stult

Sure, ultra-hazardous activities are regulated differently from other activities, including under tort law, but generic AI tools are not ultra-hazardous by nature. No piece of software is, until it is connected in some way to real world effects. Take an object-detection algorithm. There's absolutely nothing inherently dangerous about identifying objects in a video stream. But once you use the algorithm to create an automatic targeting system for a drone with a grenade strapped to it, it does become hazardous. But that's no reason to regulate the algorithm as if it were hazardous itself, at least no more so than it is to regulate the drone. As you point out, we regulate hand grenades. We do not regulate the boxes hand grenades are delivered in, or the web framework used for building a website that can be used to purchase hand grenades.

> Preload a city's, county's etc. entire set of laws

You would also need to load an enormous amount of precedential case law, at least in the US and other common law jurisdictions. Synthesizing case law into rules of law applicable to a specific case requires complex analysis that is frequently sensitive to details of the factual context, where LLMs' lack of common sense can lead it to make false conclusions, particularly in situations where the available, on-point case law is thin on the ground and as a result directly analogous cases are not available.

I don't see the utility at the current performance level of LLMs, though, as the OP article seems to confirm. LLMs may excel in restating or summarizing black letter or well-established law under narrow circumstances, but that's a vanishingly small percentage of the actual work involved in practicing law. Most cases are unremarkable, and the lawyers and judges involved do not need to conduct any research that would require something like consulting an AI assistant to resolve all the important questions. It's just routine, there's nothing special about any given DUI case, for example. Where actual research is required, the question is typically extremely nuanced, and that is precisely where LLMs tend to struggle the most to produce useful outputs. LLMs are also unlikely to identify such issues, because they are issues for which sufficient precedent does not exist and therefore the LLM will by definition have to engage in extrapolational, creative analysis rather than simply reproducing ideas or language from its training set.

Hence C laws

None of the positions seem to actually be remote, so this posting is deceptive. Really freaking annoying actually.

It would help if the legal profession (at least in the US) did not actively select against numeracy. The running joke in law school is that everyone who was a decent student but bad at math goes to law school because the LSATs do not have a math component, unlike the GREs. I’m an engineer and former lawyer, and when I was in law school I actually ended up taking a really disproportionate number of tax law classes because I was so much better than my average peer at basic arithmetic that I could always beat the curve in classes that required basic calculations. And we’re not talking about even intro level algebra, just basic arithmetic. Addition, subtraction, some multiplication and division.

By ignoring math skills altogether, the legal profession necessarily ends up selecting for individuals who on average have a poorer understanding of anything quantitative or statistical, and often that also correlates with poorer understanding of all technical and scientific matters, because the same innumeracy means those individuals don’t make it far any sciences at all, even in their secondary or tertiary education. So they often don’t even have strong fundamentals on the basic concepts in those areas.

As an engineer, it was maddening to witness and absolutely pervades every aspect of the US legal system. Which both translates into reaching false factual conclusions as in OP, but also infects legal reasoning with a ridiculous lack of rigor (a point which lawyers often dispute vigorously, but they have no concept of just how much rigor they lose by eschewing mathematical and especially statistical reasoning), and it also translates into poorer operational efficiency and administrative efficiency in the legal system, and poorer outcomes from the regulatory system, which, while often informed by technical concerns, remains a primarily legal- and lawyer- driven process. That opens up regulatory bodies to subjective political concerns where there is absolutely no reasonable technical case to be made. This phenomenon is most obvious in recent history with climate change related regulations, but is pervasive throughout the US policy making establishment, and contributes substantially to our challenges with special interests and regulatory capture.

The US has developed both low-cost interceptors like you suggest and Directed Energy Weapons for a counter-UAS role.

https://www.army.mil/article/268403/army_joint_force_work_to...

DEWs are likely to win out in the long term (at least for defensive purposes) because the cost per interception is extremely low when compared even to the price for the 30mm autocannon ammo used by Gepard and similar SPAAGs, never mind when compared to the price of a guided drone or missile. DEWs are also capable of intercepting a broader range of targets, including mortar and artillery rounds. With increased power levels, DEWs should be able to defeat ballistic and hypersonic missiles as well. The US DoD has already developed DEWs more than capable of the 50kW power levels required to destroy a small-UAS like an Orlan and is in the process of integrating them with Strykers and other armored platforms so they can be deployed en masse, but these systems are still in development and thus have not yet been made available to the Ukrainians. e.g., for a system that is already production ready: https://www.rtx.com/raytheon/what-we-do/integrated-air-and-m...

But that's why you don't hear much about re-establishing SPAAG production lines or scaling up interceptor drone production. Everyone knows something cheaper and more capable is coming very soon. Interceptor drones may play a role in clearing out enemy-controlled airspace prior to an offensive action when there is a need to destroy drones that are out of line of sight (although I expect DEWs mounted on attack helicopters or large drones to fulfill that role longer term), but defensively nothing will be able to compete with DEWs on cost and capability in the very near future.

> Of course the only way to win would be by pushing your enemy back down the stairs and then with whoever is left above you, retake the entire castle.

> Yeah, that’s unlikely to work out.

> Frankly, if you find yourself in this position the castle is probably already lost.

This argument directly contradicts the instructions of Vegetius, whose De Re Militari was *the* definitive military manual of the Middle Ages. Specifically, in book 4, he writes,

> Innumerable instances are to be met with where the enemy were entirely cut in pieces, even after they had penetrated the body of the place: this certainly will happen, if the besieged continue in possession of the ramparts, towers, and highest parts of the city... The sole resource after a place is forced, either by day or night, is however to secure the ramparts, towers, and all the highest places, and to dispute every inch of ground with the enemy as they advance through the streets.

It's hard to overstate how important Vegetius was to Medieval strategists, who treated his work practically as gospel and would generally try to follow its recommendations to the letter. So I seriously doubt any medieval strategist would have viewed the only conduit to the critical and decisive high ground positions as tactically irrelevant.

> For your particular usecase, yes.

The use case they are describing is a standard web server or web application. That's a pretty important and widely applicable use case to dismiss out of hand as "your particular usecase".

IME there is a code quality/sanity valley of death between about 50-1000 people, with the lower bound shifting upwards a bit depending on the maturity of the company's founding/early stage technical contributors. You get more bang for your buck with five of the right people than 50 of the wrong ones in the early stages, and sometimes that determines five years later whether you are drowning in tech debt (bad) or 1000 years ahead of the competition (inefficient, probably) or maybe just barely staying one step ahead of the reaper (optimal, probably). But a lot of companies limp into the valley, get enough revenue to stay alive for a while, and then slowly fade into obscurity the second a nimbler competitor shows up on the scene and eats their lunch. That's why small companies are fun. All that visible impact and you know you won't be mired in the bureaucratic death spiral of a medium-sized company lurching into obscurity!

What an incredibly, idiotically reductionist interpretation of cherry-picked data this is. First of all, he selected a fairly random, non-representative set of goods and services. I can immediately think of a glaringly obvious counter-example that his dataset excludes: energy. It's one of the most highly regulated industries in the US and yet energy inflation has been barely positive over the last 20 years (https://tradingeconomics.com/united-states/energy-inflation), completely contradicting Andreesen's hypothesis. So basically all of his conclusions are unsubstantiated and likely false on that basis alone. But there's more stupidity here.

He also does not define "heavily regulated" or how one can measure the level of regulation applicable to a given good or service in a way that allows for apples-to-apples comparisons to other goods and services. But even just taking a colloquial understanding of what the term means, car manufacturing is incredibly highly regulated. Perhaps not so much as medical services, but certainly far more than household furnishings or clothing. And yet the inflation curves for those goods are nearly identical to the one for new cars. If regulation is the sole explanation for the differing inflation trajectories between each of these goods and services, then why is such a highly regulated good performing identically to substantially less regulated goods?

There are many more flaws in his thinking but I won't bother digging into it because this is a childishly stupid article that would never have made it to the front page of HN if it weren't written by an especially wealthy idiot.

> Not rich enough to fly charter

The growing backlash against the environmental impact of private jets may change this calculation a bit.

> Another example is when our rights were violated the lawyer we talked to said it was a violation but there's not a good case because the judges only care if there was significant financial or physical injury involved.

It’s not that judges don’t care, it’s that courts are there to provide a remedy where one party has injured another. If there is no actual injury (damages in legal parlance), then the court has no power to provide a remedy.

Although just practically speaking oftentimes in these situations the expense of a lawsuit dramatically outweighs any potential compensation. Judges tend to view those cases as a waste of their time because the plaintiff is usually just trying to make a point or to punish the defendant rather than seeking meaningful compensation. It’s basically lighting a big pile of cash on fire to make someone feel better. Most judges believe the courts aren’t there to act as platform for grand standing or moralizing or punishing wrongdoers (at least in civil suits without punitive damages). The court system is already overloaded just dealing with more substantive cases where injured people need to be made whole in a significant way.

> Not drugs or mental illness, study says

Meanwhile, in the actual body of the article,

> Aldern, a data scientist and policy analyst in Seattle, and Colburn, an assistant professor of real estate at the University of Washington’s College of Built Environments, said they are not suggesting that mental illness, addictions and other issues are not contributing factors to homelessness.

Sometimes I wonder if newspaper editors actually bother to read the articles anymore

> Not even sure what a formal education in this context would entail.

An MBA. This is exactly the sort of material MBAs study and there really isn't any other formal educational credential that qualifies someone to evaluate strategies for making a business succeed. Making this particular ad hominem criticism especially ridiculous.

I think there is a mismatch between the types of web content that the general public most interacts with and the types of web content which tend to have lots of development resources devoted to them. The ratio of SWEs to users is much, much higher for niche (typically B2B or internal business line) applications, which often require a fair amount of interactivity and custom logic. Whereas the vast majority of random web content used by the general public probably can just be hosted as static sites or server-side rendered apps with better performance and lower cost. HN then just reflects the underlying demographics of the industry and has a disproportionate number of people who work on niche applications where SPAs are still the more appropriate tool.

Which makes sense given the economics. Consumer facing content is generally free to users and funded by ads so the dev costs need to be amortized across more content, which tends toward a lower touch, more scalable approach like static pages where there isn’t much if any custom logic on any given page of content. Businesses on the other hand are willing to fork over cash directly to solve their problems so you get higher touch, more custom solutions where SPAs are relatively more useful.

Also software engineering by its very nature automates routine tasks, reducing their costs and manual complexity dramatically over time to the point that a SWE isn’t needed to carry them out. So none of us work on static sites because at this point it doesn’t take a highly skilled engineer to set up, host, and scale one. One day SPAs and cloud native and k8s and all the latest fads may reach the same level of simplicity, and we will all be working on tuning hyper parameters on ML models or whatever else becomes the de facto cutting edge for the field.

I think the big mismatch here is that romantic desirability is largely determined not by someone's subconscious assessment of their future happiness with a potential partner but rather their perception that the potential partner would help them produce successful offspring. A conventionally attractive partner might not make you happier, but they are more likely to be healthy and therefore to pass on healthy genes to your children.

But over the last 50ish years in the west, we've begun to expect more from our romantic relationships, including many different, potentially inconsistent objectives beyond just the production and raising of children. We want friendship, intimacy, meaning, happiness, etc. And romantic relationships where the production of children is not a goal, and in some cases is not even possible, are increasingly common too (eg many--but not all--LGBTQ+ relationships, child free hetero marriages, and so on).

Not that this is a bad thing, but on some level we are still monkeys driven by the need to procreate and we are following those same subconscious instincts even when they are inconsistent with our conscious desires to form long-term happy relationships.

We don't have that phrase in English but it translates very clearly. Interestingly, your description of Orban's behavior exactly matches a somewhat famous game theoretic analysis of international leaders' behavior called "the two-level game." https://en.wikipedia.org/wiki/Two-level_game_theory

He's not unique in this regard. Every national leader has to consider the interests and perceptions of their domestic power base when negotiating with international leaders. Very often a mismatch between their rhetoric and actual behavior indicates that they are trying to sell an unpopular but necessary international agreement to the people back home, or that they must accept an international reality that does not match their domestic political framework.

For example, Putin finds himself in a difficult position right now because he has convinced his people that the Russian military is invincible and the Ukrainians are dominated by a small number of Nazis who need to be removed from power. Meaning, his domestically acceptable "win set" only includes scenarios in which Ukraine cedes large swathes of territory and makes at least some notional concessions around governmental reform toward "de-nazification." Anything else admits the political framework he uses to justify his power is premised on falsehoods and risks destabilizing his regime.

These positions are of course not consistent with reality and therefore not within the Ukrainian win set, and so no agreement will be reached until something changes. I suspect that is why Russian journalists/propagandists have recently started recasting the war as a conflict against NATO, because that will make defeat more palatable to the Russian people. Getting beat by NATO doesn't mean the Russian military was horrifically incompetent and their government deeply corrupt, but rather that it was an unfair fight sprung upon them by evil westerners. It feeds their sense of grievance while excusing Putin and his lackeys from any personal responsibility for the widespread military failures.

I mentioned cannons as the absolute worst case option in case US air-to-air missiles suffer from the same issues as Russian SAMs appear to, and from the perspective of extant US capabilities that could be adapted and rapidly fielded for immediate use. My ultimate point being that we're just way better positioned to handle these drones right now and into the near future than the Russians are or will be.

It's a very solvable problem though. At least for countries other than Russia, which might not be technically nimble enough to adapt mid-conflict and under a strong sanctions regime. S-300s/S-400s/Buks and the somewhat rarer western equivalents like the Patriot system just aren't designed to target something flying that slow with that small of a radar cross section. That doesn't mean the drones aren't reasonably easy to detect and target, just that the targeting systems need to be fine-tuned a little.

I doubt NATO would have this problem, not only because Turkey is part of NATO but because US tactics would probably be more effective against similar drones. We rely on fighters to establish air superiority more than SAMs, while the Russians do the opposite. Worst case scenario, F-15s/16s/22s/35s can use cannons and pilots are more adaptable than radar systems. I also suspect that the F-35's sensors at least are far superior to whatever the Russians are relying on.

I am also somewhat suspicious of the reports we have heard about them striking deep within Russian territory. Not because I doubt their veracity, but because I am certain we are not getting all of the relevant information. I wouldn't be surprised to find out if there was some supplementary SEAD (suppression of enemy air defense) going on that Ukraine and its western allies are remaining tight lipped about. It wouldn't shock me to find out that NATO has some clever trick for blinding Russian SAM radar in a way that the Russians haven't figured out yet. The Russians would never ever admit the existence of such a capability publicly because it would totally destroy demand for their military exports, and NATO wouldn't want to tip their hand in anyway that would give the Russians an opportunity to adapt.

> I'd have to imagine it's more cost effective (in terms of kilograms of high explosive on target per dollar) to make some GPS-guided or laser-guided mortars

The problem is scale. No one in the private sector makes or needs mortar rounds, so the DoD doesn't benefit from buying COTS components for them. Whereas you can pretty much just strap a grenade to a quadcopter with a few minor tweaks and you're in business. Which is more or less exactly what the Ukrainians have been doing with their surplus Soviet anti-tank grenades.

Yes indeed, many of us do. Nearly everyone I've ever dated gets mad at me at some point because I can't keep cookies in the house or I will eat them all. You are far from alone, and in fact I suspect the overwhelming majority of people are the same. Hence the obesity epidemic.

Personally I find that my self-control is limited to the same extent that my time and energy are limited. If I'm tired, it's harder for me to be disciplined, and the place that I most often tired is at home at the end of the day, when temptation to over eat is at the greatest. I believe there is a fairly substantial body of evidence that shows I am not unusual in this regard. This phenomenon is why groceries stores put candy in checkout lanes, when you are mentally tired from exercising discipline by not picking up the delicious-looking tray of cupcakes in the bakery aisle and thus are most susceptible to temptation.

I think some people can just use willpower to maintain their discipline in every situation, but those people are extreme outliers. Like Navy SEALs or Olympic athletes or other extreme high performers, and not even all of them. Those who can are experts at delayed gratification and are able to visualize what they want and then never deviate from their plan for how to achieve it, regardless of their current emotional state or energy levels. But that is vanishingly rare and often their behavior is a result of some deep trauma (eg someone who is super disciplined about their diet because their dad died early from an obesity-related illness) rather than a positive focus on a certain outcome. The truth is that the vast majority of people just follow their natural inclinations and do whatever is easiest for them. The path of least resistance.

As one such weak-willed person, my approach has been to structure my life to remove temptations, much as the GP comment above described their approach to curating their internet content. For example, I have lived almost exclusively in big cities for my entire life until the last year, when I moved to the middle of nowhere. Which means I can't get takeout food delivered and it's much harder to go out and get ice cream on a whim. Compared to living in a major city where I could get Michelin star restaurant food delivered in a matter of minutes, the temptation is just much, much less powerful. I order my groceries online too, which also dramatically reduces the tendency to impulse buy anything unhealthy (I get to skip the checkout lane with the candy, for one thing). On the other side of the caloric equation, I set up a home gym so that travel time to the gym doesn't become an excuse not to exercise. Basically I add barriers for bad choices and remove them for good ones. I am in the best shape of my life as a result.

I guess what I'm saying is that most people achieve good behaviors in part because of their intentional discipline and in part because their environment is conducive to those behaviors. The real trick is figuring out how to intentionally structure your environment to match whatever behaviors you want. It's effectively borrowing motivation from your peak-energy level periods to provide discipline during periods when you don't have the energy to generate it yourself. It can be incredibly hard to figure out, because you need to know how to change your environment and how you will actually react to that new environment, which isn't always predictable.

That is such an absurdly specific but accurate description. I used to manage a site that had a lot of geospatial elements, including various custom maps and obscure (outside of the GIS world) data formats like geotiffs or KML or DTED or the WMM. Coming from a non-GIS background and without the need for an expensive solution like ArcGIS, I have no idea how I would have sanity checked any of it without QGIS

Cool new business model, harassment as a service. We‘ll get your info off google for the low low price of $1.99 per fake death threat

Faces are helpful for communicating around emotionally nuanced issues, which is usually not part of routine communications. So if it's a potentially emotionally charged conversation, maybe that should be the signal it's time to require cameras on.

I’ve had a very similar experience and my solution was to incrementally move into more and more math intensive jobs, starting from regular old SWE working on a web app to working on an aerospace-related web app that involved lots of physics and geospatial calculations and then moving toward MLE/DS jobs. All without a STEM degree, teaching myself the math as I go. It hasn’t been easy but I enjoy what I do more and more over time.

There was just an article from deepmind on HN about this topic the other day[1], but basically IIRC it argues that all of the LLMs are horrendously compute inefficient, which means there’s a ton of room to improve them. So those models will be optimized over time just as the consumer hardware will be improved until eventually one day the two trends will converge. It’s just a question of when that will happen.

[1] https://news.ycombinator.com/item?id=30987885

This touches on the controversy that sparked Coase's Theory of the Firm. He proposed the idea that firms exist to reduce transaction costs as an explanation for why capitalist firms look like mini-soviets internally, with highly centralized control and planning. Which seems like a contradiction to classical economists. If free markets are so effective, why can't you run a company like a free market? eg have workers bid for tasks from managers. Turns out that is wildly inefficient because of the transaction costs involved in coordinating competitively instead of cooperatively (at least in successful companies). I suspect pandemic WFH has revealed there are much more efficient ways to coordinate cooperatively than simple centralized planning, and those methods generally do not require as much managerial facilitation because they rely much more heavily on asynchronous communication.

It’s a distinction without a difference. For human purposes, nukes can effectively destroy us completely. We don’t need to also physically rip the planet apart and destroy all the single celled organisms to ruin the planet for ourselves.

Surpass in destructive power doesn’t mean much since we can already destroy all life on the planet with nukes. That’s what makes the doomsday weapon in Dr. Strangelove so ironic and funny. But if a new weapon type that at least approximates nukes in destructive power while also surpassing them in ease of proliferation, cost of production, or detectability? Those are much scarier propositions.

I don’t think any of the irreplaceable uses of plastics are really major contributors to the problem. eg if syringes were the only plastic item around, this would not be an issue simply because there aren’t that many syringes and they aren’t generally reused (and so don’t shed plastics as much as water bottles with tops being screwed on and off frequently). It’s the common, day-to-day plastic items like food containers and synthetic clothes. Those are hardly irreplaceable use cases, although obviously other solutions are going to be more expensive than plastics.