Yes, the manufacturer of the self-driving system. People have been asking this silly question for a decade now, even though there is no other answer. Google, at least, has stated publicly on many occasions that they are liable for the actions of their self-driving vehicles.

No, they did not. In fact they said exactly the opposite, that the cars are never remotely-piloted. They said that the cars occasionally request guidance, which means something like "Should I go this way or that way?", then the car acts on the answer.

I quit using Windows in 2000, and it's been at least a decade, if not more, since I ran into anything like that.

From an anarcho-capitalist and Austrian perspective, the attack of these subjects is severe. The core criticism is that welfare economics often dresses political value judgments in mathematics and presents the result as scientific optimization.
The Austrian objection starts earlier than the libertarian one.
There is no measurable quantity called âoesocial welfare.â
You value a steak dinner. I value the money more. We trade. Both reveal, through action, that each prefers the new situation.
Economics observes the voluntary exchange.
Now a welfare economist writes something like:
Social welfare = Aliceâ(TM)s utility + Bobâ(TM)s utility + Charlieâ(TM)s utility.
The Austrian response is: what units are you adding?
Utility is ordinal. You prefer A to B. This does not mean your satisfaction is 17.3 units and mine is 12.8 units.
You cannot scientifically establish that taking $1,000 from one person causes less lost utility than giving the money to another person creates.
You can support redistribution as a moral or political position. But calling the resulting calculation a social welfare function does not transform the moral judgment into an objective measurement.
âoeSociety choosesâ is dangerous language.
Individuals choose. Individuals act. Individuals own things. Individuals bear costs.
âoeSociety decided to spend $10 billionâ usually means a political process selected an expenditure and taxpayers were compelled to finance it.
From an anarcho-capitalist perspective, aggregating millions of people into a fictional decision-maker hides the essential question:
Who decided?
Who pays?
Who benefits?
Who refused?
What happens to the person who says no?
Social choice theory deserves some credit here. Its own results expose serious problems with turning individual preferences into a coherent âoewill of the people.â
The Condorcet paradox shows that majority preferences can cycle. Arrowâ(TM)s theorem shows that no general ranking system satisfies several attractive conditions simultaneously.
The anarcho-capitalist reaction is almost sarcastic: you spent decades proving mathematically that there is no coherent social preference ordering, then continued discussing how experts should optimize social welfare.
Pareto efficiency is much narrower than political rhetoric suggests.
Pareto efficiency has a legitimate analytical meaning. The trouble starts when economists move from voluntary exchange to hypothetical compensation.
Suppose a regulation gives Group A benefits economists estimate at $100 million and imposes costs of $60 million on Group B.
Some welfare analysis says the policy produces a $40 million net social gain.
The libertarian response is simple: Group B lost $60 million. Did anyone ask them?
If A gains $100 and B loses $60, saying âoesociety gained $40â treats separate people as entries in one accounting ledger.
An anarcho-capitalist rejects the premise. A benefit to one person does not cancel an imposed loss on another person merely because an economist performs subtraction.
The knowledge problem destroys the fantasy of optimization.
This is the Austrian argument associated especially with Friedrich Hayek.
Economic knowledge is dispersed. Prices contain information produced by millions of independent decisions. Preferences change. Local circumstances change. Resources have competing uses.
A central analyst does not possess the information needed to calculate the âoeoptimalâ allocation.
A market does not require one person to know everything. Prices coordinate plans without a central mind directing the entire system.
Welfare economics often asks, âoeWhat allocation maximizes welfare?â
The Austrian response is, âoeYou do not know the relevant preferences, opportunity costs, entrepreneurial discoveries, future alternatives, or counterfactual prices required to answer your own question.â
The calculation problem is worse without genuine market prices.
This is the argument strongly associated with Ludwig von Mises.
A bureaucrat deciding whether resources should produce railways, hospitals, housing, batteries, or server farms needs meaningful prices for capital goods.
Those prices emerge from exchange, private ownership, profit, and loss.
Without genuine market pricing, planners are not optimizing. They are allocating according to administrative rules, political pressure, historical budgets, lobbying, and guesswork.
A spreadsheet does not solve the economic calculation problem. More computing power does not solve a missing-price problem.
âoeMarket failureâ analysis often compares reality with an imaginary perfect market.
This is one of the strongest libertarian criticisms.
The usual pattern is:
Real markets have imperfect information, transaction costs, externalities, monopolistic tendencies, and unequal outcomes.
Therefore, government intervention might improve the result.
The missing step is institutional comparison.
Government officials also have imperfect information. Regulation has compliance costs. Voters are rationally ignorant. Agencies seek larger budgets. Politicians respond to concentrated interest groups. Regulations create unintended consequences.
The relevant comparison is not:
imperfect market versus perfect government.
The relevant comparison is:
imperfect market process versus imperfect political process.
Once you make that comparison, many clean textbook conclusions become much weaker.
Social choice theory accidentally supplies ammunition to libertarians.
This field is less inherently collectivist than welfare economics.
Social choice theory demonstrates that collective decision mechanisms have deep structural problems.
Majority rule can cycle.
Agenda setters can influence outcomes.
Strategic voting changes results.
Different voting systems produce different winners from the same underlying preferences.
There is no neutral mechanism for converting individual rankings into a single collective preference under all desirable conditions.
A libertarian conclusion follows naturally: if collective choice is structurally problematic, reduce the number of decisions imposed collectively. Leave more decisions with individuals, families, firms, voluntary associations, insurers, cooperatives, charities, and contractual communities.
Social choice theory often asks, âoeHow should everyone collectively choose one option?â
The anarcho-capitalist asks, âoeWhy must everyone choose the same option?â
That question cuts much deeper.
The deepest conflict concerns consent.
Mainstream welfare economics often focuses on outcomes.
Anarcho-capitalism focuses heavily on means.
Suppose forced redistribution produces a statistical improvement under some selected welfare function. The anarcho-capitalist still asks whether coercion became legitimate because an economist assigned weights to different people's utility.
From this perspective, the central problem with much welfare economics is not bad arithmetic. The problem is a category error.
Economics studies choices, scarcity, exchange, prices, production, and consequences.
The moment an economist says, âoeThis distribution is socially better,â a moral judgment has entered the analysis. The economist should identify the ethical assumptions instead of presenting them as a technical output.
The harsh Austrian verdict would be this:
Welfare economics starts with subjective individual preferences, admits they cannot be directly measured or meaningfully added across people, constructs a mathematical social welfare function anyway, inserts political judgments into its parameters, and then announces an âoeoptimalâ social outcome.
Social choice theory is more interesting because much of its best work demonstrates why the phrase âoesociety prefers Xâ is often logically unstable.
The anarcho-capitalist alternative is less ambitious and more disciplined: voluntary exchange, private property, freedom of association, decentralized decision-making, and liability for harms. Instead of trying to maximize an imaginary aggregate welfare number, allow people to pursue different goals and coordinate through consent.
The strongest criticism is not âoeall welfare economists are socialists.â Many are not. The stronger accusation is that the framework gives technocratic politics a scientific-looking vocabulary. Words such as optimization, social welfare, efficiency, and compensation criteria often conceal the real political questions: whose property is taken, who decides, who pays, and whether refusal is permitted.

Years back I was interviewing people for a coding position. We went through the standard tech stuff and then did a bit of project to see how they thought. We said (this is circa 2009'ish I think) - imagine you're on a team creating a new phone. You don't have time to test all the functions, so which would be your top two functions to ensure working?

All a bit Kobayashi Maru - obviously you can't release a phone testing only two functions, but we wanted to see what they'd prioritise. The very best answer we received was this one: "I would make sure it has the ability to call emergency services." Their thinking was that this was likely the most critical feature of a phone for both a user, and also for the manufacturer to avoid being sued. Absolutely great answer.

And yet here we are, with the post above. Taking the thinking of this interviewee - the ability to work with emergency services is important for general society, for the user of the vehicle (so they don't get in trouble) and for the manufacture of the vehicle (so they don't get fined/sued/both). Absolutely critical.

Dude, you're being ridiculous.

That's clearly not the intention here if the end result of passing through the luxury terminal is boarding the same airplane as those masses, and it is. It's obviously just to make the airport part of the travel experience nicer, in ways that would be too expensive to apply to the regular terminal. It's the same thing as airport lounges (I'm a Delta Sky Club member myself, a privilege I pay money for so I have the option of a nicer place to wait, the availability of hot showers on long trips, food, drinks, etc.) just scaled up to cover the whole airport process... right up to boarding time when the people get shuttled to board with everyone else.

LOL. I specifically mentioned that I use correct capitalization... so of course I screwed up the capitalization.

Tiny, tiny minority. And if you think you want that, you're wrong!

Also, it's worth pointing out that finding matching pages in a database of pages is indeed trivial -- and building that is utterly insufficient, because for any query that trivial matching algorithm will return a huge number of pages. Thousands, even for the most obscure technical terms, millions or tens of millions for more-common words.

The hard part of building a web search engine (and it's very, very hard) is ranking the results once you've found them, so the thing the user wants is on top. That was, in fact, Google's big innovation: PageRank was Larry Page's idea for how to rank pages by examining the link structure of the web and prioritizing pages with more inbound links. That specific mechanism quickly broke down when SEO companies began exploiting its structure, but in addition to being gameable, PageRank had another problem: What if the search terms are used in multiple domains? The classic example is the query "python spacing". Am I looking for information about how large an enclosure I need for a captive python, or am I asking about indentation in programming?

So Google, and every other competent search engine, has shifted towards supporting queries in natural language, as well as using contextual information when available, such as the user's search history -- in the "python spacing" example, unless the user is a zookeeper who also writes code, their search history will point to the correct domain.

If you're writing queries as lists of terms that you want matched in pages you're doing it wrong. You'd actually be unhappy with a search engine that gave you exactly that, and you're also artificially reducing the effectiveness of the much better search engine you're using. Try typing questions instead, e.g. "How much space does a 10 foot Python need?" (correct spelling, capitalization and punctuation are not really required, but I use them anyway). This will give the engine more contextual clues about what kind of thing you're actually looking for and you'll get better results.

That said, it should be pointed out that if what you really, really want is "include only these words, exactly as I spelled them", Google will give you that. Just put them in quotation marks.

It's too bad they don't provide numbers, because the numbers are incredible. I occasionally checked the search qps numbers when I worked at Google, just for fun, and... wow. Say what you will about Google, their scale is incredible. The services work so reliably and quickly that you don't often think about what the infrastructure must be like to handle it -- and you can't achieve that kind of scale just by throwing hardware at the problem, either (though lots of hardware is required, obviously). Every layer of the stack is finely-tuned for performance, with both macro optimizations like sophisticated distributed consensus-based eventually-consistent storage and micro optimizations like libraries that squeeze maximum value out of every cycle.

Supporting tens of millions (maybe hundreds of millions now?) of queries per second against a multi-petabyte (maybe exabyte now?) database is an incredible feat of engineering, as is keeping the whole system humming along with near-perfect reliability. There are a lot of damned fine engineers at Google, and "engineer" is absolutely the right word when you're talking about global-scale infrastructure.

One of my first "Google-scale" moments was shortly after I joined in 2011. The global data center status pages had a bug, which was that the field that displayed the aggregate on-line storage (basically all spinning Rust back then, I think; the SSD transition was just about to get under way), was a Java long, a signed 64-bit integer, and it had just wrapped; Google's online storage had exceeded 2^63 bytes. That is a big number. They just updated the code to use a BigInteger instead.

No, but they'll record the punch, then call the police. The recording will be evidence in the subsequent prosecution for assault and battery.

All of the big tech companies incentivize their employees to write up as many patent ideas as possible, and anything that looks remotely plausible gets filed by their patent attorneys. This in no way means the company has any plans to build the thing patented.

Why do they do it then? To build up their "patent warchest". Every company knows that they're going to get sued for patent infringement, because it's just impossible not to. Hamstringing your engineers by having patent attorneys scrutinize everything they're doing to see if it happens to be somewhat close to a thing that someone else patented will drive everyone nuts and drive productivity to zero. And you definitely don't want your engineers searching the patent database themselves... if they stumble across something vaguely close to what they're building or thinking about building, then go ahead with what they were going to do anyway, it's now "willful infringement", eligible for treble damages.

Instead, companies file for patents on everything possible, mostly with no plan whatsoever to build it, so that when they get sued they can then dig through their warchest to find something vaguely related to what the plaintiff builds/does (assuming the plaintiff is a company that builds stuff; patent trolls are a separate issue and require a different strategy), then threaten a countersuit. Then the lawyers get together and craft a cross-licensing agreement, which rarely includes any money changing hands or anyone changing what they're building and selling.

Patent filings like this one aren't news, they're noise, and /.'s editors should know better.

Likely bots.

Spot on.

What happens when people take someone's "cute OOTD" pic and put them in another outfit without any consent? That kind of stuff will get the young people off of Insta and looking back at Snap or VSCO others.

What happens when someone swaps a paid influencer's merch for another brand?

And that's not even touching the whole issue of non-consensual sexual content.

And why exactly do regular people who post like or want that content to be manipulated? What's the value to the primary content-making user?

The Zuckodei and Pinchmans of the world seem to think that people will make public content no matter what they do. I'm not naive enough to think they'll all migrate to Proton or privacy-preserving OSS stacks, but I do think even regular people have limited patience for having their thoughts, communications, and likenesses changed and monetized without benefit to them.

I doubt it. I'd bet it's more of a problem of feature parity with OS X. Even though Windows has backup and restore, it still gets ragged on for not handling device moves or replacements as nicely -- even though the main problem is users and/or app developers not turning it on. I'll bet that first MS offered the feature and left it to users to discover and enable, but hardly anyone did. Then they started nagging people to turn it on, and most still didn't. So now they're turning it on by default.

I saw this same story play out on Android (when I worked there), firsthand. Backup was added to provide feature parity with Apple and to reduce user complaints about lost data. But approximately no one turned it on. In Android there were actually two obstacles, because out of abundant concern for privacy the backup solution required both app developers and users to opt in. User opt-in improved when users were nagged to enable it at setup time, but then they were annoyed that when they restored a backup hardly anything was restored because hardly any app developers opted in. Eventually apps were opted in by default (app devs can still opt out, but most don't) and the feature became somewhat more functional -- except for apps that were broken by it because app developers didn't expect to have their app and its data suddenly appear on a different device.

I had a front row seat to this saga because the component I owned (Android Keystore) was a significant motivator for requiring apps to opt into backup. The problem is that Keystore's core purpose is to provide hardware-backed cryptographic key material which is permanently and irrevocably bound to the device (and is strongly guaranteed to be wiped on factory reset). When an app that uses Keystore keys has its data restored from backup and then tries to load one of its Keystore keys, it gets a null pointer. If the app developer didn't check for null, that turned into a NullPointerException, and if they didn't catch the NPE, their app crashed. Net result: after a restore, most banking and many other apps crashed on startup because one of the first things they do is try to use their keys to authenticate to their servers.

I designed a key backup and restore scheme but my scheme would require app developers to opt into key backup on a per-key basis, at key generation time, because adding any automatic backup/restore solution would have broken the fundamental security property of Keystore. The scheme was moderately complex because it tried very hard to make it impossible for Google to ever access the backed-up secrets, and it relied on some internal server infrastructure whose sole purpose was to make it possible for Google to store data it could not access. That infrastructure was expensive, fragile, high-maintenance and not much used so it was at risk of being turned down because the teams that used it couldn't afford to maintain it. Strongly authenticating the new device and the user before releasing secrets was also tricky. At the end of the day, I never launched the Keystore B&R scheme (though vestiges of it remain visible in the Keystore secure key import scheme, which was designed for B&R but is useful on its own. Specifically, the import format includes a "masking key" field that isn't really useful in the normal import flow, but was crucial to the scheme that kept the secrets impenetrable to Google).

At the end of the day, app developers eventually fixed their apps to deal with being restored, mostly, and forcing users to make a backup/no-backup decision during device setup generated high adoption. Android B&R works fairly well today (except where device OEMs screw it up), though people still complain that the whole system isn't as smooth or as seamless as iOS', I think mostly because Android apps can opt out, but iOS apps can't (AFAIK).

If you pay it off every month, it's not really debt, is it? You never pay any interest, and the only way it translates into actual debt is if you have more expenses than you can cover -- but in that case you'd be reaching for debt of some sort regardless. But with a CC you do get the fraud buffer (the fact that you haven't happened to need it, yet, doesn't mean anything), and you get to delay your payments by about one month, which IMO is really quite nice. I not only get a short interest-free loan, letting me keep my cash in an interest-bearing account (currently about 1%, not a lot but not nothing), but I also have the flexibility to decide which account or accounts to keep my cash in (I actually keep most of it in a couple of brokerage accounts, not a regular bank account; yes, it's still insured by FDIC).

Plus I get 1-5% cash back depending on where I use it.

I realize that The 2-6% I get back (rewards + delayed payment interest) is paid for out of the fees the merchant pays and passes on to us all. I think it would actually be better if merchants were allowed to itemize those fees on my bill, which would discourage use of high-fee cards and change the calculation. With the system as it is, though, I'll absolutely stick with my credit cards.