Anvoker

I'm not sure I believe in FOOM, but I don't think the hardware argument is so cut and dry. LLMs have had vast improvements to capabilities attributable purely to software improvements.

Considering how calorically efficient the human brain is when it comes to the breadth and depth of general problems it can solve, I don't feel like LLM software improvements have hit a theoretical limit yet.

I don't want to be uncharitable but the Mt. Everest analogy is not that far away from directly justifying not adding accessibility features in order to satisfy the ego of specific gamers.

I agree with you that in some sense it undermines the achievement, but tending to the ego of players that feel undermined by the existence of accessibility features is not high in my hierarchy of values personally.

In the era before the advent of mass digital surveillance, it was well understood that there are many different ways to fight crime. Now every time I see this conversation being had, it's treated as if keeping any amount of privacy beyond what you personally find acceptable is tantamount to endorsing the existence of human traffickers. Warrantless spread-shot digital surveillance is now often treated as essential.

I have a question for you. Why don't you advocate for more surveillance? The more active and widespread the surveillance, the more criminals can be caught. If you think it's not entirely practical, just embark on a thought experiment with me -- assume it would be practical. Would you do it? Would you have everyone be under perfect surveillance 24/7 in order to catch every criminal? Let's call this stance surveillance maximalism.

If you agree with surveillance maximalism, then we're just very different people and I don't think we can find common ground. I hope we can live peacefully in different countries with different laws that suit our preferences.

If you disagree with surveillance maximalism, then why is your arbitrary tradeoff so good? It's not obvious why it's better. You are assuming the moral high ground, but you're also doing the same thing you're accusing me of -- you're accepting some amount of traffickers existing by not being a surveillance maximalist.

By adopting this moral high ground, the discourse is kept at a shallow level. We talk less about which surveillance measures are actually effective or not, what has happened to crime rates over time, what is the context in which crime occurs, what are the negative consequences of undermining privacy, what are the negative consequences of mass surveillance etc. Instead we waste our time and energy on cheap moral opprobrium.

And why not? Law enforcement does not represent supreme justice and good. There are higher moral principles out there. Respect for law enforcement in the absence of justice is a disaster for society. All it does is give cover for bad actors.

I feel the opposite. It is one of the genuinely competent languages out there, if you're the kind of person that isn't upset about it being a "kitchen sink". Instead, I feel like the ecosystem, community and reputation hold it back. There are nowhere as many useful and popular packages to reach for as Python for example. Unity3D always lags behind in runtime and language version and had amazingly bad package management -- this further hamstrung C#, because the language is never used in a vacuum.

This kind of opacity is unacceptably user hostile. It's not okay to treat some amount of developers as acceptable casualties, without them even knowing, in order to help enforce a restriction that only serves Anthropic's interests. And if you want to tell me this is for managing the x-risk factor, I'm frankly unimpressed.

Because the poor will suffer tremendously if they are ill equipped for the job market. If we want to have education have a strong humanistic focus then we should start by changing society in other places. Or we'll get unemployed and suffering well rounded human beings.

Regardless, at the moment education does not seem to be doing great on either the humanistic or "pragmatic" axis.

Unfortunately, framing the problem as this kind of dichotomy is something people are inclined to do because then the problem can be reduced to the unwillingness of the opposing side to face reality.

Sometimes the dichotomy is correct, but the bias exists.

What about setting environments for normies that mitigate this problem? I don't know that you can do it on Windows, but Linux offers various tools for isolation where you can give full rights to an LLM and still be safe from certain classes of disaster.

Maybe this kind of isolation neuters the benefit you're thinking of, but I do believe some sort of solution could be reached.

I recall having a conversation with a friend that was angry at this kind of thing sometimes being called "Tail Call Optimization". It's a guarantee that your program won't literally crash by stack overflow if you recurse too much using tail calls; the fact that it might have performance benefits is secondary in importance. So the name can be a bit misleading.

Back then Rust's tail call story looked bad. The problem slipped my mind for years and now I'm suddenly made aware of this become keyword, and it seems like a great idea -- to make the developer's intent more explicit. Getting better target support for this is doing god's work.

As an aside, I would slightly argue against the idea that it's useful only for a narrow set of libraries. It's useful for the developer as a user of the language to be able to express things in either iterative or recursive style without worrying as much whether the latter is going to panic. Maybe if I had to deal with all the work of making such features functional in a language, I would value expressiveness a bit less!

Trump's second term has had dire consequences for the US that are simply not on the same level as Obama, and it could have been avoided. The democrats are also terrible, but it's a matter of harm reduction. In this case, I would strongly argue that the difference was greater than 99 and 100.

You should vote to do harm reduction. Elections happen regardless of what you do. Whether 30% or 100% of Americans vote, the winners of elections still get access to the same amount of state power. The system does not require our political participation to continue to exert control our lives. Abstaining from voting is not an effective tactic in reducing the legitimacy of the system. That tactic might work in other situations, but not in this one.

I hope you will keep your distaste for both parties, but still vote for the lesser evil, even if it's distasteful. Because I think we should help that one person not get stabbed. And if indeed you have voted for the lesser evil and my post has a tone that assumes otherwise, I apologize.

The fact that the police won't use surveillance data in a way you would consider good and effective does not mean you don't live in a surveillance state.

One of the key aspects of police and surveillance states is that the incentives are structured so that the policing and surveillance need not be done with the interest of public welfare in mind. As you said there is no political will.

They are a tool like anything else.

Exposure and Response Prevention therapy works. You will never get fully well without exposure. However, it requires that you find stimulus of a magnitude that makes you uncomfortable, but doesn't send you outright spiraling. You need to keep steady while experiencing it for a while.

Content warnings give you the ability to estimate what intensity of negative stimulus you will experience, and this is important when dealing with actual triggers.

Not everyone is yet at the phase where they can handle a certain level of exposure. For some unfortunate cases it takes a long time to be well enough to start being able to handle exposure.

That being said, I do think content warnings need to be specific, not generic. The most useful ones are spoilers, not generic messages to put you on guard. Careful Ao3 authors do a better job at this than most games. There are technical solutions that allow interested parties to get this information without having to spoil the default audience, but we live in a busy world that has a lot of things to care about other than this.

If you've ever had trauma, especially recent, you'll appreciate well done content warnings. You don't want the dramatic plot twist to happen to be exactly the topic you've been trying to avoid so that you can slowly get better.

If you've experienced a certain kind of trauma, it's not a matter willpower. It involves a loss of control over one's emotional response and thoughts which can be triggered by things that relate to your trauma.

Don't knock on content warnings just because they lack rigorous evidence or because "trigger warnings" became the butt of jokes for a while. They have a genuine utility.

Were unions involved in Epic's layoff?

Often when I see layoffs like this, I can't help but think "Wow that company has so many employees and yet, in practice, does so little". This perhaps a rather uncharitable sentiment, but I can't help but have it.

Yes, Unreal Engine keeps getting improved, more Fortnite content gets produced. But there is a general lack of innovation, one that I find personally painful when I look at Epic's recent-ish track record. Needing to fire this many employees is not just a result of market conditions, but also a straightforward consequence of not being able to leverage them for sufficiently lucrative outcomes.

Companies with this amount of capital are well positioned to take multiple strategic bets which aren't at all safe bets, but pose no real financial risk for the company in aggregate. Why do these bets end up being taken instead by indies with much more to lose? Well, partly because indies often _need_ to take riskier bets to carve a niche. But the other side of the coin is, what I can only surmise, a lack of imagination and adventurousness on the part of management. They could be funding many experiments and seek to have another hit like Fortnite, perhaps in a somewhat different market. Having to seek another hit while your finances are declining is less pleasant.

When a company loses its edge in this way, as long as it hasn't _really_ captured a demographic or created some very sticky ecosystem, it's bound to get whittled down repeatedly. I doubt that Epic will suddenly get more creative and adventurous at this point, but perhaps necessity will have its part to play.

(Aside from all of that, I agree with most commenters here that the layoff is being handled about as gracefully as one could reasonably hope.)

By accepting that valuing the life of a cop over a criminal is a good reason for American cops to behave the way they do, what you end up encouraging instead is cops valuing their own lives over those of citizens to a shocking extent.

One issue is that security theater creates demand for itself. Do things that induce worry and a tendency towards paranoia in the more susceptible parts of the population and then you will gradually raise the general alertness of the population. This then manufactures a desire for these measures. It largely rides off of people's general unwillingness to entertain just how many of the measures are ineffective or nonsensical. "It can't all be pointless. Surely some of it must make us safer." It's not an unreasonable belief in itself, but everyone having that attitude lets security theater grow cancerously.

Agreed, and not just engineering quality. The company I've been with for 8 years is on its last legs and massive overhiring has had a huge hand in it. It cost a tremendous amount of money and made management more difficult, reducing the quality of the decisionmaking in many areas. It made its strategy deviate further from doing the things that would have made the company sustainable. It also devastated employee morale with the subsequent repeated waves of layoffs.

> The second sentence is false. Lots of people blindly follow things and don't understand consequences until they brick their devices. Those who don’t break something won’t notice if they’ve silently backdoored themselves.

"Lots of people", how many though? Can that number be reduced? What number would be acceptable?

I feel like it _has_ to be possible to devise an unlocking procedure that dissuades most people from self-harm.

The problem is often treated as intractable, but intuitively this seems really unlikely to me. I don't think more than a tiny percentage of Xiaomi owners, for example, would go through the bootloader unlock process which often has a mandatory wait period attached to it without a reason more compelling than an impulse to randomly and blindly follow instructions on the internet.

I would like to see user studies with good methodology before other people decide to barter long-term freedoms away for insufficient benefit.

Why do I so rarely see people who are concerned about the security issues of bootloader unlocking calling for designing hassle and warning into the process. Instead, it's more common to hear that in the name of the average user, all escape hatches must be removed.

I felt like outside of bug fixes and QoL improvements, Revision's maps and soundtrack were a downgrade compared to the original. I feel quite strongly about this, but I recognize it's also largely up to taste and it's valid to like those changes.

But for me, they made the experience worse and they were enabled by default. As a result, I gave it a thumbs down on Steam and did my best to explain my thoughts in detail. I received a couple dozen negative comments over the years on that review, largely in the vein of how dare I give negative feedback to a labor of love provided for free. That kind of argument did make me feel guilty, like I was being unfair to the developers. I eventually changed it to a positive review, and now I regret doing that. I allowed my genuine opinion to be clouded.

This was an extremely tame internet conflict overall, I'd feel ashamed to frame myself as a victim over so little. What I'm trying to say is that both sides are capable of failing to genuinely engage with the other.

It's definitely true that Revision has been to some degree unfairly attacked. There are purists who do not give it a fair shake and make ludicrously confident statements, peddling opinion as fact. But there's also legitimate reasons to dislike it. Not knowing you, I am not at all accusing you that you'd be lacking nuance on this topic. I'd just like to say as a general statement that discourse ends up healthier when people care about distinguishing between people who disagree with you versus people who disagree with you _and_ that are acting in bad faith.

Are you sure that the jankiness you describe is not related to you not being close enough to the enemy for it to count as a point blank hit? The engine does a simple distance check which is rather unforgiving, and if you utilize the range your melee weapon affords you, you will fail this point black attack check. This is just a guess on my part, but it's the #1 reason I've seen people be perplexed at seemingly inconsistent melee stealth takedown behavior.

If you first push yourself as close as you can get to the enemy model and attack the general area of their torso, it works every single time. But again, this is just me taking a guess as to what the problem you're referring to is!

I don't think any amount of ethics education will matter in the end in the face of the incentive structure that appeared in the industry.

Strong cultural norms (e.g. hacker culture) might help for a while. But incentive structures eternally erode opposition.

It could make it easier for developers to band together and try to collectively veto things like this. But corporations with money can always buy the expertise of people, have them undermine the community, create their own parallel communities and influence public opinion and legislators.

FAANG salaries supercharge people's cognitive dissonance. They will find ways to excuse, minimize and ignore their contribution to the current situation.

Even HackerNews developed a sub-subculture of people that were constantly going on threads and calling remote attestation worries as "FUD".

It's unclear how to preserve cultural norms that stand in the way of market dominance. The only thing I can think of is having competing interests in the market. But whenever these align -- hell breaks loose.

I wonder if you could obtain stimulants and if it would help your particular situation. It's not like ADHD meds are somehow chemically designed to help only with ADHD. They can help with working memory issues and with chronic fatigue like symptoms.

I'm thinking about something like Concerta, aka extended release Methylphenidate. Probably low or medium dose. I'm specifically mentioning this because you seem like you could be at risk of burnout or just pushing yourself too hard. And being on a lower extended release dose limits how much a stimulant enables you to push yourself too hard, in my experience.

The idea isn't entirely novel: - https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9214912/ - https://pubmed.ncbi.nlm.nih.gov/26991608/

I've read studies referring to stimulants as helping "normalize" an ADHD brain over time, so it's not unthinkable it could help "normalize" a brain in other situations, though what exactly that word means is probably medically tumultuous. But, in short, I think stimulants have the ability to change a brain over time, not simply be a "pick me up".

I don't know what the chance of this helping is, but I felt obligated to mention it, because in your situation I'd be willing to try anything that didn't have a chance of further messing me up. And stimulants in therapeutic doses have an extremely low rate of long lasting side effects, but again, your situation could be really different.