l'empathie mécanique @dpwiz@qoto.org

@danilo > Give me a week and you can have plans for a scalable fusion reactor design.

Oh... Trading "decades" of building AI for finally getting fusion after a week instead of perennial "in 20 more years" is so, so, so worth the deal.

Even if it burns out after that single miracle, having cheap clean energy will solve any climate-related problems in no time. We have solutions already, the problem is they aren't particularly energy-efficient.
With a few more hints at efficiency geoengineering can be achieved to marvelous results.

@danilo Incremental improvement in carbon capture is all it has after all the time? Did it even have "Engines of creation" in its corpus? We are a living proof of programmable nanotechnology being possible. Even humans can solve it on their glacial bureaucratic timescales given incentives. With AIs, this step is basically unavoidable due to its omni-useful nature. Whatever the task you can postulate "brb, inventing nanotech" will be one of the first replies from a truly capable AI.

If the story is about a chatbot instantiated for hype, then yeah, everything checks out. But if the "1000 person-years compressed into 1 hour" is the real deal then the ending doesn't make any sense.

send purr requests on mewhub

@leobm The key point for me was understanding what "type A is inhabited by values x, y, z" is about.
Next, if types are some propositions in logic, then values are their proofs.
So, by writing code that compiles you're proving that some logical statement holds.

Interestingly, if you can't say something with types, no way you can write actual code that gonna work.

And of course, if a type system admits bullshit, then you can bullshit compiler with your code.
Every industrial language admits BS, but some of them make it easy to lie, cheat, AND steal (esp. with dynamic types), while the others make that painful.

A good type system can review your design before you have committed to implementing something unrealistic, being a net positive on effort spent pinning down all the important stuff.

@ebel 🌈, if you squint hard enough

@boilingsteam whatever, as long as it runs Steam. Preferably SteamOS.

Windows multi-store schiso-tablets... No thanks.

@arialdo There's [Void#](https://hackage.haskell.org/package/base-4.19.0.0/docs/GHC-Exts.html#t:Void-35-), for the times when you really want to a-void something

@arialdo What do you want to understand? The function ignores its argument to produce String. It may as well be `f :: a -> String`.

The `error` argument isn't evaluated and so no actual errorirng happens.

@reidrac@social.sdf.org @doragasu Just use Cop-s Lock, like Google does on chromebooks.

@pmidden Even better: commas are whitespace (clojure)

@mhoye Reclaim *all the things*!
Don't let memetic ideology limit your options (that are nice by themselves and properly fit together). Memelords can go wear themselves.

My, the times are fast now. It's been only a year, and we already have the art style of *vintage* neural images.

https://nitter.cz/WeJe3396/status/1740442739034988768

@aburka @Viss @forrestbrazeal @amyengineer There are a few attempts on that. The thing generates its own code to run a next iteration.

@glynmoody If only they could stop mixing with civilians and using public infrastructure to stage attacks. Their leaders are their worst enemies in this, with all the "victory is near" bravados while shooting their own children for "stealing" food.

@someodd Yay, more Haskell games! dpwiz@gh

@gd that's almost a tautology. It is correct under the given set of axioms.
Then, ex falso quod libet somewhere in there and you have a prod full of bugs.

@aras Reminds me of this trick https://petapixel.com/2021/07/26/unwrapping-buzz-aldrins-visor-in-moon-photo-reveals-what-he-saw/

@aras impressive feat of un-rendering!

@amiloradovsky and by whom