t

efficiency is begging for full on class conflict. they are burning the puppet show that kept them in comfortably in power

the only thing worse than an accelerationist is one who does it on accident, i feel like im watching a trainwreck in slow motion

call me a cynic but the government exists in large part to insulate the ruling class from the rest of em. the bulky and inefficient bits allow us to maintain some illusion of an egalitarian democracy and still slouch towards a slightly better future. getting rid of that illusion in the name of

"surely you're not advocating fo-"
again, no, but it's not hard to imagine that a lot of people are looking at this thinking they have no other choice, the gov just colluded w the courts to screw them over

helpful voice in the back: "a class action lawsuit!"
thats what Mass vs EPA was, and everyone just watched an industry insider buy his way into a position of power that afforded him the ability to overturn it

and since free market incentives don't exist to reign in CO2 emissions due to the diluted, global nature of their impacts, the public might feel left to enforce the costs back onto producers themselves... gee I wonder what that looks like

hmmm not endorsing, just observing here. but the DOE throwing the kitchen sink at delisting CO2 from the Clean Air Act feels a little destabilizing

regulatory control is the public's means of addressing negative externalities, eliminating that begs for those costs to then be settled out of court

ooof I'm actually [redacted] and based this number off nameplate capacity and not annual output, actual number is much closer to 10 W/m2...

there are some plants running higher than the 15 W/m2 limit proposed by Smil today in the UAE, but still bad look dude. dont do math angry

"but what about weather dependency? and nighttime??"

batteries are pretty cool tech here and seem to be improving in density and dropping cost too... rmi.org/the-rise-of-...

again not discussed in the BHL report

Chris however wrote the BHL report in 2024 and winds up making representative claims about solar using data so far out of date that it ignores a major industry inflection point

which doesn't feel like good faith argument imo, but who am I to say

Grid Storage: Batteries Will Win A short and spicy post. There remains, even in 2023, a substantial fraction of the “future of energy” hivemind who are still convinced that the solution to all our problems is to build …

(side note you should read all of Handmer's work on batteries and grid storage: caseyhandmer.wordpress.com/2023/07/12/g... )

in Smil's defense, he published his paper on energy densities right when solar manufacturing broke into dedicated silicone manufacturing for PV cells, which nearly doubled learning rates in the field. tough time to make predictions

from Casey Handmer: caseyhandmer.wordpress.com/2024/11/09/s...

which seems a little odd in light of the fact that there are plants running close to 43 W/m2 in Mexico right now
www.power-technology.com/data-insight...

if we were to ask Smil in 2010 what todays densities would look like, he'd suggest ~15 W/m2 (a small improvement) and call even this unlikely (see quote from paper). Chris' use of this data in his 2024 book would suggest he agrees with that claim even today?

back then Smil came up w an average solar density of 10 W/m2, which
a) is a global average rather than regional and
b) doesn't really line up the 1.21 homes / acre cited in the BHL figure using 2024 EIA home energy estimates of 10.8 kWh / year (someone check my math?)

but whatever...

Fig 1.3 and 1.4 cite data from 2015 and 2022, and Fig 1.2 doesn't have a date. The Vaclav Smil paper it cites is available online tho, published in... 2010. Old data isnt exactly a crime, buuut this is "Energy Systems of *Today*" vaclavsmil.com/wp-content/u...

The section titled "Energy Systems Today" harps on renewable intermittency leading to higher energy prices, and bemoans the need for significant land use due to low energy density. a few figures are cited

Bettering Human Lives 2024 - Liberty Energy Liberty Energy has updated and expanded our Bettering Human Lives report. Issued in January 2024, the third edition of this report contains an in-depth look at

report is available here btw: libertyenergy.com/resource-lib...

a while back I was reviewing the 2024 Bettering Human Lives report, the author of which is now the head of the DOE. its still sitting in a draft somewhere, but I want to nitpick some of the bits on renewable energy for a minute

my two most inflammatory opinions appear to be
“killing people is bad” and
“there are consequences for electing stupid people”
regardless of audience i am going insane

summer climbing conditions out there

(it was v good and you should read it (if you want to))

hearing these are now collectors items

i respect you sticking to your luddite guns. something noble in the conviction even if its wrong :)

anyways finally this all comes full circle to me thinking ai demand for power is only going up, and if you care about emissions not also going up your level of urgency should be increasing proportionately

so uh. sko nuclear and solar panels

all of this could be null if adoption (and subsequent inference costs) get outpaced by power plant construction and interconnection, but i think thats unlikely

short term estimates should reflect the same inference growth rate, and the size of each new connection needs headroom bc nobody wants to build whole new data centers every time a larger model drops. they’ll all (want to) be overbuilt

bc of this, long term grid planning should expect to expand capacity at the same rate we can come up w problems to apply intelligence to… or in other words as fast as we can build power plants

**if someone has good arguments for this not being the case, i.e. that there are things permanently beyond human comprehension, i could probably be convinced otherwise

*i think the assumption that we wont see decreasing returns on the application of intelligence itself does a lot of work here

i currently fall into the David Deutsch camp of “general intelligence is a universal explainer in a causal universe”, so more intelligence will always be useful**

gains in efficiency lower these per query costs(see deepseek etc), but the overall demand for intelligence probably isnt decreased. we always* have more things to apply more intelligence to. yay, jevons paradox…