advael

Better evidence is always good, but we much more drastically need better policy. Psychadelics' potential benefits are still being explored, but their risks are pretty well-documented and quite minimal compared to a broad swath of substances legal for adults to acquire. There is no reason consenting adults should be criminalized for seeking them out

Lots of people like that analogy because it's extremely self-congratulatory. It also makes no fucking sense

Despite the fact that people keep insisting on the buzzword "AI" to describe these large neural networks, they are more succinctly defined as approximate computer programs. The means by which we create them is through a relatively standardized family of statistical modeling algorithms paired with a dataset they are meant to emulate in their output

A computer program that's specified in logic is already a usable representation that can be used to understand every aspect of the functioning code in its entirety, albeit some of it may be hard to understand. You don't need to consult the original programmer at all, let alone read their mind

In contrast, a function that is approximated in the manner described needs the training data to replicate or make sense of it, and in fact is even necessary to assess whether the model is cheating at the benchmarks its creators assess it against. The weights themselves are a functional approximation, not a functional description

For the purposes of the ethos of free and open source software, it is obvious that training data must be included. However, this argument is also deployed in various other places, like intellectual property disputes, and is equally stupid there. Just because we use the term "learning" to describe these systems doesn't mean it makes sense for the law to treat them as people. It is both nonsensical and harmful to say that no human can be held responsible for what an "AI" model does, but that somehow they are "just like people learning from experience" when it benefits tech companies to believe that

Just counting things I've personally seen, using "untrusted" userspace software and accessing a USB webcam. It's possible these particular examples are no longer true. I haven't been a windows user for about 15 years. But both surprised me quite a lot when I saw people I know do them and it seemed indicative of a systemic separation of concerns problem

At this point few besides the pedantic or motivated seem to want this, given that they've been massively diverged projects for at least half a decade

It'd be a lot less easy if the OS didn't seem to require full privilege escalation for a lot of tasks you don't need that for in linux. One of the major problems that leads to escalation is poor separation of concerns

I think that statistic is a good indicator that we have significant broken incentives in the way liability can be shunted away from the people with decisionmaking power in businesses. The "corporate veil" for example is a deep loophole in the whole concept of rule of law, and I for one would prefer it not to exist

Yea, like I'd rather have a doctor who actually listens to what I'm saying than one who got top marks in medical school, and honestly network engineers I've worked with who strike me as "higher IQ" often fall into the pitfall described by the adage "the perfect is the enemy of the good" (which I empathize with, because I am also a high-IQ person with this kind of bias, but it's actually quite counterproductive at times)

My education has equipped me quite adequately to fend off such critiques, and frankly I think it's a misuse of terminology to claim that studies "prove" anything, because the nature of empirical evidence even under the most generous assumptions is that a body of evidence can support or contradict conclusions, but "proof" as a concept comes from logical modes of inference, which only work within contexts where the conclusions drawn pertain to assumptions that are defined in a stricter sense than any of these fuzzy real-world questions permit

Neither of your examples really contradict my point. If we are measuring a subset of things you need a brain for, of course factors that make your brain generally less effective are going to show up on them. Lead poisoning and inadequate nutrition also decrease libido for example, and I don't think it's reasonable to conclude that this provides strong evidentiary support that libido should be our proxy for people's general capabilities and declare it equivalent to "intelligence", though there are certainly those who would try to make that case too

I've always thought the Flynn Effect was yet another Moore's Law, and both demonstrate Goodhart's Law. We didn't get transistor density doubling year over year because this was a natural tendency of some physical law, we got it because of mountains of money and engineering effort poured into doing that. Similarly, we didn't get rises in a measure of what society considers intelligence because of some genetic or uncorrelated environmental factor, we got it because what we're actually measuring with IQ, since it was invented by Binet, is education and selection of minds toward a particular category of reasoning tasks. I think Binet did something impressive when he distilled the priorities of the entire education system into a psychometrically-evaluable quantity that's relatively stable across individuals, but the way people overstate the generality and broad applicability of this metric baffles me. Like people will point out that IQ correlates with lifetime income and also SAT scores. But those two things not only correlate with each other for a much clearer reason (We gate educational outcomes with standardized tests that resemble IQ tests, and also gate economic outcomes with educational credentials), but that reason also is a causal explanation that mostly amounts to a decision made by human institutional policy, rather than some grand enduring fact about the nature of intelligence

Like look, I get how people get so attached to this stuff. I've got a high enough IQ score to be beyond the statistical sensitivity of Raven's Progressive Matrices which mostly seems to mean that I'm great at test-taking and puzzle-solving, and I get that IQ seems like this really powerful explanation of variance on vibes from an internal perspective. It feels like people above a certain point on the curve are just so much more vibrant and curious sometimes! But I've come to believe that this is more an artifact of how we shunt certain people within the education system than some deep difference between people on some generally-applicable measure of capability. Lots of social science theories about assortative mating and such seem to observe that people with brains that work similarly will also tend to like each other more, and we have vast edifices that determine other socialization-determining factors based on these kinds of heavily-correlated metrics.

That is pretty funny, and the whole idea that you can't make money packaging open-source software in a way that's more appealing to people is definitely funny given that this is the business model of a lot of successful companies

I do however think this leads to a lot of problems when those companies try to protect their business models, as we are seeing a lot of today

Generally speaking a lot of "essential tools" in "cloud computing" are available as free, boring operating system utilities.

Principle of least privilege trivially prevents updating system packages. Like if you don't want people using apt, don't give people root on your servers?

You know you can just do this with keyauth and a cron job, right?

Of course they do. Every company doing predatory shit has some slick liars on payroll to come up with a story about how their lock-in policies and kafkaesque contract terms and surveillance and "opinionated design decisions" (or, "attempts to technologically control their users' behavior" for those who prefer plain English) are for your own good, actually, and any attempt to corral their behavior is an attack against their users

The messaging of companies in regulatory cases has become rote and predictable as the sectors they occupy have grown more concentrated. There is no reason to heed it at all in the present environment

I'm not sure that's a warranted assumption based on this result, exciting as it is, we are still seeing replication of an extant testable world model, rather than extrapolation that can produce novel mechanics without them being in the training data. I'm not saying this isn't a stepping stone to that, I just think your prediction's a little optimistic based on the scope of that problem

Dang this is the first paper I've seen in a while that makes me think I need new GPUs

I mean it sounds like a way to trick people into letting a giant company build a surveillance system, and I imagine aliens observing humans in this moment in history might conclude that's what we want, but for science or the benefit of the people "surveyed" it's mostly downsides

Mostly because of a very successful propaganda campaign by people who sought to loot the post-war economic boom

Generally speaking, the sad truth of a complex economy is that coordination is hard, and there's usually a short-term privatized gain to be had by someone willing to poison the future and the commons. No amount of benefit to humanity overall or even the specific society such a person lives in will convince a person who simply doesn't care about anyone else. Fortunately for humanity, a very small minority of people actually operate like that. Unfortunately for humanity, some of them have managed to accumulate a lot of power

Kellner's law. I like it. It's an excellent illustration of the cowardice and hostility toward customers that comes from being in the subscription-twiddling business model

If Kellner wanted to make a deal with cable subscribers, why didn't he advertise cable that way? "If you buy our package and agree to watch ads when they come on in order to support our monetization model, you can watch over a hundred channels!"

But he didn't do that, of course. Instead, when his business model clashed with reality, he blamed the customers for violating an imaginary contract they never signed. Just because execs want to imagine violations of their assumptions as crimes doesn't mean that they actually are, and we should be using every meager lever of power we have access to to prevent governments from making laws to appease people like him, because every time a law like that is made, more people's rights are made subject to the arbitrary retroactive demands of delusional businessmen

This is amazing

You're keeping charts, right? I wanna know what the top hit on that block is next month

Haven't yet seen a reason to use VSCode and can't anticipate I will

As an editor it's a pain in the ass, as an ecosystem it feels like a fucken freemium mobile game for how polluted and nickel-and-dimey extensions are, and worst of all it's owned by microsoft so while this article is informative on the specifics, I knew from the jump there was some legal fuckery involved because there literally always is

I can't be a total purist and never use a tool that's connected to a large tech company's ecosystem, but where at all possible one should, and this is only becoming more true

Could and would

If this model of "theft of service" is legally enforceable, then significant swaths of this industry are engaging in false advertising and possibly full-on fraud by not making this agreement they're entering into clear to customers prior to their purchase of the device

From an ethical perspective, you will never convince me to feel bad about any "theft" of this kind, and this kind of dilution of concepts like "theft" and "crime" is likely responsible for the emotional valence of those terms reading a lot less significantly negative to people

Claims of automation really fall apart for me when they rely on outsourcing. Even setting aside ethical considerations, this business relies on there being a stable source of cheap labor. Moving that labor to remote monitoring and presumably some kind of local intervention skeleton crew for when they report something may be more in line with the experience customers want than a gym that is perceptibly staffed, but automation is a strong word for that

In trying to write a book, it makes little sense to try to "compete" on speed or volume of output. There were already vast disparities in that among people who write, and people whose aim was to express themselves or contribute something of importance to people's lives, or the body of creative work in the world, have little reason to value quantity over quality. Probably if there's a significant correlation with volume of output, it's in earnings, and that seems both somewhat tenuous and like something that's addressable by changes in incentives, which seem necessary for a lot of things. Computers being able to do dumb stuff at massive scale should be viewed as finding vulnerabilities in the metrics this allows it to become trivial to game, and it's baffling whenever people say "Well clearly we're going to keep all our metrics the same and this will ruin everything." Of course, in cases where we are doing that, we should stop (For example, we should probably act to significantly curb price and wage discrimination, though that's more like a return to form of previous regulatory standards)

As a creator of any kind, I think that simply relying on LLMs to expand your output via straightforward uses of widely available tools is inevitably going to lead to regression to the mean in terms of creativity. I'm open to the idea, however, that there could be more creative uses of the things that some people will bother to do. Feedback loops they can create that somehow don't stifle their own creativity in favor of mimicking a statistical model, ways of incorporating their own ingredients into these food processors of information. I don't see a ton of finished work that seems to do this, but I see hints that some people are thinking this way, and they might come up with some cool stuff. It's a relatively newly adopted technology, and computer-generated art of various kinds usually separates into "efficiency" (which reads as low quality) in mimicking existing forms, and new forms which are uniquely possible with the new technology. I think plenty of people are just going to keep writing without significant input from LLMs, because while writer's block is a famous ailment, many writers are not primarily limited by their speed in producing more words. Like if you count comments on various sites and discussions with other people, I write thousands of words unassisted most days

This kind of gets to the crux of why these things are useful in some contexts, but really not up to snuff with what's being claimed about them. The most compelling use cases I've seen boil down to some form of fitting some information into a format that's more contextually appropriate, which can be great for highly structured formatting requirements and dealing with situations which are already subject to high protocol of some kind, so long as some error is tolerated. For things for which conveying your ideas with high fidelity, emphasizing your own narrative voice or nuanced thoughts on a subject, or standing behind the factual claims made by the piece are not as important. As much as their more strident proponents want to claim that humans are merely learning things by aggregating and remixing them in the same sense as these models do, this reads as the same sort of wishful thinking about technology that led people to believe that brains should work like clockwork or transistors at various other points in time at best, and honestly this most often seems to be trotted out as the kind of bad-faith analogy tech lawyers tend to use when trying to claim that the use of [exciting new computer thing] means something they are doing can't be a crime

So basically, I think rumors of the death of hand-written prose are, at least at present, greatly exaggerated, though I share the concern that it's going to be much harder to filter out spam from the genuine article, so what it's really going to ruin is most automated search techniques. The comparison to "low-background steel" seems apt, but analogies about how "people don't handwash their clothes as much anymore" kind of don't apply to things like books

I think there's an explicit rule on hn against metacommentary on voting, but it can be a useful signal of the vibes within the community as a whole, and I do think that critically examining the VC model of tech startups kind of edges close to the sacred cows of this particular community (it is a public forum, but it's also ycombinator). But even if you're not bought into that particular organization's messaging, I do sympathize with not wanting to feel like your life's work is executing a bunch of cargo-cult behaviors to keep some investors who make all the real decisions happy with you. While I do think what I said was based on my best attempt at a sober assessment of reality, a synthesis of my own direct experience and the described experience of friends and colleagues and glances at the overall landscape from broad reporting done by various sources, it's a really bitter and painful thing to understand

Not to stereotype but I'm not sure I've ever read a British writer who didn't at least a little bit sound like that. I wonder what it is

That would be true if this was a market where funding mostly came from making products and generating revenue

It isn't. The majority of tech companies are speculative and funded by venture capital, and what's worse, many of the largest companies, which actually do generate revenue, are monopolies

What this means is that efficiently competing is only a viable strategy if you can convince a small pool of potential investors - many of whom think "competition is for losers" - that it's a good play, while also avoiding being bought out and crushed by one of a few trillion-dollar companies

The dynamics you're describing are those of an open market, and with wealth disparity at an all-time high (so your customers have less influence on average) and financial scams fairly rampant due to decades of lax enforcement, most companies right now are not effectively operating in an open market

Buying SaaS is trendy conventional wisdom. It has benefits and downsides, but in my experience the biggest benefit is that VCs think it's smart to do and also often benefit from a company buying the products of other companies. Whether it helps with things like making products faster or better overall is close to immaterial in this environment

The anthropomorphizing pattern I picked up on was the whole phrase "chat with someone", because while I think LLMs are interesting and useful tools in a lot of ways, it's a pretty drastically different experience from talking to a person, and I think that a lot of the marketing of LLMs relies on people doing this and kind of "filling in the gaps" for what they imagine these models to be both doing and capable of. These differences are pretty stark and meaningful, and the strongest sign of that to me is that I've noticed there are a lot of people who don't bear this in mind and interact with the things often who are starting to exhibit what I would previously have identified as signs of significant social withdrawal, except instead of sounding like, I dunno, their favorite youtuber's political polemic or somesuch - which has by and large replaced the characteristic atrophy of verbal fluency we may have seen in a pre-internet era - their stilted speech trends more toward the professional and somehow both confident and airy tone of popular language models. I worry that this anthropomorphizing mindset may carry some negative cognitive consequences in the medium to long term, analogous but different from those of the rise of social media

As far as my job hunt goes, I'm not finding myself rejecting positions out of disappointment, but noticing that I'm often rejected by C-suite people after being technically vetted, often after having what I believe was a pleasant, positive, and often even generative conversation about what the company's plans with AI are and how I might help accomplish them. As I said before, I really do try to stay positive, and I think when putting my best foot forward, like in a job interview, I tend to succeed at that, but my experience leads me to be more negative about this moment in the industry when I am more candid, such as in this forum. I think if you're going out for work in whatever the tech bubble du jour is, the people you're talking to have really different biases and expectations from those hiring for more "boring" development jobs, in a way that makes me want to just go out for more general roles and lie low for a while, except that since I've been working on primarily ML-related projects for most of the last decade, it's also hard to convince people not in that area that I have adequate relevant experience. In this context and with bills to pay, it's hard to stay optimistic