the amount of cooling you get is proportional to the difference of component temperature to ambient temperature. thats why modern chips are engineered to run much hotter.
Actors tend to lead to bloated incomprehensible code. Even when you do want game objects to look like independent message passing actors. Faking it is always easier.
Actors are self defeating because they turn everything into a distributed system.
That's what I was thinking. I'm trying not to write Actors off because I'm not an expert on game dev and can imagine the possibility of an unusually large-scaled game. But yeah, if you're writing a typical game, I imagine you're best off doing it the typical way that's been optimized for.
think of corrugated metal. sound waves and heat will travel slower in the wavy direction than in the straight direction. you can stamp some more complicated pattern onto a piece of metal to control how sound waves travel through it. almost like a sound circuit. (eg. passive sound canceling metal panels. things like that)
you could come up with some material that naturaly has those kinds of properties without being wavy. but mathematically its the same thing.
> Nobody knows how even the current generation of LLMs work (at the plumbing layer, sure, but the high-level behavior is a complete mystery).
i dont think thats fair to say tbh. clearly language has alot of patterns in it. so it should be possible to come up with a network that finds them. if the loss converges to something reasonable and the amount of training data is greater than what the network is able to memorize, then the only possibility is that the network learned whatever general patterns the data happens to have.
the jacobi identity is [x,[y,z]] - [[x,y],z] = [y,[x,z]] ie. it says the lie (pronounced "lee") bracket is almost associative but you have to add a [y,[x,z]] when you rearrange brackets. it has absolutely nothing to do with disambiguations.
both of these are reasonable. if you have an `x` such that `x + n = x` implies that `n = 0`. (assuming x still has an additive inverse)
in other words you just invented modular arithmetic which is a very reasonable thing to invent.
1/0 is maybe a bit trickier and leads you to invent projective spaces.
Anyone interested in such a thing would just keep their money in T-bills via a money market fund. Fast access, zero credit risk, but there is the hassle of sweeping/topping up the bank account used for transactions.
I understand many banks offer "insured sweep" accounts that do just this but of course this costs them cheap deposits. Go ask the various VCs why they didn't press SVB to make this service available to their portfolio companies.
Even $30B (as a single claim) would be difficult to handle. Again, name a company you would trust to hand over that much money then and not topple over.
