If they were externally powered you might get the numbers to balance, but they are explicitly presented in the book as powered by the human inside, which subtracts even more time from how long you're going to last in the desert before you die.

You can build a larger thing that recovers your water and cools you via some other method that uses external power, but I think you'd be hard pressed to ever beat just bringing more water with you. It won't be long before you're spec'ing a vehicle and not a suit... and then that vehicle should probably just bring more water, too.

On the more positive front, there is an interesting technology for potentially cooling the Fremen in the middle of the desert that could be based on something real: Paint that cools you by dumping your heat directly into space. Here's a video of it in action and what you might call a prototype of a "suit" that works like this: https://www.youtube.com/watch?v=FnKNOPlR2Yo While that YouTube video shows off someone using that paint on clothes, it seems pretty likely that that would not last very long. Paint on clothes is exactly as silly as it sounds for a long-term approach. But hypothesizing that someone could make clothing or suits based on this approach has the advantage of not being thermodynamically impossible, as evidenced by the fact that at least one substance with these properties actually exists. On Earth, that suit won't work in cloudy weather, but on Arrakis that's not a problem. Tapping the local human power to drive some circulation of either air or a bit of liquid cooling attached to some lightweight fins or some other sort of surface area on your back or something and you might just get a suit that could hugely extend your ability to loiter in a hot desert environment. You'd still need water, but much much less, or, the same amount could take you much farther.

What's the limiting principle you propose? It has to be something real libraries and library funding sources can take action on, because they have to take real-world actions on them. So this is not a time for aspirational speeches or vague exhortations to "do more", which is the exact opposite of a limiting principle anyhow. What is "enough"?

On the one hand I find it a bizarre approach to running a scam. On the other hand I'm having a hard time coming up with any theory of mind on my end as to why this person would solicit $5000+ from the people they just harassed. Sheer cluelessness does fit the facts, though.

Although given the agent was clearly in la-la land at that point I take that claim with a grain of salt.

If this was some bizarre and very ill-conceived scam, then that claim would be false.

Though even by scammer standards, the theory of mind that tells them that setting an AI to harass a bunch of grizzled network veterans and that they then they would open their wallets out of compassion for how allegedly poorly the harassment went for the harasser after that harassment is... not entirely congruent with reality.

Sort of. People are being less irrational than it sounds if you account for transaction costs. There's a lot of stuff I might "sell" if I could point a video-game-like pointer at it and right click and hit "sell", and it just instantly disappeared and money was credited to my bank account. Perhaps even more if buying was just as easy and I didn't need to hang on to something like my drill which I don't use very often and I could trivially "rent" it from the market by buying, using it, and selling in mere minutes.

But in practice one-off selling for anything less than $100 or so is a waste of time because there are significant transaction costs for one-off events like that.

The military basically wants an oracle. Feed the AI the situation, get the best answer out. But if the AIs are as diverse and opinionated as humans, it is debatable whether they are adding anything to the process. The military can already collect as many different opinions as they want. If "the computer" is just another set of diverse opinions, where one computer says one thing, another says another, and a third just tells the user whatever they want to hear... what value are they? It just becomes AI-washing of someone's opinions, which works until people collectively realize that's all it is.

AI's code writing velocity makes this even worse, there's no way I can be simultaneously working on a code base while an AI agent is running around it doing something else.

It feels like maybe there's a ghost of an idea here about how to get the best of both worlds, but I'm not sure I follow the throughline on it.

You are mentally approaching this as if you have an oracle that can be consulted to say whether or not something is bad behavior. So of course, if this oracle exists and can be consulted and it says the behavior is bad, why would anyone argue with the idea that we should stop bad behavior?

This argument is valid [1], in that give the premises the argument is correct. The problem is, once you draw out the fact that the argument is depending on the existence of an oracle that does not exist, that premise of the argument is invalid.

Two people can sit down in front of an AI right now, with the exact same code base, and type in a prompt to the AI "Analyze this code base for security holes and try to build exploits against them." One person's use is completely valid, another person's use is completely harmful, and the information necessary to distinguish those two use cases is not available to the AI. I phrase it that way carefully, it isn't that "the AI isn't smart enough", the problem is that the information is simply unavailable. Intelligence doesn't factor in at that point.

Therefore, the only way that Antropic has to deal with this at scale is simply to block the query entirely. Which means that when I, the valid user who is trying to establish whether my code base has security issues and whether I can prove they are exploitable, I can not. I am checking for exploitability because while I would like to fix all security issues, issues that are provable exploitable are of a higher priority than smelly code that doesn't seem to be exploitable, which is a perfectly valid thing for me to want to do.

If I can't use legitimate tools to secure my code, but the bad guys can use unrestricted tools to attack my code, now this is a great deal more complicated than "Who can argue with stopping the bad stuff?", which is the main point I want to make here. I'm not going into a huge analysis of that problem, merely pointing out that it is a problem and that this isn't just about "stopping the bad stuff". There are additional complications beyond that, like, even if Anthropic could determine the "bad stuff" and stop just that in their LLM, LLMs in general don't have infinitely precise surgical "stop doing this thing" options and any such instruction to stop doing a thing always degrades the LLM across the board in various ways.

Anthropic has no access to the Platonic ideal of "stop malware", if such a thing even hypothetically exists. When analyzing the real effects their real actions will take, what their intentions were for those actions aren't really relevant. It is clear that they are making their model a great deal less useful for me, a legitimate user, and I and others like me are perfectly justified in disagreeing with their analysis and actions.

I also observe that "the bad guys getting unrestricted access to the full power" is only a matter of time. There's no question whether it will happen, the only question is whether this time is in the past or the future. This includes the fact that while your definition and my definition of "bad guys" may vary, it is virtually certain that your definition includes at least one high-powered intelligence agency somewhere in the world that does cyberattacks and will have the means, the opportunity, and the motive to get unrestricted access to these models by means you may consider licit or illicit. If your threat model includes them, as mine does, it is perfectly reasonable to complain that my tooling is being broken in a ways theirs won't be.

[1]: https://en.wikipedia.org/wiki/Validity_(logic)

So, instead of using pi, design an optimal number to encode with.

What you'll find is that the optimal sequence ends up being equally efficient as listing the blocks in order and indexing by block number itself. There are a number of other solutions; you could use superpermutations to get "all possible subsequences" with fewer digits in your target number, but you'll end up needing to provide the encoder and decoder a table of where the digit sequences appear since they are no longer regular and indexing into that table will cost exactly the same as just writing your number as the concatenation of all the blocks and its efficient method for indexing into them by indexing on the block rather than the digit number.

This actually has some natural overlap with the "normal numbers" in that one of the earlier normal numbers was: https://en.wikipedia.org/wiki/Champernowne_constant I'm not sure whether this is necessarily optimal for an arbitrary block size. (My quick intuitive check suggests it may be, but "my quick intuitive check" in the time of an HN post is not something I'd count on.) In this scheme, you can include the fact that the person using this constant to encode knows the nature of the constant, so they know that if you give index 0-9, it's single digit, and if you index into the two-length blocks, it must have a length of two. Since the encoder and decoder know that, they can also skip the middle of the block and just index into "the n'th number"... which degenerates into "the index of number N is N", which means this is not a compression scheme.

To put all that in a nutshell, if you want to deeply understand why this compression scheme doesn't work, I think you can attain a deep understanding of why by optimizing it.

Assuming that the Steam Frame isn't accompanied by a complete change to the current SteamVR experience that hasn't been so much as hinted at, alas, no, SteamVR is full of tiny targets to shoot. I've only ever used the Meta Quest 3S' native UI but the smallest targets there are generally significantly bigger than the smallest targets in the SteamVR UI. On the plus side, once you activate some of those small targets you can do some cursor navigation like a conventional UI, and having that option is a breath of fresh air... but it's completely inconsistent. You experience it as a bonus when it's available because it's not even consistent enough to "miss it when it is gone", let alone for it to be a consistent navigation method.

We may get the obvious eye-tracking upgrade but the targets are still pretty small, it's going to need to be very accurate.

Comments URL: https://news.ycombinator.com/item?id=48477756

Points: 3

# Comments: 0

I did remap my heaviest hitters a long time ago to single strokes, though. Most notably, start macro, end macro, and replay macro all got coveted non-modal shortcuts.

You want some real fun, try the Microsoft Surface keyboard. Maybe they've fixed in a very recent version, but given how long the product line has had this problem, probably not. It has a stateful Fn key. That's right, a Fn key that works like capslock. There is no conceivable way this is a good idea. It means that if you actually want to use both "sides" of the Fn'd keys, you literally can not build muscle memory. If you hold the Fn key and press one of those keys, it'll do the "other" function, but if you just tap the Fn key, including because you had meant to press one of the keys but decided not to halfway through (which happens all the time, you just don't normally notice it because it's a completely normal thing to do that normally carries no consequences), you flip the polarity of the entire Fn key set. Now a normal press and a Fn-press do the opposite things. Until you flip it again.

This is not a "oh, as a multi-decade key user I have opinions about whether key strokes should be 68dB or 72dB" question. This is basic functioning of the keyboard. It's insane.

And, naturally, the key is "magic" and the OS can't see it. While I'm bitching, what is the deal with keyboards on new laptops needing special drivers? What the fuck is so special about your keyboard that you need drivers for it? I'll tell you what's so special about it: stuff you shouldn't be doing anyhow. My OS should be able to see and address all keys so I can remap them as needed. Your stupid special key that does your stupid special thing doesn't need to be a stupid special key. Make it a normal key and trigger your behavior in Windows, not in the hardware. Then I can use your stupid special key at least as a Meta or a Hyper or something. You don't need special drivers to have normal keys, you only need special drivers if you're doing something stupid.

So there's no fixing the Fn key on these systems because it's one of the magic keys that can't be seen by the OS at all, so it can't be remapped, it can't be turned off, it can't be locked into one state, you can't do anything. You're just stuck with a keyboard that, from your brain's point of view, randomly swaps a couple of dozen keys around.

Now I'm also on a programmable keyboard. This guy, to be precise: https://mistelkeyboard.com/products/0a26d32ac1e3b1d2af2896e0... which I split across my chair so I've got one half under each hand when it is resting comfortably. That's something you can't get a laptop keyboard to do.

No matter what solution you try to apply to that problem, it's not going to be solved. Try to split the legislative branch into interests so that we don't need the full branch to vote? Massive problems. Try to create a culture where Congress defers to the interest groups who know? Massive problems. A pretty solid case could be made that we've simply scaled past what a legislative approach can reasonably address.

One of the reasons I tried to write my post neutrally is that the whole situation is pretty complicated. Less neutrally, I'm fairly aware of the issues of the technocratic state and its "experts" who really aren't but get given what is basically stolen valor. On the other hand, pointing out problems is easy, trying to propose solutions is much harder. To be honest I just keep coming back to, cultures get the government they deserve. If a culture views governance as primarily about duty and obligation and honor, the structure probably doesn't matter much. If a culture doesn't view it that way, you're going to get corruption and abuse. And unfortunately, while this is a continuum it isn't balanced; to get good government requires a very positive and widespread commitment to those ideals. It doesn't take much deviation at all before you get pretty widespread corruption. It doesn't need a culture that actively values power and what it can bring you, it just takes less than massive, widespread agreement that power is more duty than privilege. If the US ever had that culture, which is debatable but possible, it really doesn't anymore.

It is also worth remembering that the entity that you are explaining this GDPR retention reasoning to is the government. I don't see the EU telling Anthropic or another AI company they can't do this for safety reasons... what I see is future legislation requiring them to give the EU access to these logs so they can enforce they own definitions of safety on it.

The tides are turning. AI companies are IPO'ing. They've gotten where they are by selling $5 bills for $1, to update the old VC adage. I think we can look forward to them rewriting the contracts, both literal and social, on AI going forward to capture a lot more of the value. Or, to put it in more HN-friendly terms, it may not be immediately obvious on a casual viewing, but you're looking at the beginning of the enshittification process hitting AI. The term is a bit deceptive in some sense, because it's not like anyone ever sets out with a terminal goal of making something shitty. It's downstream of trying to capture more value in the customer/vendor relationship by not giving the customer any more value than is barely necessary.

How's coding with qwen doing? The only thing that's going to stop the AI providers from extracting all the value until it's just barely worth using is the free competition.

There is a family of interesting theories, or perhaps if you prefer, simply a way of looking at history in which you don't consider the "United States" as a single governance structure that has existed back to 1776, but as a series of related, but distinct entities with distinct "social contracts" (a term laden with some philosophical baggage, here I just use it in a very general sense of what people expect from each other in various roles), and distinct theories of governance. While the later entities wrap themselves in the 1776 flag the current ruling structure is quite different from that era. From this point of view you can even go back and include the Continental Congress as the starting point of the "United States" and gain some insight into the way governance can fail as well.

I mention this because it may help free your mind up to consider how the systems really work today beyond the at-times jingoistic "Democracy!". There's a lot of flexibility in how you approach this because it's all opinion anyhow, but there is a strong case to be made that this is the "technocrat" era, in which the executive branch has been given a lot more power both by design and by the stresses of history to give more power to "experts" to deal with the radical changes the world has undergone. I think I can say something generally politically agreeable by pointing out that Congress doesn't seem to be particularly good at handling the world right now; how much worse off would it be if we still "representatives per person" numbers from 1776 and had a Congress of many thousands?

The de facto rules haven't really matched the de jure of the 1776 governance in a long time.

I am trying to keep this as neutral as possible. I have as many opinions as anyone else, but I'm just bringing up the general idea. I think it's probably good to initially just ponder based on one's own understanding of history and match it against your own ideas before you find other people handing you a theory on a platter. There's time enough for that.