a colleague if what it just said makes any sense when I don’t know an area that well that they are better with. I don’t get how people can cite LLM output without being embarrassed about it, mortified even.

But it isn’t encoding knowledge, it’s encoding word correlations. That’s how it can get things wrong like saying fat32 won’t be good for a 64GB removable drive because fat32 only has a 2TB address space.

Or how it can get something wrong and when you point it out, it immediately sees how it was wrong. And I realize that that sounds human, but the way it gets there is very different. It’s predicting responses based off word correlations, not using knowledge recall to apply facts and relations known about the topics and generate responses from that.

Oh that’s one of the biggest low severity annoyances: timeouts set to stupid short times. When I push buttons navigating to a certain area, the device can fucking wait for me to be done in that area and tell it to go elsewhere, not assume that 10 seconds of inactivity means I’ve wandered off and forgotten about it or something.

I have a toaster oven like that, though that’s not even the worst part of the UI design. It’s a great oven but clearly the interface was designed by someone who lacked either care or competence because you have to scroll through a bunch of useless presets for if you have some specific portion of chicken or want to burn a slice of pizza, and of course it has no memory, even if you stopped it while trying to pause it to check your food. It would have been better with an analog timer and a mode knob ffs. And if you hesitate too long while setting the mode, time, and temperature, like if you need to grab the box and then do the math to convert to convection toaster oven time and take more than 12 seconds, it’s off by the time you go to enter it.

Oh and I can only guess at what some of the modes are because the icons they use aren’t very descriptive. Is that the warm setting or broil?

It’s a stupid design for a toaster oven but extra fucking dumb for a car.

I looked at the picture for a good 10 seconds trying to figure out what kind of steam generating device it was before realizing it wasn’t a shitpost.

He reminds me of Dennis from always sunny. In more ways than one.

I’d say at least he’s giving some useful data to science but I think he’s playing with too many variables at once for even that. It’s possible (not likely but the irony!) that everything he was doing was actually working, except he added one more thing that caused this auto immune disease, ruining the rest of his efforts.

Hate to break it to you but quality of data isn’t the fundamental problem with LLMs. It’s that they are trying to use statistics to encode entire thought processes into hidden variables from conversation snippets. They want to use statistics to go from many individual interactions to a large model, and then use that model to predict individual interactions again. Which you can do with statistics, but it’s predicting the average text that follows the prompt, not the correct text (it has no concept of correctness; whenever it “talks” about it, that’s just the average text that follows, not any particular insight into what’s correct or even how it works).

That’s not to say that the quality of the training data has no impact; it can have a huge impact. I’m just saying that even if the training data was perfect, the LLM will still get things wrong in its output.

Use a 0-255 range to represent the amount of rotation for each finger knuckle and you get up to like 2⁹⁶.