The only thing that changes is the data that is passed to the LLM, which for each iteration includes the last token that the LLM itself generated. So yes, sort of. The LLM itself doesn’t change state; just the data that is fed into it.

It’s also non-deterministic insofar as similar inputs will not necessarily give similar outputs. The only way to actually predict its output is to use the exact same input - and then you only get identical token probability lists on the other end. Every LLM chatbot, by default, will then make a random selection based on those probabilities. It can be set to always pick the most probable token, but this can cause problems.

…The Planck constant is a set distance. It’s being used here to define other things as a constant. The very article on the kilogram states outright that SI units’ foundation is three constants: “a specific transition frequency of the caesium-133 atom, the speed of light, and the Planck constant.”

Vacuum permeability is how much electric currents affect magnetic fields in a vacuum. As in, when there are no other interferences. It is not variable. The speed of light when there is no interference is not ‘affected by’ vacuum permability any more than a cartesian graph is ‘affected by’ the set of real numbers. Vacuum permeability describes vacuum, it doesn’t define it. Same goes for permittivity.

You ARE correct about the kilogram no longer being a number of specific atoms, though.

The two constants - the speed at which light moves, and the unperturbed ground-state hyperfine transition frequency of cesium - can be combined to define every measurement of time, length, and velocity. They are the constants by which everything else is defined.

Throw in mass, which is easy - a certain number of atoms of a specific element will also have a universally constant mass. Combine it with the other two constants and you have force, energy, and work, and voila, you can describe nearly everything in classic physics.

That was actually partly what I was thinking of. Whether you liked the voice is up to preference, but you’d have a hard time arguing that the performance was worse than typical dubs from the era.

Installers deal with more than just unpacking files into a folder. There are often prerequisite shared libraries that are included in the installer that AREN’T in the game directory, which may or may not need to be installed along with the game depending on if your system already has it.

So just double-clicking the .exe after copying the folder to a new computer is not reliable in the same way GOG’s installers are.

I’m pretty sure that’s what you said last time.

The issue is more that English dub voice actors tend to be … not as good, and wind up sounding stilted. Watch the dubbed version of Princess Mononoke, and compare it to … uh…

You know, it’s been so long since I watched any English dubs that I’ve plum forgotten what some of the worst offenders are. Naruto? Sailor Moon?

I can’t help but wince at the thought of not seeing the spider before plugging in the ethernet cable…

The best thing about it was all the functions that existed for string manipulation, since every data record was just a big string with demarcations for columns (and values and subvalues, making it technically a 4D database). You could use it to consume nearly any data format and throw together a quick-and-dirty parser in minutes. Really good for rapid proof-of-concepts, but occasionally challenging to maintain data integrity when you have all the rope in the world to hang yourself with.

I miss working with it, compared to SQL which is stupidly rigid and obnoxiously fragile in comparison.

Anticompetitive behavior is tolerated much more from companies that aren’t the market leader, and Apple Books is far from the market leader.

ffs, you do not own the games you bought.

You haven’t owned the games you buy since software licenses became ubiquitous.

Any business that has over 4% market rate should be forcibly split.

My dude, 4% is insanity. If you did that, there would be - at absolute bare minimum - 26 different companies for any given market, including distribution. No one would be able to scale to serve global demand in any way, shape, or form. You’d have to route packages through three different companies in order to get a product to your friend in Switzerland. Any product that got popular would see its parent company split. If someone wrote an OS that was too good and everyone decided it was fantastic, the company would somehow have to split in such a way as to divide the OS share between two companies, which would then negate the point of having the same OS as your buddy.

The problem isn’t market share, it’s enforced market share. Anti-competitive practices. Lobbying. Buying out competition. These are places that Steam demonstrably is more ethical than its competition.

All Epic had to do in order to capitalize on Fortnite was to make a launcher that was better than (or at least approaching the quality of) Valve’s. Instead, when the starting gun went off, they started walking backwards.

The problem is that, allegedly, there are threatening emails from Valve to developers who tried to sell for lower prices on other platforms (NOT Steam keys). If this is true, then there is actual ammunition against Valve.

Too on-topic, not enough self-aggrandizing tangents.

Rebel Assault was neat, but the controls were terrible. Using a joystick to point a cursor on the screen is frustratingly inaccurate - and not all joysticks were created equal. Cheaper ones were inaccurate and had weird axis profiles thta made the game even harder than it should have been - and trying to play it with mouse was at least as frustrating.

If the robot is the right size, it would be able to move vertically.

Mind you, you’re right, but new methods of locomotion can come in useful in surprising ways that you and I haven’t thought of.

You still lose the internal state between each token in the database output. It would let it plan, but it would still be externalizing that planning, one token at a time. Condensing all of the internal state into a single token at a time still means huge losses in detail as well as fragmentation of responses, resulting in all the problems that you see with LLMs.

Somehow the actual internal state needs to not only be preserved, but fed back into itself. That’s how brains work. Condensing it into tokens isn’t enough.

Except nothing will cause the function to collapse. With real superpositions, observing the state causes it to have always been only one specific position the entire time.

Running cables through conduit is one example that comes to mind. You need something to drag the lead, and if there’s corners or something that thwarts using a fish tape, this kind of robot would be perfect.

How is this not in a NotTheOnion community?

There’s been significant advances in super-cheap batteries that have lower power density, but excellent resilience, utilizing common materials. Those are ideal for solar storage where space isn’t an issue.