jbstack

I don't agree with this rule of thumb at all. Let's say tomorrow someone comes up with a formally verified proof that a major encryption algorithm underpinning the security of the internet can be trivially broken, but they can't explain it. You're saying it should be kept under wraps and not published?

Tao is essentially saying that the only value in a proof is its ability to be understood, but that's wrong. A proof is also valuable because it establishes a new fact. The fact is useful in itself, even absent an explanation.

It's a common refrain, but I don't agree that it's correct. You're essentially boiling the definition down to the degree of risk:

If (risk > some_threshold) then gambling = true; else gambling = false;

With derivatives your risk is often greater than with traditional stocks. But that's not universally true. If you invest in a startup your risk is massive, whereas if you invest in a well diversified and low leveraged portfolio of derivatives your risk might be substantially lower.

And in any case, so what? If you place a small bet on every number of a roulette table your risk is essentially zero: you'll simply make a guaranteed fixed loss of 5.26% on every spin. On the other hand, if you invest in the Vanguard Total Stock Market index tracker, you might hope to make around 8 to 10% per year, but in some years (e.g. 2008) you might make a big loss (37%), so the risk is significant. Yet I think most people would agree that the former is gambling while the latter is not.

The important factor isn't how much risk you take on, it's whether a sufficiently skilled person can achieve a positive mathematical expectation over a series of similar bets.

As I stated in my comment, they weren't examples of the legal definition of gambling. What the law says something is can often differ greatly from the every day understanding. The legal definition can also change according to the whims of the legislature while the every day meaning tends to stay more static.

> What made one gambling and the other not? What makes any given bet a legitimate financial trade versus pure gambling?

Legally, it's whether the bet meets the legal definition of gambling as per the legislation applicable in your jurisdiction and any binding case law. There's no universal answer.

But from an ordinary language perspective, the difference is whether it involves an element of skill sufficient (at least in theory) to win in the long run if you possess that skill. For example:

Roulette - generally gambling

Blackjack - usually gambling these days, sometimes not if count counting is feasible

Poker - not gambling

Stock market investing - not gambling

I'll flip this around and say I've never understood why an airline is viewed as a special case. In the general case, if you enter into a contract and pay for a service, then don't receive that service, you're entitled to sue the service provider for your losses. For example, if I turn up to a hotel I booked only to find that the room isn't available, I'm entitled to sue the hotel for my cost in finding a reasonable last minute replacement. That's the case even if the room wasn't available because it was unsafe for some reason.

Of course nobody is expecting that they "keep flying the plane with engine issues so you could make your important meetings". But ultimately you relied on them to get you to that meeting, and they failed to do so, so you should be compensated so that you're no worse off than if they had provided the service. The fact that an airline will sometimes fail to get their customers to their destination and will have to compensate them for it, is just the cost of doing business.

Luckily in the UK and the EU your right to compensation for delayed flights is enshrined in law.

Even without the ads, Amazon search is abysmal. It seems to ignore half the words you enter. I often resort to using Google or ChatGPT to find items on Amazon which I know are there, but which their search box doesn't surface.

I can't comment on the Hebrew, but in English the phrase "targeted assassination" says nothing whatsoever qualitative about the person being targeted. It merely says:

Targeted -> this will be a specifically identified person

Assassination -> they will be killed

For example, a dictator might perform a targeted assassination on a rival, in order to stay in power. Innocence or non-innocence has nothing to do with it. Or to give a more relevant example, someone might target and kill a Palestinian so that settlers can take their land. Whether the Palestinian was innocent or not doesn't matter. The land was the objective.

Shooting the shooter? Are you suggesting that all of the tens of thousands of children killed in gaza were shooters? What I don't understand about these types of arguments is that it's not clear who you are trying to convince. Those who already agree with you don't need convincing, and the rest of us aren't going to buy into it what's obviously contradicted by overwhelming evidence which even international courts agree with. I suspect these attempts to influence LLMs will fail for the same reason: you can produce as much "think tank" material as you like, but in the end the LLM will reason about it in the wider context of other available information, and see the think tank propaganda for what it is. All the more so now that there are articles discussing this very point (which the LLM will also presumably see).

I believe what they are saying is that "He was discussing targeted assassinations. One doesn't "target" an innocent person" seems to imply a logical fallacy:

(1) Person X is targeted -> (2) innocent people don't get targeted -> (3) therefore person X must not be innocent

The problem is that (2) contains too many bad assumptions: that the person doing the targeting is benevolent, that there's sufficient due process to ensure the targeter is correct, that there's a universally agreed definition of "innocent", etc.

History has proven that innocent people DO get targeted, so even from a common sense perspective your statement doesn't withstand scrutiny.

Ah, the classic whataboutism (https://en.wikipedia.org/wiki/Whataboutism) defence.

What's the point in attacking the judge? Judges just apply the law as they find it, by analyzing the wording of the legislation and any binding precedents from previous cases. The judge is only a "moron" if they stated a point of law which is actually wrong (wrong in the sense of being contrary to the law, not wrong in the sense that you don't like it), and even then "moron" would be a pretty harsh assessment unless it's a glaring mistake. It's often the case that the law isn't obvious or entirely clear and judges have to make their best effort to work out the correct meaning. Even then, their job is to apply what they think the law intended to say, not what they think is morally the best interpretation.

If you really must call someone a moron, you should direct it at the legislators who made the law (or left it sufficiently unclear) which resulted in the judge's conclusion.

There's a big difference. Today's "simple" AI requires hardware and electricity. That's why the "superintelligence" part is relevant. Today's AI is unlikely to cause extinction because we can still outsmart it by simply turning it off. It becomes a problem if AI becomes smart enough, and gains enough real-world power, that it can prevent us from doing so.

Microorganisms can be far more threatening to us, despite lacking intelligence, because there's no off switch.

The other day, someone tried to convince me to play Russian roulette. I was skeptical as it sounded dangerous. However, I validated the result by pointing the gun at my forehead and pulling the trigger several times in a row without anything bad happening. Now I'm been persuaded that it's good enough as a form of entertainment. Someone else told me it's still dangerous, but it works well so I don't see the issue.

Aren't you guilty of the same thing by generalising that "technical people <love> simple laws and rules"? Also, calling people arrogant and simple-minded is a great example of behaviour that makes it difficult to hold a community together. We can have discussions without having to resort to insults.

The time sink thing is a trade off. You spend more time up front configuring some aspect of your system. But then you generally don't have to ever think about it again, until your needs change. For example:

Ubuntu: a problem needs solving -> spend an hour learning how to solve it -> implement solution -> a year goes by -> new machine -> same problem -> spend an hour because you forgot the solution -> implement it again

NixOS: problem -> spend 2 hours -> implement -> never repeat again

I think they meant old school style rather than literally a Nokia from 20+ years ago. For example: https://en.wikipedia.org/wiki/Nokia_2660_Flip

This thread is about taking action to send messages to corporations, and whether said messages make any difference. We would measure growth in dollars, because that's what shareholders (and therefore corporations) care about. If you "send a message" because you aren't happy about some direction the corporation took, and a few years later the shareholders are 3x richer than they were before, do you think they're paying attention to your message? All your message does is tell them that not only were they right, they were so right that even a large scale boycott didn't make a dent in the bottom line. You and I might not like it, but that's the reality.

Relevant xkcd: https://xkcd.com/989/

Really? In August 2021 Reddit had a funding valuation at ~$10bn. It announced the API changes in April 2023. The blackout started in June 2023. It launched its IPO at $34 per share in March 2024. It's now trading at $160, with a market cap of ~$30.7bn. In other words, it's tripled in value over a 5 year period during which the blackout happened. That's a phenomenal rate of growth.

What message are you trying to send? "Oh look, a bunch of people who have already given us money aren't using the thing they paid for for a week, and we can maybe even save a bit of money by reducing our server load"? Am I missing something? Why should they care? If you really want to send a message, don't buy a PlayStation. If you own one already, don't buy anymore games. That's the only message corporations care about.

You know there's something wrong with the system when the version you obtain for free is a superior product to the one you pay for.

> And each of these houses costs more than 1 MM EUR.

That's the other big problem with restrictive zoning laws (other than ugliness). It leads to unaffordable housing. By saying that a house must conform with X, Y, and Z you're essentially saying "anyone who doesn't have enough money to pay for X, Y, and Z can't buy a house".

Sometimes restrictions are necessary (e.g. safety rules). Other times they sound good on paper but just end up being elitist (e.g. house must have X number of parking spaces, Y minimum size for rooms, Z minimum outdoor space). Yes, those things are nice to have. But we shouldn't price out people who can't afford them. It's better that someone starting out in life be able to buy a house with smaller rooms, than not be able to buy anything at all. We've become so obsessed with raising standards across the board that we've forgotten that the people we exclude from those standards end up with worse standards.

Also, not everyone has the same priorities. Some people might like their bedroom to be huge. Others might just see it as a place to sleep and would rather spend their money on living in a good location, or some other priority. Restrictive zoning laws remove options by making buildings converge on a limited range of designs.

> Reading AI output is extra effort. It's verbose, frequently contains all too plausible nonsense, and is increasingly jargon dense. I recently got this sentence from Claude: "NATS control-plane events: stream leader election / R3 quorum re-form during pod churn." Jesus. I had to lookup almost every word to make sense of this.

I'm always puzzled when I see this kind of commentary. How are people using LLMs that they find this type of thing problematic? Just do:

- "It's verbose" -> prompt "Explain briefly"

- "plausible nonsense" -> Yes, it happens, but less frequently than with older models, and arguably far less than your average human. Mitigate by prompting "Run a search to check that X, Y, Z is correct".

- "I had to lookup almost every word to make sense of this." -> Does this really happen frequently? It doesn't to me. Mitigate by prompting "Breakdown that sentence and explain each term in plain English".

It's true that the Nix language is quite difficult to learn. However, part of that is the poor documentation, and the weird situation where flakes are the de facto modern way of doing things, but are still treated as an "experimental" feature (which leads to further confusion when reading documentation based on the old way of doing things).

Luckily LLMs can help with this too. I asked ChatGPT to generate a course with just enough content to help an ordinary NixOS user configure their system, work with Nix code and flakes, and create/modify simple packages (which is rarely needed anyway). Just a few hours of that gave me a much stronger understanding than all the documentation / tutorials I had read before.

Yes, I agree. I think you misunderstood me though. I wasn't saying refrigerators aren't an attack vector. I was using them as a comparison to make the point that a policy of blocking internet access is far less effective on an autonomous robot with vision and hearing (since it can easily communicate in other ways) than for a refrigerator (where the policy actually has a good chance of disabling the attack).

You need something more than just disabling internet access if you don't want robots sourced from military rivals to remain a strong risk.

Removing an internet connection doesn't really solve much when you're talking about autonomous AI-driven robots with vision and hearing. It's a whole different category to toys, refrigerators, etc. My refrigerator isn't going to suddenly pick up a weapon and start attacking things on command. Nor does it have an easy way to receive instructions through side channels. I'm not saying I agree with Trump's policy - but removing internet isn't a solution either.

You've convinced me enough to give it a try.

I think I'd still maintain a conceptual difference between sites that are source material (e.g. documentation) or apps (e.g. Google Translate), and those that are just for browsing (e.g. news, blogs) and are more ephemeral (I may read it now, or keep it for later, but either way I'll be done with it).

You'd open a PDF, Word document, or app in its own window which you'd manage with your DE / WM. To me it feels like a bit of an anti-pattern to have some of this type of content handled by my WM, and others by a tree-style tab manager, when they are the same "type" of content. I want to be able to press the same keyboard shortcut to switch to my documentation whether it happens to be a PDF or a website.

On the other hand, there isn't anything else really equivalent to browsing-type content, so the tree-style tab manger feels more natural to me there.

Do you really need hundreds of open tabs though? To me this suggests a workflow issue rather than a tab management problem.

I only keep tabs open that I'm actively working on right now (or am imminently about to switch back to). I regularly close tabs (and windows for that matter) that are only open because I forgot about them, or that I might use again but not until hours or days later. If I need them later I just open them again.

For me, having hundreds of things open at once just creates clutter that makes it harder to focus.

I'd never heard of Math Academy. It seems interesting, but quite expensive ($49/month for software which is in beta). Would you say it's significantly superior to something like Khan Academy (which is free) to justify paying that much?

So you're fine with your code having unintended behaviour, as long as that unintended behaviour passes a test that the agent wrote?

My preferred approach is to read and understand everything the LLM produces AND have it create test suites (which I also read and understand). The LLM can help you with that too - just have it breakdown and explain the code at each iteration.