notabot

Type cd then press Alt+. and it’ll copy the last argument of the previous command, make sure it’s what you want, and press Enter.

Indeed. Whilst I very much hope that people are turning against using LLMs, that’s not what this chart shows. As you allude to, it’s just the sentiment of reviews mentioning AI, of which a large number are probably complaining about the results, rather than expressing opposition to the use of LLMs.

Yeah, right, I’m sure that’s exactly what you’d do with them.

Notices your username

Ah, you might actually do it…

Regardless of your status at parties, I actually prefer your wording, it seems to scan a little more easily.

The problem is that Höcke probably looks at that and thinks “It was a good start”. People like that feel no shame for this sort of thing, and want more of it. It could be a powerful message to others though.

That’s a brilliant series that had me chuckling along to chemical formulas, he has a fantastic way with words.

If you enjoy this topic, I can highly recommend Ignition! An informal history of liquid rocket propellants by John D. Clark. He an excellent writer, and makes the whole subject wryly humourus as well as understandable to the lay person.

If you’re tired of dealing with them, you don’t even need to express empathy as the gp post so eloquently described, just don’t throw it back in their faces if they show signs of breaking away, even a little bit. If you don’t have the energy to respond to them and try to expand that break, maybe someone else does, but if you are hostile you’ll just drive them back into the darkness and make them that much harder to ever pull away.

None of this means you have to welcome them with open arms, or even forget what they’ve done. Just give them a space to be less awful than they were. As the gp post said, thinking strategically we need to break away as many supporters of maga as possible, but that doesn’t mean they immediately, or automatically, become allies.

I should say that I’m referring to publicly traded companies here, private companies have different incentives.

Directors, including the CEO, don’t have a mandate to increase shareholder value, no, but under UK, and as far as I can see, US law, there are many rules around things like “actions prejudicial to a minor shareholder” and what is actually in the company’s interests, that the shareholders can, and do, sue both the directors and the companies for actions that do not increase shareholder value, especially as companies often hold their own stock on their balance sheets. They can do this because of the way both statutory and case law has formed. If statutory law, at least, made it that companies had a duty of care to their staff, could not lay off or reduce payments to more than some small percentage per year, and had to pay the costs of remidiating their environmental and societal damage as they inflicted it, these cases would largely vanish, and directors would be incentivised to optimise for those things instead.

Basically, if you make getting rich as a shareholder or CEO easier to do without screwing people over than by screwing them, CEOs and their boards will be less willing to screw everyone over in their pursuit of wealth.

This is something I think people forget too often. When you build and tune a system to incentivise anti-social behaviour, don’t be surprised when you get anti-social people doing anti-social things. We currently legally demand that management always maximise shareholder value, so that’s what they do, and that’s the sort of person who makes it to higher levels of management. If we changed that goal, the outcome would change too. Unfortunately the lawmakers who would have to make those changes are too invested in the current system for it to happen easily.

“Running the gauntlet” was a military corporal punishment that involved running between two rows of you comrades while they hit you with sticks. Basically it now just refers to overcoming a public trial, such as an obstacle course like this.

Maybe just for a feature wall, or a nice border?

To make little charms or pendants, you can use “polymer metal clay” (aka PMC). It look like normal clay, it’s really easy to work with whatever tools you have, and small pieces can be fired on the stovetop or with a small blowtorch. As you do that, it hardens into the metal you chose. You can get anything from copper though silver and gold.

It’s not wildly expensive if you just want to make a few small unique pieces for your hoard, it’s quite forgiving to work with, and it’s really satisfying seeing your creation emerging from the flames.

I’ve heard that sometimes monuments are a bit like flock cameras, and they just fall over for no obvious reason. It’s a real shame, but there you go.

If you don’t mind using openscad with it, it actually works quite well. 😁

Not eating glass is wOkE.

If “different typos” averaged out to “nearly no typos” the same logic would have different words average out to nearly no words.

No, because typos are irregular, so combining multiple sources does not reinforce them, whereas “words” (tokens would be a better term, because they’re not always full words) tend to be used in similar ways, reinforcing those patterns. As you say, context is relevant, an LLM isn’t just looking at the last token to decide the next, but at a much larger window. That does allow it to adjust to tone, as the probabilities of certain tokens, and so words, will depend on that tone, and the type of words used, and thus context, of a conversation. If emojis are used a lot in certain contexts, those patterns will tend to be reinforced in their training, and so produced more in their output.

As to filtering their input, at no point did I say they ingest “the entire internet”, so quoting it seems rather disingenuous. They scrape as much text as they can get, both online, and by OCRing books, as we’ve seen with the recent upset about the number they destroy. What the commercial models do with this afterwards is uncertain, as anything they say is likely to be misleading for commercial purposes. I think it’s a fair assumption that they want good quality data, however they define that, but filtering it all manually is obviously much too vast a project to do entirely manually, so it’s done heuristically, which has the obvious problem that it’ll let through low quality sources some of the time, lowering the quality of the overall data set. You only need to read the anodyne screed they produce to see how all of the little quirks and nuance that marks human communication tends to get left out, leaving LLM prose feeling rather vacuous and repetitive.

The employees of the AI companies will lose their jobs.

Spend the bailout money helping them instead. Maybe even make some sort of society wide safety net, so people don’t need to fear losing their jobs through no fault of their own. Let the companies sink or swim on their own merits.

The fact that they “average” their inputs is why there are comparatively few typos (different sources have different typos, so they average out), not too many emojis or internet lingo (again, different sources use different ones in different places, so they average away), and why they produce such tedious stock output (it’s an average of the inputs, so all the little quirks and idioms that make human communucation more vibrant have been blended away).

I’m sure there is some filtering on the inputs to try to remove the worst of it, but ultimately it’s still just taking the rest and building it’s probability tables from that, which leads to the homogenised outputs we see.

Mind you, having said there are fewer typos, the last time I bothered trying to get one to write some code, it managed to misspell a popular library name in multiple places, which gives some indication of how bad the inputs are, how bad the tokeniser is, or possibly both.

It is a horse made of knifes, in the shape of a knife, that wields knives. It has difficulty getting through screening at airports.