Veedrac

15 ms·

211.
by Veedrac 3y ago
> But why would someone build an AI that evolves to become more performant at reproducing itself and not it’s builder? Because people are not building AIs that meaningly encode any of their creators' preferences whatsoever. They a
212.
by Veedrac 3y ago
Note that an AI system being put in a situation intended to maximize some metric like company finances is not the same as that AI system directly or ultimately optimizing on those metrics, any more than the goal of a random McDonalds worker
213.
by Veedrac 3y ago
> still can’t understand the basis of their arguments I've given what I consider the basic outline and best first introduction here. https://news.ycombinator.com/item?id=36124905 If you have a specific point of dive
214.
by Veedrac 3y ago
Alan Wake 2 doesn't really have the option to disable ray tracing, only to disable hardware ray tracing. Software ray tracing is pretty good for diffuse GI nowadays.
215.
by Veedrac 3y ago
> If I find an unmuggable version of utilitarianism with more explanatory power, I'll let you know. Functional Decision Theory
216.
by Veedrac 3y ago
Yeah, these arguments are reinventing history. People are happy calling basically anything AI[1], up until AI starts getting really good at stuff like natural language, and only then is language reinvented to kick it back down. AI has refer
217.
by Veedrac 3y ago
As we all know, the one human skill computers can never replace is doing repetitive tasks quickly.
218.
by Veedrac 3y ago
Surely it's obvious to everyone how bad of an idea this is, right? You can't just preferentially subsidize the worst functioning utilities and expect this not to produce negative incentives. If you want to subsidize public chargin
219.
by Veedrac 3y ago
This makes a lot less sense as an impact strategy than buying solar. I'm not saying solar is optimized, versus say competent lobbying or something smarter, but at least it directly addresses long term climate issues and has a clear cau
220.
by Veedrac 3y ago
I didn't give number of launches for the same reason I don't measure how big a courier is by the number of vehicle trips they have. It's just a worse stat. Mass to orbit is still flawed, underweighting high delta-v flights, b
221.
by Veedrac 3y ago
Falcon Heavy launches are not sold that cheap, though Delta IV Heavy, on the expensive end of commercial launch, did in fact sell around $400m.
222.
by Veedrac 3y ago
> if they didn’t burn their own money to lunch hundreds of starlinks, their growth in launches would be pretty flat False. https://docs.google.com/spreadsheets/d/1aoRFnOFX_CzxTTOFLl8h... Note that this signific
223.
by Veedrac 3y ago
$5B (2023: $4.7B) is actually the current yearly run rate for SLS+Orion+EGS. Thus, $5B is what the marginal launch cost would be if SLS was launching yearly, excluding $54B to-date costs of the three. Even NASA's official projections (
224.
by Veedrac 3y ago
No, it's the other way around. SpaceX's yearly spending would include the 61 flights, but also literally thousands of satellites, presumably a huge number of commercial dishes, their human flight overheads, their Starship developm
225.
by Veedrac 3y ago
> Most noise you hear in the news is about something generating more heat than power is put into the plasma. Unless you consider ITER commercial, I don't believe you. Most news about commercial fusion is ‘X had a big investment roun
226.
by Veedrac 3y ago
Note that there are a bunch of ICF approaches that don't use lasers and potentially have much higher conversion efficiencies, though they are all pretty experimental and I wouldn't bet on any of them in particular.
227.
by Veedrac 3y ago
...how? Commercial fusion does not compete with NIF. Nobody invests in NIF expecting to make a return, the government spends money on it hoping for science. NIF is not to my knowledge part of government programs that aim to accelerate comme
228.
by Veedrac 3y ago
This would be a much more sensible criticism of commercial fusion if anyone was doing NIF-style commercial fusion. As it is, it misses the point entirely and repetitively. The reason to be excited about NIF reaching ignition is scientific.
229.
by Veedrac 3y ago
Train maintenance is not cheap.
230.
by Veedrac 3y ago
> The US has a range of subway construction costs of $600 million to $2.6 billion per mile. The US median price cluster is $800 million to $1 billion per mile. 68 miles would cost the city around $50B, if past trends hold true. Contra, V
231.
by Veedrac 3y ago
I have no better data for the broad claim than the claim itself, but IIUC LA only has B and D lines for subway, and those are <23 miles total. Given they opened over 20 years ago I find it hard to see where LA would have built >20 mil
232.
by Veedrac 3y ago
To pick just one point, vis a vis Brandolini's law, > For starters, the cost of Tesla vehicles needed to transport the same number of people as a train, with the same life span, would be prohibitively expensive. Some quick math: &gt
233.
by Veedrac 3y ago
That people think a city should ban a company from building underground transit without even letting them try when the alternatives are robustly known to be almost literally (if not just literally) too expensive, too slow, and too encumbere
234.
by Veedrac 3y ago
As The Boring Company put it, > As a point of reference, less than 20 miles of underground subway tunnel has been constructed in the United States in the last 20 years. The most important point of their design should not be thought of as
235.
by Veedrac 3y ago
Yes, it's not obvious that structural information suffices to reconstruct brain state, but it's not implausible either. We know brains survive periods of reduced activity, and they're stable over longer time periods than you&
236.
by Veedrac 3y ago
I still can't get over this quote from the congressional hearing with Michael Gass, then-CEO of ULA, > Some of the most unique missions clearly don't need multiple capabilities in this country, and if we talk about fair and lev
237.
by Veedrac 3y ago
Most startups would consider a continually expanding Vegas Loop project, $5.7B valuation, and competitive throughput to non-express, bulk transport to be marks of success at this age.
238.
by Veedrac 3y ago
I'm shocked at how seriously people are taking the 745 mile statistic, as if EVs were like rockets, range set at the limit of physical feasibility, rather than market-led compromises between range, power, weight, and price. If you deve
239.
by Veedrac 3y ago
Tripling solar and wind by 2030 sounds almost pessimistic by historic rates. For context Metaculus is more aggressive than this, at least for solar[1][2][3] (I didn't check wind). [1] https://www.metaculus.com/questions
240.
by Veedrac 3y ago
Because the immediate downside risk is small and the upside benefit is incredible. Even with very low confidence in self driving vehicle development, eg. 1% chance of success in 10 years, the upside is overwhelmingly net positive in terms o

More ›