tal

You can get wrong answer with 100% token confidence, and correct one with 0.000001% confidence.

If everything that I've seen in the past has said that 1+1 is 4, then sure --- I'm going to say that 1+1 is 4. I will say that 1+1 is 4 and be confident in that.

But if I've seen multiple sources of information that state differing things --- say, half of the information that I've seen says that 1+1 is 4 and the other half says that 1+1 is 2, then I can expose that to the user.

I do think that Aceticon does raise a fair point, that fully capturing uncertainty probably needs a higher level of understanding than an LLM directly generating text from its knowledge store is going to have. For example, having many ways of phrasing a response will also reduce confidence in the response, even if both phrasings are semantically compatible. Being on the edge between saying that, oh...an object is "white" or "eggshell" will also reduce the confidence derived from token probability, even if the two responses are both semantically more-or-less identical in the context of the given conversation.

There's probably enough information available to an LLM to do heuristics as to whether two different sentences are semantically-equivalent, but you wouldn't be able to do that efficiently with a trivial change.

One major problem with very-long-data-retention formats is that the hardware to read the things may not be around in a surprisingly short period of time. Like, if you assume that this format isn't bumping up against fundamental physical limits, then it will probably be supplanted down the line by something else, and people will probably stop making the devices to work with them before long. The devices to work with the media won't last as long as the media, and there probably won't be new ones produced.

https://en.wikipedia.org/wiki/BBC_Domesday_Project#Concerns_over_electronic_preservation

The BBC Domesday Project was a partnership between Acorn Computers, Philips, Logica, and the BBC (with some funding from the European Commission's ESPRIT programme) to mark the 900th anniversary of the original Domesday Book, an 11th-century census of England. It has been cited as an example of digital obsolescence on account of the physical medium used for data storage.[1][2][3][4]
This new multimedia edition of Domesday was compiled between 1984 and 1986 and published in 1986.

In 2002, concerns emerged over the potential unreadablility of the discs as computers capable of reading the format became rare and drives capable of accessing the discs even rarer.[14][15] Aside from the difficulty of emulating the original code, a major issue was that the still images had been stored on the laserdisc as single-frame analogue video, which were overlaid by the computer system's graphical interface. The project had begun years before JPEG image compression and before truecolour computer video cards had become widely available.

I think that realistically, if you want to maintain something for very long-term archival use, it's probably going to need to be rolled over into a new format periodically.

I love all open source desktops equally

https://en.wikipedia.org/wiki/Common_Desktop_Environment

If the US federal government creates a website to let Europeans bypass restrictions on hate speech in Europe and some Europeans in countries that don't restrict porn create a website to let people in US states with state-level porn restrictions, I'd be pretty comfortable with the outcome.

Our last, best hope for the subsidy model was Valve, a company that famously rakes in money hand over fist and launched the original Steam Deck at the unbeatable price of $399 through a “painful” amount of subsidy. If Valve did the same for the upcoming Steam Machine, it could have legitimately competed with the PlayStation and Xbox for your living room TV.
But Valve has all but dashed those hopes through a series of moves. In late December, it discontinued the $399 Steam Deck, raising the starting price to $549. In early February, it announced that the Steam Machine had been delayed due to the memory shortage and that the company would have to reset expectations on pricing. And now, even the $549 Steam Deck OLED is out of stock specifically because of the memory crisis.

I was pretty confident that Valve was not going to subsidize the Steam Machine from the start, even before Valve said that it would be priced comparably to a PC and even before it said that it was delaying determining pricing (which was a good sign that it hadn't locked in a contract price on components). I commented along those lines here.

Consoles can do the razor-and-blades model because they are a closed platform. If you buy a Playstation, it doesn't do you much good unless you use it to buy Playstation games. So each Playstation purchase is very, very probably going to be used to purchase Playstation games. Sony can crank up prices on those and make their initial loss back.

But the Steam Machine is open. I can go run whatever on it. I can just take the thing and, say, make it a media server or whatever. And if Valve subsidizes it, people will just buy it instead of a comparable PC and then run whatever they want on it. Doesn't make much sense for Valve, just because of the nature of the machine.

fedinsfw.app

I am thinking that lemmynsfw.com remaining down is going to result in a bunch of new NSFW instances showing up.

[continued from parent]

The plans and measures for Army expansion to meet the crisis of 1940 were matched by a naval expansion program, designed to provide the United States with a "two-ocean" Navy that could cope simultaneously with Japanese naval power in the Pacific and with the naval power that Germany and Italy had or might acquire in the Atlantic. On 7 June, the Navy's General Board proposed a building program that would about double the existing strength of the Navy in combat vessels. Congress approved the program on 19 July, and by the fall of 1940 the Navy had begun construction on more vessels than it then had in actual service. Outside of the military services, mobilization called forth a host of new civilian agencies under the Advisory Commission to the Council of National Defense to supervise the gradual transformation of the national economy from a peacetime to a wartime basis.

Germany's occupation of Denmark raised immediate problems for the United States with respect to the future of Greenland and Iceland. The Danish colony of Greenland was completely unprepared to resist a German attack or occupation. Since Greenland was considered a part of the Western Hemisphere, the United States opposed its military occupation by British or Canadian forces; such an occupation might give the Germans an excuse to attack this northern flank of the hemisphere. At the same time, the United States Government was as yet unwilling to commit itself to protection of Greenland with its own forces. It limited its actions to opening a new consulate at Godthaab, the Greenland capital; to the establishment of a Greenland patrol by Coast Guard cutters; and to the sale of a small quantity of arms and ammunition to Greenland authorities to be used for protection of the cryolite mine at Ivigtut.

The planners assumed that Germany and Italy could not launch a major military attack against the Western Hemisphere until they had defeated Great Britain and gained naval control of the eastern Atlantic. It now appeared that British naval power based on the British Isles could be maintained at least for another six months. Even if the Axis Powers then gained control of the bulk of the British Fleet, it would take them six additional months to assimilate British naval strength and prepare it for offensive operations across the Atlantic. The United States, therefore, probably had at least a year's grace in which to complete its military preparations. By the end of that year (roughly, by October 1941), American mobilization under the long-range program was expected to produce the 1,400,000-man Army and enlarged Navy that would be strong enough to resist successfully any Old World military aggression against the New. During this year, too, the United States could afford to keep the bulk of its fleet in the Pacific to check Japan. On the other hand, if, as seemed increasingly probable, Japan should in the meantime strike southward in the western Pacific, the United States could not afford to commit a major portion of its naval strength in an effort to stop Japanese aggression. American naval power must be kept mobile, free to shift to the Atlantic to deal with any emergency that might arise there.

In the event, Nazi Germany never defeated the UK, and then attacked the Soviet Union instead of going after the Americas, and was at war with two major powers already when it declared war on the US, so those contingencies never really arose.

The US also intended to make use of Latin American labor to greatly increase its productive capacity, in that contest of productive capacity. That was only ever partly activated --- the cataclysmic materiel-heavy cross-Atlantic war with the US and Canada fighting the Axis wasn't the way that things played out.

https://en.wikipedia.org/wiki/Bracero_Program

The Bracero Program (from the Spanish term bracero [bɾaˈse.ɾo], meaning "manual laborer" or "one who works using his arms") was a temporary labor initiative from 1942 to 1964 between the United States and Mexico that allowed Mexican workers to be employed in the U.S. agricultural and railroad industries.[1]

The program, which was designed to fill agriculture shortages during World War II, offered employment contracts to 4.6 million braceros[4] in 24 U.S. states. It was the largest guest worker program in U.S. history.[1]

The war plans were not conditional on atomic weapons --- they were made prior to the creation of those weapons. However, in that timeline, assuming that the Manhattan Program tracks something like its historical rate in our own timeline, the war would probably have run longer and thus the US and Germany would still be fighting when the weapons became operational. The US would probably thus have wound up utilizing atomic weapons against Germany; the late war US projections were that the US would scale up to mass production of atomic bombs within months. My guess is that the WW2 death toll would probably be higher than it is in our own timeline.

The next morning, six infantry divisions, two motorized brigades, a paratrooper unit, and hundreds of aircraft, including 186 Heinkel bombers, launched Operation Weser Maneuver. Denmark capitulated. Norway resisted and was crushed. “Once we have the two countries,” Goebbels recorded, “England will be flattened” because Germany could use Scandinavia as “a base of attack.” As for the United States? That country “is of no interest to us,” Goebbels wrote, because by the time the Americans could deliver any material assistance (eight months, in Goebbels’s reckoning) or put boots on the ground in Europe (18 months), the war would be over.

The US didn't especially want to fight Germany on land in Europe, not in the near term, at any rate, but rather across the Atlantic, a battlefield which played to its strengths. The basic US intent was to keep the Axis from acquiring a beachhead from which it could stage against the US.

If Nazi Germany declared ownership of the American colonies of European countries that it had conquered, US war plans involved occupying them and declaring war on Nazi Germany. It didn't want to have them used by Germany as a staging point for being used against the US; it wanted to force the Axis into a naval war, so that it could leverage its greater industrial power and turn the war into one of naval and air construction, in which it would be able to ultimately overwhelm the Axis.

searches

Oh, this is actually a better source than the one I was looking for.

https://history.army.mil/portals/143/Images/Publications/catalog/4-1.pdf

The Framework of Hemisphere Defense

The Crisis of 1940

Though the United States Army in the summer of 1939 was stronger and better prepared for action than it had been in the earlier 1930's, it was numerically far weaker than the army of any other world power. On the other hand the United States Navy, somewhat favored over the Army in the initial rearmament program, had a strength just below that of Great Britain. While Japan's ominous naval expansion was making protection of American interests in the Pacific an increasingly formidable task, the Navy in general was ready to perform its traditional function of providing the first line of defense in a war emergency. Nor can the military power of the United States in 1939 be reckoned solely in terms of active Army and Navy strengths. Both services had partially trained reserve components, and the nation's industrial might constituted a tremendous military asset. As World War II was to show, the military potential of the United States exceeded that of any other nation.

Sometime during September 1939, when the President was shown a draft of one long-range scheme for military expansion, he is reported to have said: "Whatever happens, we won't send troops abroad; we need only think of defending this hemisphere." The President and his aides likewise foresaw that no serious threat to the Western Hemisphere could arise unless the British and French were pushed to the brink of defeat. In that event the United States would be faced with the grim choice either of supporting Great Britain and France, "as our outlying defense outposts," or of vastly increasing American naval power to "meet the ultimate issue between us and a Russo-German Europe bent on dominating the world, somewhere in the Middle Atlantic." In an informal discussion on 19 September, the President and Assistant Secretary of State Berle
. . . ranged the globe, forecasting the division of Eastern Europe between Germany and Russia, wondering whether Western Asia was also to be divided, and guessing at the chance of an ultimate German foothold in the Atlantic. Both thought that if Germany won the war, Hitler would try to get his hands on the Azores or Cape Verde Islands, as bases for operations against the Americas. But both agreed that the war's main danger to this country lay in the alternative prospects of post-war economic chaos or a world economy dominated by the dictatorships.

The new joint RAINBOW 4 plan was based on assumptions that clearly indicated the dire forebodings of Army and Navy officers at the end of May. It assumed that, after the defeat of Britain and France, the United States would be faced by a hostile German-Italian-Japanese coalition. Its combined naval power, bolstered by portions of the British and French Fleets, would considerably exceed that of the United States. Japan would proclaim its absolute hegemony in the Far East, and might seize the Philippines and Guam.
Germany and Italy would occupy all British and French territory in Africa, and also Iceland. In Latin America, the Germans and Italians would use every means to stir antagonism toward the United States, and they might succeed in establishing pro-Axis governments in strategically located countries. Canada, remaining technically at war with Germany, would occupy Newfoundland, and the United States would have to join with Canada in the defense of Newfoundland and Greenland. Nevertheless, a considerable interval would probably elapse after the British and French collapse before the United States would be drawn openly into war.
The United States planned to counter these threats initially by occupying key British, French, Dutch, and Danish possessions in the Western Hemisphere claimed by Germany and Italy as the spoils of war. Thereafter, its armed forces must be disposed along the Atlantic front of the hemisphere so as to prevent any lodgment by Axis military forces. In the Pacific, every effort would have to be made to avoid open hostilities with Japan; if they began, the United States should base its defense on Oahu and Alaska. The major portion of the United States Fleet would have to be withdrawn from the Pacific and concentrated in the Caribbean area. Though the original RAINBOW 4 concept had contemplated defense of the entire Western Hemisphere, the armed forces of the United States for the time being would have to confine their operations to North America and the northern part of South America (approximately within RAINBOW 1 limits), extending their operations southward only as additional forces became available. While maintaining a defensive position in both the Atlantic and the Pacific, the nation would have to increase its military power as rapidly as possible, with the eventual objective of limited offensive action. In presenting the RAINBOW 4 plan to the Joint Board, the Joint Planners stressed above all the critical situation that would arise if the main elements of the British and French Fleets were surrendered to the Axis Powers. Should that happen, Germany and Italy would soon attain a naval strength in the Atlantic equal or superior to that of the entire United States Fleet. The planners estimated that the Axis nations would require a minimum of six months to recondition and man the surrendered vessels. For the United States, they pointed out, there would be two critical dates in this process: "The first is the date that either the British or French Fleet ceases to function, by reason either of destruction or surrender. The second is six months after that date. . . . The date of the loss of the British or French Fleets automatically sets the date of our mobilization.

[continued in child]

If UK doesn’t manage a free trade agreement with EU

The UK has an FTA with the EU, the TCA. It was negotiated as part of Brexit.

https://en.wikipedia.org/wiki/EU%E2%80%93UK_Trade_and_Cooperation_Agreement

The EU–UK Trade and Cooperation Agreement (TCA) is a free trade agreement signed on 30 December 2020, between the European Union (EU), the European Atomic Energy Community (Euratom), and the United Kingdom (UK). It provisionally applied[3][4] from 1 January 2021, when the Brexit transition period ended,[5] before formally entering into force on 1 May 2021, after the ratification processes on both sides were completed: the UK Parliament ratified on 30 December 2020;[6] the European Parliament and the Council of the European Union ratified in late April 2021.[2]

An operating system cannot brick a piece of hardware.

Sure it can. It may not have in this case, but there's plenty of hardware that can be bricked by an OS.

Interesting. I wonder if it'd be practical to containerize them by default.

That's also a solution for Windows users if they can't get some future version of Windows to work with the thing. You get a dinky Linux box, like a Raspberry Pi or something, and just set it up as a print server.

Wooldridge sees positives in the kind of AI depicted in the early years of Star Trek. In one 1968 episode, The Day of the Dove, Mr Spock quizzes the Enterprise’s computer only to be told in a distinctly non-human voice that it has insufficient data to answer. “That’s not what we get. We get an overconfident AI that says: yes, here’s the answer,” he said. “Maybe we need AIs to talk to us in the voice of the Star Trek computer. You would never believe it was a human being.”

Hmm. That's probably a pretty straightforward modification for existing LLMs, at least at the token level.

You can obtain token probabilities, so you can give some estimate out-of-band confidence in a response, down to the token level. Don't really need to change anything for that, just expose some data.

And you could make the AI aware of its own neural net's confidence level, feed the confidence back into the neural net for subsequent tokens, see if you can get it to take that information into account.

https://en.wikipedia.org/wiki/Recurrent_neural_network

In artificial neural networks, recurrent neural networks (RNNs) are designed for processing sequential data, such as text, speech, and time series,[1] where the order of elements is important. Unlike feedforward neural networks, which process inputs independently, RNNs utilize recurrent connections, where the output of a neuron at one time step is fed back as input to the network at the next time step. This enables RNNs to capture temporal dependencies and patterns within sequences.

I guess that'd make open-source firmware illegal.

If only there was a standard for how high the headlights are off the ground, regardless of vehicle height. Sigh, that ship apparently done sailed though…

Can't change it for existing vehicles, but can for new ones. The existing ones will eventually age out.

EDIT: I suppose that it might also be ultimately obsoleted as an issue on the receiving end if either:

Everyone winds up using a self-driving car.
If we shift to vehicles that don't have glass windows, like, just rely fully on cameras and displays. We already have vehicles doing things like providing radar HUD warnings and camera displays. We could move all the way to displays. It does make the continued functioning of the camera and display system functioning life-critical, which is not a trivial move to take --- if it fails, you're now fully blind at highway speeds --- but it could solve a lot of other issues, like helping older drivers with degraded vision. And for self-driving cars, the computer vision system is already life-critical, so...shrugs it's a problem that has to be dealt with anyway. If you're using cameras, then you can post-process the stuff and do whatever you want, merge data from infrared cameras, outline people and vehicles, get the blind spots from the pillars out of the way, whatever. Would also permit for cars to be a lot more secure and have better thermal and sound insulation.

https://www.twz.com/news-features/russia-eyes-balloon-communications-system-to-fill-massive-gap-left-after-losing-starlink

Ah, gotcha, thanks. It looks like that's a larger and more elaborate than what I was thinking of, with a ballast system and limited maneuverability. considers I dunno what the cost impact is.

However, the Barrage-1’s comparatively low altitudes could make them targets for Ukrainian air defense systems and other countermeasures.
“And what’s most important for us? To have the means that can detect such objects over our territory,” Beskrestnov suggested. “And to have the ability to shoot down such targets if they pose a threat. As far as I remember, the S-300 [surface to air missile system] can engage targets at an altitude of 20-30 km (about 12 to 19 miles).”

If it's expensive enough, then using S-300s or comparable systems becomes economical.

Ehhh....I mean, if the things have a GPS receiver, which I assume that they do, they can probably be configured to move to a given location and then only then flip on the cell radio to act as a relay.

EDIT: Honestly, I'm kind of surprised that someone hasn't tried a drone that can deploy, say hydrogen or helium balloons with a relay radio hanging from them. It's gotta be a complete pain in the ass to try to shoot balloons down, as they're cheap, and they probably linger in an area long enough to permit for operations using them as a relay on an extended basis. They can also probably get a lot higher than a comparable drone, if that's desirable.

AAA studios are doubling down on micro transactions, “performative” social justice, AI slop, and live service slop. - AC Shadows

I think that one factor driving either microtransactions, freemium, free-to-play stuff that does data-mining, or "incomplete" games with expansions is resistance to a higher initial price. I mean, if a studio isn't making their return on the initial price, they're going to look for alternate routes. AAA games cost more than ever to make these days. If people say --- and I've seen plenty of people on here do so --- "I absolutely will not buy a game with an up-front price of more than $N"....but then they're okay playing freemium stuff or games with microtransactions, I mean...that's what game studios are going to do.

I'm generally okay with an expansion model, because I like the idea of giving the studio the option to expand really popular games, and it de-risks things for both the player (you just buy the base game and get expansions if you want) and the publisher (you don't put down a ton of money to create massive amounts of stuff for a flop). Plus, some of my favorite games (including indie and open-source games, like Caves of Qud) have very long development cycles, and selling expansions is one way for larger developers to do a long development cycle...though honestly, I do agree that I miss the "just pay and get a complete game" approach, for a lot of games.

I don't care about the gambling (if it's for money --- I'm fine with variable-ratio schedule reward stuff in games, like having random loot drops in games). We know that that strongly appeals to humans.

https://en.wikipedia.org/wiki/Reinforcement

Variable ratio schedule (VR) – reinforced on average every nth response, but not always on the nth response.[16]: 88 (ex. Gamblers win 1 out of every 10 spins on a slot machine, however this is an average and they could hypothetically win on any given turn)

Variable ratio: rapid, steady rate of responding; most resistant to extinction.

...and I'm amazed that people would be dicking around with cryptocurrencies as a form of recreation, but if you put out good games that have erotica that appeals to me in 'em, I'll buy those. Video games are not incompatible with erotica. There are a lot of really bad video games with erotica out there, too.

In just the US consumption of Tiktok is up 39m hours a day compared to pre-COVID figures.

I really don't like short-form video, but I do appreciate that a lot of people do like it.

In the 2025, American consumers spent roughly $5bn on Onlyfans.

My suspicion is that streamed pornography is likely presently more of a substitute good for stuff like static pornographic movies than video games, though I'm willing to believe that I could be wrong.

During this 2025 period, AI apps that allowed for "role play, erotica, and art" have soared. The latest tracked statistic for installs for this software came to just under one billion worldwide.

Well, yeah. That's new tech.