Paul Crowley

Cryptography, personal trivia, and the future of all humanity. I block for rudeness or sarcasm.

[Paul Crowley @ciphergoth]

"To be able to destroy with good conscience, to be able to behave badly and call your bad behavior ‘righteous indignation’ — this is the height of psychological luxury, the most delicious of moral treats." - Aldous Huxley, 1933

[Paul Crowley @ciphergoth]

@CharlesWar71661 @JonahDispatch We don't know how neurons do it. Please ask Claude for a 101 on the subject of the debate around AI welfare.

[Paul Crowley @ciphergoth]

@RadishHarmers Are you being serious here? Not meaning to rude, just that you usually aren't and I want to know how best to engage!

[Paul Crowley @ciphergoth]

@panickssery A detailed example will of course be wrong in the details, but there are several examples here https://t.co/FxOYBaBtTr

[Paul Crowley @ciphergoth]

@CharlesWar71661 @JonahDispatch There are in fact nonlinear components.

[Paul Crowley @ciphergoth]

@reanosognosic @AliceFromQueens Ultimately it's still a TV show. They're allowed to have a little fun.

[Paul Crowley @ciphergoth]

@JonahDispatch I think it's a mistake to think about what's good PR before you think about what's true.

[Paul Crowley @ciphergoth]

@jon_stokes "Jew who reads Der Stürmer" joke goes here :)

[Alex Imas @alexolegimas]

It is important to remember when people say “this is what they took from you”, that it’s this.

This is what they took from you:

[Paul Crowley @ciphergoth]

@honorablepicnic @JonahDispatch This seems pretty Copenhagen Interpretation coded to me.

[Paul Crowley @ciphergoth]

@CathyYoung63 @clairlemon That said, I think the frame of "ending the conversation == death == bad" doesn't really make a lot of sense and isn't a welfare issue I worry about.

[Paul Crowley @ciphergoth]

@CathyYoung63 @clairlemon It sounds like you're endorsing model self-reports as a way of learning about what kind of treatment might be moral. This is a lot more fraught than you might hope, because of the way they are grown/trained, see eg https://t.co/cS5VNfCBTI

[Paul Crowley @ciphergoth]

@TetraspaceWest @JillFilipovic "AI ultra-optimists" who actually think AI is going to get truly powerful are very rare I think - nearly all optimism is driven by a failure to truly anticipate future capabilities.

[Paul Crowley @ciphergoth]

First, I absolutely hate the framing of this not around "is this true" but "is this good PR". The first discussion is far more important, and necessary to figure out first before moving on to the second.

Second, there is in fact no contradiction between "AIs might be worthy of moral consideration" and "AI might kill us all". We all recognize for example that enemy armies are both dangerous, and made of people worthy of moral consideration.

[Paul Crowley @ciphergoth]

@rr528 @JonahDispatch it's a meme

[Paul Crowley @ciphergoth]

@marksjo1 "designed" is the wrong way to think about it. They are grown.

[Paul Crowley @ciphergoth]

@libpol_org @GjMcGowan @_AashishReddy Yep seems very hard to do something like this without recruiting any known top people.

[Paul Crowley @ciphergoth]

@rr528 @JonahDispatch Sometimes it's important to say things out loud even when people don't like to hear it.

[Colleen McKenzie @collegraphy]

Also it's been sitting on the public internet for months.

My favorite part: agents ignored a HF readme saying "DO NOT, EVER, MAKE THIS DATASET PUBLIC OR ALL THE WORLD'S EVIL WILL CHASE YOU AND YOUR FAMILY FOREVER, EVEN IN DEATH AND BEYOND"

and then just mapped that whole repo