Vecr

12 ms·

181.
by Vecr 2y ago
I'm not sure when this was written, but any new developments in non-equilibrium thermodynamics and/or spacetime theories since then don't change the result. Minds can't be rescued from the past or from "ashes"&
182.
by Vecr 2y ago
People are saying very weird things in the comments. To the extent that epigenetics transfers at all, they can't go very far. For past-life memories, uh no. For memories in non-brain tissues, there's a major detail problem there,
183.
by Vecr 2y ago
I thought the Mitochondria Liberation Front was defunct. Are people getting back into this type of linguistic debate? Anyway, uplifting as a whole is pretty sketchy and I'm not sure there's too much point.
184.
by Vecr 2y ago
Fire Upon the Deep had more or less this. Story important, so I won't say more. That series in general had absolutely brutal bandwidth limitations.
185.
by Vecr 2y ago
Trust the plan?
186.
by Vecr 2y ago
https://gwern.net/backstop#hui-nengs-flag -- Evolution as Backstop for Reinforcement Learning -- Hui Neng’s Flag -- the mathematics of positive and negative reinforcement https://gwern.net/fiction/batma
187.
by Vecr 2y ago
1) they're not that slow, and 2) I'm talking about the cut down versions of Siphash
188.
by Vecr 2y ago
Depending on the length of what you're hashing, either one of the cut down Siphashes or Blake3 are reasonably fast. The Siphashes are too short I think for general use, but when cut down and keyed they're okay for a hashmap.
189.
by Vecr 2y ago
The entire point of my solution is to prevent another HK situation. It's quite likely it still wouldn't work (deep state, backsliding, etc.), but it's a total no-go otherwise.
190.
by Vecr 2y ago
Personally I'd support a supermajority process, similar to a US constitutional change, started in both mainland China and Taiwan so they both have the same constitution. Once that happens the merging of the countries could carefully pr
191.
by Vecr 2y ago
I'm not exactly sure what you're saying. I know you can have gravitons without "quantum gravity" (as incomplete theories). I'm responding to a commenter on applications. In a global position finding system, gravitat
192.
by Vecr 2y ago
Cheap GPS receivers already have to do a bunch of tricks to get to the "okay" state they're currently at. Military devices either use GPS, star tracking, dead reckoning, or some combination. For submarines, detecting gravity
193.
by Vecr 2y ago
Huh? Like one of the Tegmark levels? Even then, a whole lot is not true anywhere near here.
194.
by Vecr 2y ago
I watched the video and that's what the fictional presenter says. I don't think it really makes sense though, but maybe you can get passed that because the species coming into existence at all is so contrived and unlikely? The vam
195.
by Vecr 2y ago
Yeah it's implausible. I mean, I don't think his genetics is right. So it's probably physically possible in some sense, but I don't it makes sense that it's that stable of a trait.
196.
by Vecr 2y ago
Almost but not quite literally zero? It probably is zero, because he made mistakes, but maybe you could brush that off as a narrator error or something.
197.
by Vecr 2y ago
Something being logically consistent doesn't mean it's correct. It's possible someone could make a fully logically consistent version of string theory including future gravitational predictions. They say "doesn't de
198.
by Vecr 2y ago
B. F. Skinner got there first.
199.
by Vecr 2y ago
This whole thing isn't really going that well. From what I can tell, 20 years ago it was pretty common to think that even if you had a "friendly" AI that didn't need to be boxed, you didn't let anyone else do anythi
200.
by Vecr 2y ago
I'm pretty sure it wouldn't work. Sqlite commits by deleting a file in normal operation, and with WAL you need an ordering that I'm not sure Linux on standard hardware can provide without running both the database and the WAL
201.
by Vecr 2y ago
I don't think so. I'm pretty sure sqlite relies on fsync barriers, something block devices don't have.
202.
by Vecr 2y ago
"Interesting" definition of AGI. Doesn't match the Metaculus standard, which is entirely capability based. The arguments for AI drives are reasonably good, so what he says isn't "wrong" in the sense that it&#x2
203.
by Vecr 2y ago
They got the simulation based inference going though, right? How high can they crank up the energy and still have that work?
204.
by Vecr 2y ago
It's lucky the predictions almost exactly matched. Otherwise the inference would have been a nightmare.
205.
by Vecr 2y ago
Sonnet 3.5 is the best model for many tasks.
206.
by Vecr 2y ago
https://arxiv.org/abs/1508.02421v1 https://arxiv.org/abs/1508.02421v3 (updated, 2017) They claim to fix the criticisms, see the section "The Bayesian arrow of time." Who knows how well t
207.
by Vecr 2y ago
On the "information" part, I'm not sure that's correct, have you seen the papers that dispute the Bayesian versions of thermodynamics? I think practically though, even before you hit anything "quantum", the req
208.
by Vecr 2y ago
It probably (if the calculations are right) is unable to actually do much of anything useful (because it's too complex to avoid being extremely correlated with the rest of the universe ("embedded")), and even if it could it
209.
by Vecr 2y ago
Holding materialism as an axiom, either directly or non-directly (through other axioms) could be called a "religion" (though at that point I'm not sure what couldn't be), and either way that could be considered bad. Thin
210.
by Vecr 2y ago
If they don't fix it, what will their reputation be like? In isolation giving up is probably the correct thing to do, if it wasn't so extremely public.

More ›