VoxelVexillologist
🇺🇸 Multidimensional Radical Centrist
No bio...
User ID: 64
VoxelVexillologist
🇺🇸 Multidimensional Radical Centrist
3hr ago
Yeah, I try to be very careful to avoid confirmation bias, and only for relatively minor things. I have to do that fairly frequently with LLMs in other contexts, so it's not an unfamiliar skill to me. A couple extra strength exercises (which make sense to me given my knowledge of the anatomy) isn't likely to hurt anything.
But I doubt most patients are avoiding it. Frankly, I'm probably not completely either.
VoxelVexillologist
🇺🇸 Multidimensional Radical Centrist
5hr ago
I don't think HIV was well-known until well into the 80s, even if it was recorded earlier. There are plausibly a dozen or two documented-but-small-scale diseases today that will blow up in the next decade or two. Mpox wasn't new, but the mutation that was spreading was, AFAIK.
VoxelVexillologist
🇺🇸 Multidimensional Radical Centrist
6hr ago
In my experience, the AI can go an extra step beyond what WebMD could previously. Sure, there was a list of possible causes for your symptoms, but now it'll ask you followup questions to narrow it down, or suggest simple diagnostics ("does it hurt when you...").
VoxelVexillologist
🇺🇸 Multidimensional Radical Centrist
7hr ago
I decided long ago that if I stumbled upon a proof of P=NP that I'd decline to publish it, quit my job, and find a cabin in the wilderness like Uncle Ted (without the explosives). If P!=NP, publish away. But "actually asymmetric cryptography is fundamentally broken" seems like something to really think through the disclosure of, given it's ubiquity. I would not trade that for the ability to optimize travel routes for the sales department.
VoxelVexillologist
🇺🇸 Multidimensional Radical Centrist
7hr ago
A guy said something silly amongst walls of text and 40 hours of video
If you need reams of rigorous argument to make your point, that's fine. Some assertions are hard to make: in math some proofs are just long. But you'd better have rigorous argument, because otherwise you just have a sophisticated gish gallop. Waving your hands and making specious points in the middle of a long argument is a sophomoric approach to get the grad student TA to pass your essay, not a professional one.
VoxelVexillologist
🇺🇸 Multidimensional Radical Centrist
7hr ago
Frustratingly, unless your primary concern is assigning blame (which seems likely to fall on both China and the US, and nobody is practically going to be paying out for this), the final assessment of origin isn't even that meaningful. If your goal is to avoid a similar recurrence, the fact that it seems plausible to have been a lab leak should be sufficient for enacting the same set of rule changes as if we knew it was. We can just make sure the next one is not a lab leak, and it costs us relatively little. For what it's worth, we can try to avoid future zoonosis too.
VoxelVexillologist
🇺🇸 Multidimensional Radical Centrist
7hr ago
I think whether the doctors like it or not, their patients (maybe not the geriatric ones, but often they have younger family helping) will be asking AI questions. I will admit to using it for "does this merit going to the doctor?" Sorts of questions, and for "recommend some PT exercises for this nagging issue" (seems a blurry line between sports medicine and personal trainer). I trust the doctors I see, but copays are high enough that I don't go for just any sniffle, and I don't want to pester them with questions outside of appointments.
From using it in other fields, I'd expect it to be good at transcribing verbal notes (I hear doctors hate charting, maybe it could do that interactively with the patient?), and at pulling obscure facts from the training data. I know pharmacists are currently supposed to be checking drug interactions, but adding an extra layer of verification there seems an easy addition. I'd expect it to be good at triage, but I'm not sure exactly what that would look like.
I would not expect mass layoffs in medicine as a result. Demand for medicine is very elastic, and if appointments were cheaper, we'd probably just have more of them. The limiting factor seems to be the 300 seconds you get with a physician, and how many hours they have in a day.
VoxelVexillologist
🇺🇸 Multidimensional Radical Centrist
8hr ago
AIs can spot things, tiny grammatical issues with quotes or repetition that humans miss.
They are good at this sort of thing, but I find it hilarious that I also have trouble with models flummoxing tool calls (a JSON payload) with unterminated strings and braces. I had it try to fetch about:blank recently.
This isn't free-tier models, this is the large, recent open weight models. But most of the time it does it correctly.
VoxelVexillologist
🇺🇸 Multidimensional Radical Centrist
18hr ago
Lots of math results used to be kept secret, all the way back to the Pythagoreans. We should ask them how they feel about fava beans.
VoxelVexillologist
🇺🇸 Multidimensional Radical Centrist
21hr ago
I am aware of people rejecting it at small scales, but I'd guess they are a small minority of at least my social network. I haven't heard a big company come out against it, for example, but lots of examples of them embracing it. I'm open to other data, but my guess is something like 5% (loud) rejection or less.
VoxelVexillologist
🇺🇸 Multidimensional Radical Centrist
22hr ago
Where do these kids come from?
If it's anything like my local school district, it's partly that people are choosing other education options (homeschool, private), and partly that people are having fewer kids and the youth demographics are different than the overall numbers. At least here, youth population is up (not even true everywhere), but district enrollment is down a lot.
VoxelVexillologist
🇺🇸 Multidimensional Radical Centrist
22hr ago
It's really the fundamental tension of post-scarcity, isn't it? Some people enjoy what they do, some do it for the paycheck, and others are a mix of both. I'm not opposed to automation (AI these days) replacing drudge work nobody wants to do, but I really would rather I get the benefit of post-scarcity before my paycheck disappears. In practice, we've mostly been fine so far because things mostly change slower than you'd think from the headlines.
On the gripping hand, I'm not even sure society is ready for post-scarcity: it's not at all clear to me that everyone would be better off if we could all stop going to work tomorrow and still get paid. Some people would do well and find new friends and hobbies, I'm sure, but I suspect quite a few would end up worse than today in some metrics due to the usual vices (drugs, alcohol, World of Warcraft). As much as I like the idea of The Culture/The Federation, I'm not wholly convinced it isn't a bit dystopian in its own way. I hear a couple of Banks' books consider that, but I haven't gotten to those.
For pretty much every other sort of technology with safety concerns, we don't just say "don't let it get out of the box". We also don't throw up our hands and claim it needs a general-purpose "alignment" solution and nothing else matters, which is what Anthropic and OpenAI seems to be doing. Generally we start putting boxes in boxes of layered safety and only have tragic outcomes when all the holes in the "swiss cheese model" align.
Look at civil aviation: it's tremendously safe overall, but only because of rules and practices built up over a century of accidents. But it's not solved by a "one weird trick" alignment, but by many layers of safety and security. Accidents, when they do happen, typically could have been stopped by one of the layers catching it.
In that recent Amazon crash, the pilot touched down despite the other human voice (a qualified pilot!) in the cockpit calling it too fast, despite rules requiring stabilized approach configuration minutes before they tried to extend the flaps as they crossed the runway threshold, despite numerous automated call outs from the aircraft safety systems ("sink rate", "minimums", "too low terrain" twice --- I don't believe any of those are normal in a landing). If they had gone around on any of those indicators as generally required in their training, it'd have just been another day.
From what I know, the rules for nuclear and chemical plants are similar: we design them so many things have to go wrong before a single safety issue occurs. Any we (should) take all such things (near misses) seriously. But our AI corporate overlords seem to think they know better with "alignment".
VoxelVexillologist
🇺🇸 Multidimensional Radical Centrist
1d ago
Would it be wrong to suggest that we should codify requirements to improve education for smart kids too?
VoxelVexillologist
🇺🇸 Multidimensional Radical Centrist
1d ago
Is the carbonated grape juice in the US typically bought as something other than a non-alcoholic alternative to a wine-drinking event? The white variety I've seen before at weddings and parties and such for kids when the adults are drinking sparkling wine (champagne, or other types). Not that anyone would object to an adult choosing it too (pregnant, alcoholic, religious objections, just a non-drinker).
I'd bet you can charge more for something people are already buying with much more expensive sparkling wine.
VoxelVexillologist
🇺🇸 Multidimensional Radical Centrist
1d ago
Be thankful you don't have to deal with raw CoTs (unreadable now)
I look at them for GLM-5.3 Flash (previously GLM-5.2) occasionally, and they're still comprehensible English, usually even well-structured. It's in English even when I prompt it in another language.
VoxelVexillologist
🇺🇸 Multidimensional Radical Centrist
3d ago
Due to the sorts of inequalities that people complain about regularly, banks have a lot more resources at their disposal than hackers typically do. Plausibly, they can afford Fable and Astra tokens to pentest and harden their own networks ahead of script kiddies having the means to buy them. I wouldn't expect that to last forever, but if they put effort into it they might not lose. Nation-state actors are different, but even then Western nations can afford better models generally.
Not a strong hypothesis, but pithily "the only thing that can stop a black hat with an LLM is a white hat with a better LLM."
Those sorts of targeting concerns are [ETA not] strictly new: there are plenty of WWII accounts of poorly-designed torpedos swimming in circles back toward their launchers. But they are valid concerns, and there are better IFF techniques (not AI) available for that sort of thing.
VoxelVexillologist
🇺🇸 Multidimensional Radical Centrist
4d ago
Can the fruit fly grok, say, the axiom of choice? It's both a simple and a very abstract concept. I'm not sure, but it's an interesting question.
I'm sure at least someone has considered more difficult target discrimination problems, which seem to edge closer to the "poetry" use case. "Is this language Russian or Ukrainian", or discerning between common Eastern Bloc hardware used on both sides seems relevant and difficult, especially with active countermeasures on silhouettes and shapes going on. That meme with three fingers seems relevant.
VoxelVexillologist
🇺🇸 Multidimensional Radical Centrist
4d ago
The usual arguments for the platform-ization of the Internet focus on social network effects and corporate centralization. Your argument has me wondering if there's another element of security that's been ignored. The big platforms can afford security professionals; my basic personal site is more likely to be running an unpatched (who has time for prompt patching) WordPress with an expired SSL cert. Let's Encrypt is finally a free option, but is going to start issuing 45-day certs by 2028.
Beyond that, the desire for a public web presence seems to have wanted over time. I don't want to be internet-famous. IMO the "I don't post anymore" problem is evidence that whatever effort the platforms put into "trust and safety" doesn't seem to have delivered either. But I do miss those naive times some days.
Also one of the things the Yuddites seem incapable of doing is figuring out that superintelligence does not mean omnipotence.
I find myself frustrated at a lack of what I expect would be cool theoretical results on bounds for intelligence. It seems clear that tiny microcontrollers are never running meaningful models, but I'd have expected some "this much hardware/weight count for this much intelligence" bound. But I suppose we still don't have a raw measurement (maybe even definition) of intelligence.
My intuition is that my (not-web-enabled) toaster is never going to achieve much useful thinking, and that actual intelligence has diminishing returns with scale in ways that make "super intelligence" useful, but not unboundedly powerful. But I don't have a strong argument for that.
Pithily: Modern AI has a Claude, but lacks a Claude Shannon.
The mood is already turning against AI even while the labor market is strong, it’s hard to believe that won’t get much worse as large scale layoffs start to bite.
I don't think "large scale layoffs" is necessarily the only future here. I'll certainly admit it's possible, but historically automation hasn't always gone that way. A century ago, we had lots of accountants, and most of their time was spent manually balancing arithmetic, maybe with a mechanical adding machine. Along came the personal computer (earlier for large businesses with mainframes), and most of that work can be automated pretty trivially. Yet there are more accountants today than there were a century ago. What changed is that they're not doing arithmetic, and the complexity of what they do has grown substantially. For all the complaints about the growth of the tax code (and frequent related stunts by politicians), we are able to handle that complexity because we can offload the line-by-line balancing to software. Some of that complexity is probably wasteful, but some of it is also better auditing and accountability, or targeted governmental incentives — some, again, is probably wasteful, but some seems reasonable.
I don't have quite as coherent an argument (still a work in progress), but it's also the case that "recession caused by mass AI-driven unemployment" is actually a self-own by corporate AI overlords in limiting their upside. If you replaced half of human workers tomorrow at a cost savings, any not-truly-essential business is going to lose customers who can't afford their wares: tech companies fit pretty high on Maslow's pyramid, and most of the base ones are long-commodified — at one point, most of us were farmers, then along came the tractor. Nobody in the West is starving because they can't get a farmhand job. The cratering of farm employment is, in fact, why I get to work an office job, rather than out in the fields somewhere, and I'm not sure I'm worse-off for it.
But that latter part is, again, a half-formed argument.
VoxelVexillologist
🇺🇸 Multidimensional Radical Centrist
4d ago
I think we're ultimately going to have to agree to disagree here: I think the moral valence of the two parties here is less clear. If pressed, I'd probably weakly lean towards Israel (that Golda Meir quote on loving children doesn't seem completely wrong), but they're hardly blameless victims.
I appreciate the forthright discussion, though.
I'm not sure it'd ever fully replace ownership. Many people don't treat their cars as interchangeable parts today. If you have a work truck with tools in the back, you may not want to move those often. Or reinstall a bike rack every time you use it. I do think it could replace a lot of uses, but probably not all of them.
More options
Context Copy link