@VoxelVexillologist's banner p

VoxelVexillologist

πŸ‡ΊπŸ‡Έ Multidimensional Radical Centrist

1 follower   follows 0 users  
joined 2022 September 04 18:24:54 UTC

				

User ID: 64

VoxelVexillologist

πŸ‡ΊπŸ‡Έ Multidimensional Radical Centrist

1 follower   follows 0 users   joined 2022 September 04 18:24:54 UTC

					

No bio...


					

User ID: 64

And I’m not sure if you realize this, but increasing nameplate capacity 30% in 2 decades is absolutely awful.

That's fair, although IIRC per-capita consumption has declined (mostly through improvements in efficiency --- lighting, HVAC efficiency, better insulation, etc), so it's not quite as dire as it sounds.

whatever the American equivalent of Sodexo is,

Sodexo exists in the US. There might be others, though.

I think Abbott is at least partially swaying with the political winds here, but ERCOT's 100% increase in the last two decades pales in comparison to their forecast of 300% growth in the next 6 years. I think there are probably legitimate questions of capacity buildout at much higher rates than baseline, and "not committing to anything further" doesn't sound entirely foolish.

On the third (gripping) hand, I really question the hyperscaler economic case where they're able to pay N times the current price for huge amounts of energy, but the proles will still have enough money to pay them for AI services and for utilities. I don't know about everyone else, but "paying to chat with Claude/Gemini/ChatGPT" falls pretty low on Maslow's heirarchy of needs for me, well behind "keep the lights on and HVAC going". Even if they expect B2B to pay the bills, consumers are still an important part (ultimately, the endpoint) of the economy: it can't be B2B all the way down. Models are getting more efficient and I somewhat expect that quite a few use cases need less hardware than currently forecast.

This feels somewhat constrained to the local regulatory environment. We can just make investments in the grid faster. In the last two decades, total US nameplate capacity is up 20-30%, while ERCOT (Texas) is up over 100%. They don't want you to know this, but you can just build things: Texas is more friendly to renewables than California these days!

I lean towards supporting vigilante justice only in cases where the vigilante then turns themselves in and pleads guilty to a hostile court.

"Why, Gary? Gary, why?"

Honestly as a case I have mixed feelings about it, but I'm not convinced Gary was completely in the wrong.

Notionally, if we build out the grid as if usage will double, if they don't we can expect prices to fall. This would among other things be otherwise convenient for environmentalists, because much of the added power (certainly the cheapest marginally) will be solar/wind, and we need a bigger grid to support EVs and heat pumps, and cheaper power makes both more attractive.

Yeah, I realized after writing this that the server racks at work, which are air cooled, probably count as "open loop" here because the large site HVAC (for offices and equipment) has a cooling tower.

if these cooling loops could use gray water or something similar

Is anyone in the data center space running open-loop cooling? Otherwise they're just using the same water round and round. It looks like it is used some places, but we could easily be adding incentives or requirements for closed loop (air of ground source) cooling and battery-backed solar on rooftops instead of outright bans.

I've been toying with the idea of working out whether faithfulness to the source material is correlated with quality of the end product.

There are quite a few movies that are usually considered to be better than the books on which they were based. Unfortunately, most of the classic examples here (Jaws, The Godfather, Fight Club, Forrest Gump) I haven't read. Not to mention that the order you're exposed to them could matter.

I have seen that claim for Jurassic Park, but I'll say I quite like both.

Is "money" really that different between France and the UK? Per capita GDP is not terribly different (UK wins slightly), and I thought living standards were quite similar. But I'll admit I haven't been to either recently.

major tourist centers

Anecdotally, many places with lots of tourism have locals that really dislike tourists. Even when tourism indirectly pays the bills.

the system HAS to make a binary determination that either makes the person completely irresponsible or completely responsible

Sure, the current system does. But as philosophers here, we could muse about alternatives. What about a fractional responsibility regime?

He justified his skepticism towards the medication by mentioning one anecdote of my younger brother taking Concerta with no discernible effects, saying that he'd rather not take any mind-altering chemicals as his output would then not be entirely his own doing.

Are you sure skepticism of stimulants to treat ADHD is completely unwarranted? Sure, the ranks of skeptics include Scientology and RFK, but even The New York Times has published such opinions relatively recently [1] [2]. I'm not going to call myself a complete skeptic here --- I know people for whom the medications seem to work well --- but the diagnosis criteria are really a spectrum, and the rates at which we're prescribing these drugs (23 percent of 17-year-old American boys!) is concerning with respect to how we've pathologized cases that really can't be that far from the median, and the drugs have very real side effects. To the extent that they're performance-enhancing for most people, there is a real question of fairness and applicability, but I won't claim to know enough to have a strong opinion about that here. See also the fraction of high-achieving "mentally disabled" students at top-tier universities demanding special accommodations.

Medication might well work for you. I'm not a doctor, and I'm definitely not your doctor, but I would recommend a little sympathy for differing views here. There might well be other reasons to cut off contact, but IMO "ADHD skepticism" seems, alone, well below that bar.

Shortly before Trump came down the escalator, America was at less risk of a race war than at any time since the Monroe administration, and possibly since the founding.

The escalator (June 2015) was less than a year after the first Ferguson riots (which spanned several dates, the last of which was after the escalator). I would agree that race relations probably reached their best point during the early years of the Obama administration, but there was a demonstrable slope toward weaponizing it from as early as the White House Beer Summit (6 months into office). I think many people expected better improvements under Obama's leadership (I'm not going to say those expectations were, in hindsight, realistic), and more than a few voters soured on the idea in both directions.

Before mass migration we could functionally ignore HBD and be nice and get along because the country was 87% white

I'm not going to say you're fully wrong here, but history suggests that the 87% of what is currently considered "white" did not self-identify at the time as homogeneous. The largest single lynching in US history was of Italians (and why we celebrate Columbus Day). And they're hardly the only such group.

The perceived unity of the mid-twentieth century seems IMO more a product of the trauma-bonding of the wars, the looming Cold War threat to unite against, and mass-market broadcast culture. TV, radio, and print media all existed, but were restricted to scales that gate kept access and enforced some degree of "same screen".

Epstein seems an interesting comparison, because he also killed himself [citation needed], doubtlessly because of, at least in part, a bunch of negative press coverage. Surely if we need these sorts of protections for plagiarizing, fabulist sociology professors, we also need them for pedophile financier socialites: think of how much blood is on those journalist's hands, too!

As an example, does anyone even remember that Stanford's (white) president was forced to resign in 2023 over research misconduct brought forward by the student newspaper? Fabricating (literal data manipulation!) Alzheimer's research in Nature you'd think would be a bigger deal. He's still a professor there, even.

I have observed in the past that warfare has gotten, for lack of a better term, more abstract in the last century or two. Time was, you fight because the enemy would sack your cities, take your lands, put you and yours to the sword, or similar. But those things are unseemly by modern Western standards (not saying they don't happen, though). Modern states fight for principles or alliances or things like that, and we'd likely prefer to force surrender (or at least negotiations) rather than aim for maximum destruction. See the current Iran conflict: we certainly could strike bridges and power plants, but that's not abstract like we'd like it (not judging, merely observing).

It seems the logical conclusion is that we'll move to warfare between entirely-robotic forces, with any human casualties seen as "gross, icky, war-crime material". On the other hand, the abstraction only works when all parties accept the implication: the Taliban didn't have the good sense to give up in the face of overwhelming force that wasn't accompanied by wanton destruction. Neither does Iran, it seems.

Symbolic AI which was a thing in the 70s-80s, ie almost 3 decades before the transformer.

I have a few AI books from that era that I've been meaning to flip through again just to get a kick out of the huge vibe shift in the field. When I remember acquiring them 15-20 years ago, it seemed it was all attempts to create bespoke approaches to each aspect of the human experience. Hand-coded language models ("verb", "subject", "object"), visual edge detection, spectrograms for speech recognition, physical world models, and so forth. Nowadays, "lolz, just throw data and compute at it" seems the preferred approach.

To be fair, the limitations of the bespoke methods were known at the time: different words can serve different roles in different contexts. Sometimes we noun verbs and verb nouns. And so on for the other spaces.

I would love to get Epsteins ability to get random super successful people to email me back.

In the academic/philanthropic space here I think it might be sufficient to just have money. Full professors at research institutions are, from having talked with several, largely focused on raising money to pay for grad students and postdocs to do the real work (and a large fraction for university overhead). Other university leadership is looking to put names on new buildings or departments. The startup/VC ecosystem seems largely similar, where I've seen founders stoop to surprising levels in the interest of closing a funding round, and only later go on to great success. These people aren't fools, but having it known that there are nine or more digits in your bank account makes them take you seriously, laugh at your dumb jokes, and tell you how great your ideas are. The comparisons to, uh, escorts (which Epstein also seemed to like hanging out with) write themselves.

From knowing a bit about the topic (very peripherally related to some research interests), I doubt it'll do as much as they propose. Existing algorithms might well work, but can't allow for an open "is this watermarked?" oracle, or it becomes fairly straightforward to start stripping watermarks, or even deriving the ways to generate them.

And if every attempt to check the watermarks requires going through Anthropic, it doesn't seem to be that much better than "is this in our set of generated artifacts?".

And that's in addition to the inherent fragility of such watermarks against edits and the amount of background data needed to hide one (I don't have a quantity in mind).

I'm legitimately curious about how attorney-client privilege here interacts with therapist-client privilege.

As tempting as it is to socialize with the people you know at a conference (fellow attendees from your organization, most likely), I'd recommend you at least make an effort to talk to someone new (daily?). It'll vary based on field, but "what is your research interest?" or "are you presenting? Tell me about your topic" are pretty evergreen. Some conferences provide meals, which I've found to be a good time for this. I'd expect many/most of these attempts to go nowhere (maybe they're a vendor of something you don't need), but the practice of talking about it and making connections is useful, and sometimes you learn something or meet someone interesting. If you want to be particularly savvy, find a paper/presentation that sounds interesting to you and try to chat with the specific author/speaker.

communists to start a total extermination race war

Can I ask the Finns, or the eastern half of Poland? Or Tibet.

I assure you that women with dance training DO do it though.

I'm not going to say the video was my favorite, but to describe it as "silly walks" probably undersells the grace in the motions. The movements are quick, fluid, and precise, and not something an untrained person flailing about could do. It looks qualitatively different from, say, the Raygun breakdancing performance (which itself was better than most could manage). I might have missed a double being composited in, but I think Qualley did at least the bulk of the shots.

Aesthetically, I'm on the fence, but at least on a technical level it's impressive.