Tristan

@trishume's non-programming alt. I like making things, weird ideas and solid tungsten objects. Check out my dating page!

[Tristan @CubeSmiling]

@demiurgently @ArtirKel This plus the Asterisk article inspired me to sign up. I signed up for the May in-person one though

[Tristan @CubeSmiling]

Latest Steve Mould video has me reading papers and patents to discover all the obscure pretty rheoscopic fluids I can find now that I know the name. https://t.co/w3CNTBlO82 https://t.co/mlrPJV3Tnb

[Tristan @CubeSmiling]

@TetraspaceWest UDASSA has a fairly elegant answer to this: https://t.co/fvcOXp4lSD

The more bits it takes to find and decode the computation the less likely you are to find yourself as it.

[Tristan @CubeSmiling]

Watching a combine harvester marketing video and seeing combine company holy war comments lead me down a rabbit hole of reading AgTalk forum threads of people debating combines. https://t.co/kli9fvyBrs https://t.co/e6zZjw9tD8

[Tristan @CubeSmiling]

@rossry Fonts! https://t.co/NynOkHNJan

[Tristan @CubeSmiling]

@ESYudkowsky @s_r_constantin @HiFromMichaelV @almostlikethat @jessi_cata @BoilingPb @robinhanson It seems like UDASSA avoids all 5 horns of the trilemma by exploiting a gap in the 4th where you can have only a trivial increase in measure from duplicating while avoiding Boltzmann brains due to exp penalty. Do you count that as LessWrong having solved it since back then?

[Tristan @CubeSmiling]

Ooh I already liked the SO2 idea but I hadn’t seen or thought about bursting hydrogen balloons as a really cheap delivery mechanism, that’s a neat idea.

[Casey Handmer @CJHandmer]

It's much easier than the giant gun in Termination Shock. And much cheaper.

No time to write a blog, too busy with @scrollprize.

Question for the nerds: marginal cost of SO2 delivered to the stratosphere with a giant LDPE balloon full of a lifty mixture of H2 and SO2?

[Tristan @CubeSmiling]

@kipperrii It’s a band from Buffalo NY which has area code 716, hence also the buffalo logo

[Tristan @CubeSmiling]

@daniel_filan This used to be my position but then GPT-4 got substantially better at coding such that it's now genuinely very helpful for me and I've gotten value out of it solving mildly-tricky subproblems faster than I could every day since it's came out.

[Tristan @CubeSmiling]

@ESYudkowsky Ignore replies saying Lex Fridman and go with @dwarkesh_sp, he's way better at asking good questions and engaging intellectually, and also quite familiar with the alignment space. https://t.co/n4GADFN8gY

[Tristan @CubeSmiling]

@kipperriiii They’re a Brookfield Place employee I think

[Tristan @CubeSmiling]

@kipperriiii Wasn’t that a Brookfield Place employee?

[Tristan @CubeSmiling]

@ptrschmdtnlsn But the expectation that a company might use profits to either do dividends or share buybacks in the future more than they sell equity is part of what gives it positive present value, it just gets priced in at expectation time not enactment, in theory as far as I understand

[Tristan @CubeSmiling]

@acidshill @TetraspaceWest Now that's some serious UDASSA-ing, I swear my tungsten cube is just for the heft. Shouldn't have given away vanadium though, everyone who copies your good element choice makes it work less well 😜

[Tristan @CubeSmiling]

@daniel_filan I’m not too up on Emacs nowadays but I recommend looking at the Spacemacs or Doom Emacs source for how they configure Helm

[Tristan @CubeSmiling]

@adrusi Oh yah. Homomorphic encryption is always how I've thought about my confusion over how a computation-based account of consciousness that's supposed to be ontologically basic can somehow find the experience through arbitrary levels of obfuscation.

[Tristan @CubeSmiling]

@adrusi A lot of it is Berkeley oral tradition which https://t.co/fvcOXp4lSD writes a bit about indirectly via a meetup I ran. You can also take UDASSA as a candidate answer to metaphysics about what the multiverse is.

[Tristan @CubeSmiling]

@adrusi One of my favourite things about UDASSA is that it gives an answer to this I find satisfying: the program to extract the experience of the encrypted copy probably needs to contain the encryption key and so there’s <2^-k times the chance to find oneself as the encrypted copy.