I am very much reminded of the nostalgebraist post models may behave differently in graded episodes (a tirade).
CronoDAS
Karma: 17,077
I think the use case is attacking Israel. It’s damn hard for a terrorist that isn’t already inside Israel to cross the border between Israel and its neighbors in a truck and then start running people over. But sending drones over the border can be a lot easier.
Would Hezbollah hesitate to fire these things at Israel if it had them? Would they have better ways to kill a lot of Israelis, since they haven’t been able to successfully target their explosives at any critical Israeli civilian infrastructure?
I think Hezbollah is exactly the kind of non-state actor that would very much want this kind of weapon in order to actually use it, and they already get weapon supplies, including drones of various kinds, from Iran.
Oops. I need to find the second half of the argument—I thought it would have been in that one. Sorry.
https://alonzofyfe.substack.com/p/moral-ought-ought-not-and-reasons
https://alonzofyfe.substack.com/p/solving-the-central-problem-of-morality
How to derive an “ought” from an “is”:
https://alonzofyfe.substack.com/p/deriving-ought-from-is-hypothetical
I always thought that the problem with “belling the Cat” is not that a mouse would inevitably fail to bell the cat, but rather that whichever mouse did it would presumably be killed by the Cat shortly afterwards. So even though belling the cat wouid be good for the other mice, no individual mouse wants to be the one that has to do it
Almost nobody, anyway.
Isn’t the cost here that you need to keep track of the seed(s) used in your pseudorandom number generator and store them as a secret key? It does seem like a fairly trivial cost, though.
And does this method work if you don’t have the entire AI output from the beginning? For example, you ask the AI to write ten essays in a row, and then turn in only the last one to check if it has the watermark. If your PRNG is good, can you easily tell if a sequence “eventually” gets produced by a seed? Or do token limits make this a non-issue?
I’d be worried about random people being able to do the equivalent of teleporting a bomb into the Oval Office. Car bombs are hard enough to defend against...
CATS: ALL YOUR P ARE BELONG TO US
Barack Obama: The question of P is above my pay grade.
Shakespeare: To P or not to P, that is the question
Mad Magazine: Whether ’tis nobler to suffer, or goeth behind yon tree...
I like to joke that the reason quantum mechanics and general relativity are incompatible is that the universe is running on buggy code. ;)
[LINK] The Ferrett fails will save against the Dark Arts
As the saying goes, adding more manpower to a late software project makes it later.
“Keynes said that in hundred years the productivity would increase so much that we would only need to work 3-day work weeks.”—“Well, I guess people have a revealed preference to work 5-day work weeks.”—Sure, smart guy. Show me where can I find the job advertisements for the 3-day work week jobs. Oh, there are none! So from the fact that I didn’t choose from an empty set you can figure out what my true desire is. Amazing!
To be fair, in surveys, most people do answer that they wouldn’t want to work fewer hours if it meant that they would get less take-home pay.
You’re not wrong, but most part-time jobs are different kinds of work than full-time jobs, and they also pay terrible hourly wages when compared to the kind of work that is only available on a full-time schedule.
Paraphrasing stand up comic Paula Poundstone:
You’re probably wondering how I ended up with fifteen cats. The answer is that I had thirteen cats, and then I got two more!
Wasn’t the facial feedback hypothesis one of the things that didn’t survive the replication crisis—I think I remember it specifically being one of the things that failed to replicate?
<joke> If video games have taught me anything, it’s that you should always go out of your way to help other people with their problems. That way, you get more of that sweet, sweet EXP that you need to turn yourself into an unstoppable killing machine! </joke>
-- John Maynard Keynes (approximately)
The charitable reading is that enforcing liabilities would keep the technology safe while it’s being developed incrementally so it never actually reaches the “sudden explosive disaster” stage. (Eliezer Yudkowsky has plenty of reasons to expect that a “sudden explosive disaster” could happen anyway, including an existing, seemingly “safe” AI deciding that it has reached a point where cooperating with humans has become more trouble than it’s worth.)