I now know the Mandela effect is real because Wikipedia is telling me this guy's name isn't Clippy?
It's... "Clippit"??
Replying to @willcb
This is a brilliant pun that I think most are missing...
(The IPO is in May, if that helps)
Replying to @aadilpickle
Underestimating the capability of EAs to CDT, FDT, and EDT themselves out of this. They even have a word for this exact strategy ("mugging") and have written probably something like twenty books on how to not fall for it, and also why you should fall for it,+ AIs falling for it.
Replying to @Jess_Riedel @tenobrus
Thanks for the vocab. :) I was using the terms I was familiar with (cases of misconduct, academic or otherwise, inside of universities mostly... lots of committees), I figured they weren't most standard but didn't know the right ones.
I admit I haven't thought about the 5th one!
Happy to have arxiv.org/abs/2610.31638 out!
(Deterministic) query complexity, and quantum query complexity, and several other kinds of query complexity, are *multiplicative under composition*: D(f ∘ g) = D(f) D(g). Randomized query complexity should be too, but it's not. :)
I unironically believe this is massively missing from standard grade school civics lessons. If you don't ever stop to think about why we penalize crime, you won't have an informed opinion on what's just or not, and ultimately unjust laws are created to placate folks like you.
Once you numericalize everything, some people fall into the other trap of "okay, so then it's just game theory right?". But it's not: we don't set up criminal penalties to minimize all crime (penalties go to infinity) or to maximize global welfare, we set up laws to feel just.
Beyond the "adjust penalties to expectation value of nullification harm", there's "adjust penalties to expectation value that we catch you". If you get caught doing X and only 10% of X's get caught, then it's somewhat sensible to say punishments should be 10x'ed to compensate.
Taking away your license, on the other hand, does nothing to nullify the harm, but is very effective prevention of further harm, and also good deterrence (people who speed usually like, or need, to drive!)
So it's reasonable to say you should pay a penalty something like Expectation[property damage] as a baseline nullification component. Of course, it's hard to put a price tag on killing someone, and you can't really fix that with money either, so it's hard to figure in.
It gets messier when it's "nullification of expected harm". If you get caught speeding on the freeway, you probably did ~zero harm on your own. Nothing to nullify! But there was a *chance* that your behavior would destroy a car or kill someone. And it's largely random which.
"You stole $50,000 from this house"
-> You must give back equal money (nullification of harm), plus extra for inconvenience
-> You get a probation officer (check in that you're not doing more crimes, discourage them)
-> 30 days in jail (unpleasant, makes you+others not want to)
Replying to @tenobrus
There's a shocking number of people who have never thought about whether sentencing should be for nullification of harm, prevention of further harm (incl. via correction), or punishment/deterrence. They are very different, and all ~valid.
Replying to @elon_lit
It's mathematically neat that you can do this, but I don't see any way in which is this allows "malware" or an "exploit", and presenting it as such seems intellectually dishonest.
You're threat model assumes the user is downloading and running untrusted code -- or .so's!
Seems like some automated system saw heavy GitHub Actions use and concluded it was abusing Actions for general purpose computation and so 'flagged' (i.e. shadowbanned) me.
But this flow for recovery is broken!
@github's support could use some work!
GitHub
-Flags my account -> low rate limits
-I file ticket asking to have it re-enabled
-They respond that it violated ToS but don't elaborate where so I can't correct them
-"This ticket is archived, to respond open new ticket"
-Try to open new ticket
-Can't, rate limit 1/day🙃



