against e/acc
the answer is i'm closer to ea, as peeps in the replies who actually knew me pointed out. its interesting so much of my following is e/acc, but im honestly not that surprised. see, i'm a dedicated lifelong techno-optimist and transhumanist. and the thing is, so are most eas ⏬
@tenobrusare u closer to ea or e/acc? // do u think i'm closer to ea or e/acc?
i made this tweet near the beginning of the year and was pretty surprised then at how many people i liked and respected thought of themselves as e/acc and expected me to do the same. since then i've realized different ppl have hugely different conceptions around what it means https://t.co/i9GsJduPQQ
from my perspective, e/acc started last year as... basically a silly countermeme. a few small accounts i was mutuals with, someone i knew irl, people floating around and shitposting on the tl and in spaces about how EAs were cringe and AI was sick
a three paragraph substack post, a lot of banger tweets, no real specific ideology, not even trying to compete with EA on their terms. their terms were long carefully thought out discussions of the right way to allocate their time and resources for the betterment of humanity...
and that shit don't play well as tweets. also, it's kinda boring. and it's a lot easier to feel the rush of just shitposting about dyson spheres.
but over the last year e/acc has gained a huge following, in some ways as an ideological competitor to ea, but also among demographics who would never have been involved in the first place. it would be pretty stupid to ignore that and dismiss it without examining what ppl like
one of the biggest parts of e/acc that ppl, myself included, have resonated with is *techno-optimism*. it's undeniably true that technology has solved a huge portion of humanities problems, uplifted us from bronze age horrors, allowed us to live lives worth living
it's undeniably true that many many times in the past, fear and delay of critical advancements caused worse suffering on a massive scale. the backlash against GMOs and vaccines have caused uncountable deaths.
climate change looks increasingly like a problem in need of a technological solution, with decades wasted trying to "decelerate" our civilization instead of simply *solving the fucking problem* (with carbon capture algae? sulfur cloud seeding? who knows, but we'll do it)
and i feel the fucking personal need for us to develop lifesaving technologies as much as anyone. my mother's cancer made me yearn deep in my bones for us to figure this shit out, no matter what, so no one would have to suffer that again.
many of us around these parts are directly or peripherally working on technologies we think will save lives or advance the world, and we think this is important *despite* the real and present risks.
i worked on self-driving cars for years, and while the risks there were very real, we did everything we could to mitigate them with the knowledge that in the end, our technological advancement would make people everywhere safer.
we need to climb the Kardashev scale, not just for ourselves but for our infinite descendants who will fill this galaxy and universe with joy
but that brings me to the second aspect of e/acc, and what i view as deeply misleading choices of framing throughout the movement.
as a countermeme to ea, it was partially a cultural rejection, of grouphouses and poly and earnestness and weird grant applications. but mostly, it was a rejection of the idea that people working on ai should have to worry about it, at all
ai safety was growing in prominence in 2022. people finally started to notice what the fuck had happened 2 years earlier. timelines were updating, dramatically. everyone who had accepted ai risk as real but discounted it as "something for the 2070s to handle" had to re-evaluate
it's one thing to accept that your work might cause issues for scientists decades down the line, it's a whole different ballgame to question whether you might be directly cause the destruction of your friends and your friends children
in many circles these ideas are easy to dismiss out of hand as ridiculous. ai is scifi, fantasy, pointless. even among prestigious ML engineers many are happy and able to compartmentalize the implications of significantly increased general intelligence capability
but others weren't, others were too deep in futurism to put on blinders, and yet still felt a need to push on full-steam-ahead. and to do that you need a different style of argument.
early on, e/acc and beff pretty much directly accepted the singularity and unalignment of artificial intelligence as inevitable. their argument didn't come from that direction. beff simply claimed this as desirable.
he reframed this as "accepting the thermodynamic god", bringing what had to happen into existence whether or not humans survived.
or how weak it will be, how little data we have, how hard it is to have an effect on the real world or make scientific discoveries
most of these arguments aren't really arguments against AI risk, rather they're attempts to argue against very specific versions of the most pessimistic MIRI-style foom scenarios...
and yet u don't need foom to see how clearly AGI can cause harm to humanity. you don't need foom or magic oracles to permafuck us and create competing highly competent species in our ecological niche
@tenobrus- many models have already achieved better than average-human scores on various benchmarks
- it is possible that continued advancements in the current paradigm will top out at "merely" providing best-in-class human expert capabilities in ~virtually all domains, due to well modeling the data we have
- if this is the case, this is in itself clearly an existential threat
- all you need to create the equivalent of an alien invasion here is a few bad human actors, and there are plenty of bad human actors
- having unconstrained numbers of non-superhuman expert agents directed at potentially arbitrary goals by anyone with compute will quickly lead to terrible places
you don't need foom or magic oracles to permafuck us and create competing highly competent species in our ecological niche
and since most of the accelerationists both accept and celebrate the transformative impact AI will have... it's disingenuous as hell to talk out of both sides of your mouth about that when your real argument is "it's morally ok that our descendants aren't human"
that's the part of e/acc that i reject. the part where you take a vibe that you really want to hold tight, and you write that as the bottom line. you don't try to really argue, or think through your positions. you just wage memetic war.
so i'm not e/acc. i'm a techno-optimist. i think technology will be how we uplift humanity to the stars, and i can't wait to be there. but we have to figure out how to make sure it's *us* who are there, and not the things that killed us.
in some ways you could view this as admirable. in others, it reads very much as a death cult. either way it's a complex position, and comes more from a place of fully embracing the dangers of SAI rather than rejecting them.
and yet... many of the people who call themselves e/acc now, or argued from "its position", didn't really come from this place. and neither does beff, in many of his arguments. there's a constant bait and switch, with people instead arguing about how *easy* AI will be to align