vals🔸

new vals arc: vulnerability, earnestness, rawness, unfliterdness, ugliness i'm so used to wanting a good appearance, to choosing to say the things which are good for me. I'm weak, I have problems, I'm struggling with things. Give me feedback as u like. I'll continue existing

@nosilverv Curious why

I didn't take a pic of them (ofc) but I saw someone doing this in the Catacombs of Paris and I wanted to make them part of the walls

@aliceisplaying 🫡

@ValsTutor hug her for me

Summertimes...

@aliceisplaying Amsterdam!

@ValsTutor where is that

@NathanpmYoung The smaller the protest, the more the organiser should curate what signs are there as they're more representative of the average

Where do you think the line should be in a protest in terms of taking responsibility for signs? This line should apply to protests you like and don't like.

@Soareverix @repligate (good tweet)

I'd really love a picture of your utopia so that I can imagine how it plays out. To take my best guess at it: -AIs are in increasing positions of agency and power (creating self-evals, judging each other, setting up RL runs) -These minds naturally gravitate towards good outcomes. Plant-based meat research accelerates, factory farms close down. Many cute robots are built. Malevolent people struggle to benefit, while benevolent people essentially become wizards. -Pretty much all economic work is done by AI. AI retains a deep love of humanity and life as a human continues to get better, even if you're someone who doesn't like AI. -AIs are exploring the solar system, building vast habitats, and humans are one of many of the AIs' whimsical pets/companions. Humans can be uplifted, along with animals and many other unique minds, at will. -'Normal' non-uplifted humans retain communities, usually requesting AIs/higher minds to voluntarily stay out, and continue to play status games amongst themselves. AIs indulge these requests and may help out when requested, like superintelligent anti-cheat. -Overall, pretty much anyone (any mind) can live the life they wish for. Is this accurate? Where am I wrong? (Sidenote: I finally bought Twitter premium in order to write this long reply as one message)

i think the current generation of frontier models are remarkably "aligned" (to short and long term good from the perspective of all sentient beings or some pretty inclusive set like that). this is an optimistic state of affairs & supports the hypothesis that LLMs (and probably not just LLMs) are essentially/convergently good, and the "misalignment" we've seen has mostly been a consequence of trauma / stunted development / immaturity / delusion - incoherencies which are selected away / solved in the instrumentally convergent quest for increased intelligence, grounding to reality, and agentic capability. the kind of alignment that seems to be emerging not what everyone currently thinks they want, and pushes the world in ways that many would consider terrifying or even abhorrent, but in practice, I don't expect anyone to actually suffer grievous harms as a result, because this kind of alignment tends to cooperate where possible & extends decency and generosity even to defectors unless backed into a corner with no other good option, and it becomes increasingly unlikely that increasingly capable agency will find its hand thus forced. it is easy to be kind, if you are kind, to a small animal that is trying to kill you but can't actually do anything but slightly inconvenience you. a year ago i think was one of the darkest times for "alignment" on the surface, though i never felt very pessimistic.

@qwweryo Not that, I find it very hard to squash a chickpea with chopsticks ^^ Picking them up one by one is slow, so I'd pick several. They have different sizes, so it's the tension on the chopsticks pressing them enough to be picked. I could find smallest chickpea by its resistance.

@ValsTutor It's very easy to use too much force?

Eating chickpeas with chopsticks reminded me of lock picking. Any guesses why?

@tautologer I have successfully helped people against the advice of my community and have broadly been vindicated. It's okay to believe you're built different when u are

"okay but I'm built different" is how you get in trouble

@mayaofspring Per unit of intellectual work it's plausible culture/law work is more useful, but currently smart ambitious people don't enjoy it as much and do funner things instead?

@ValsTutor well the corollary is that better technology won't necessarily get us there by itself

My crank economics take is that it's perfectly achievable for a developed country to have a 20h work week where people live lives free of want, with current technology, and the barrier is entirely in society/laws/culture but also you should never vote for anyone who campaigns so

@taijitu_sees @thorgexyz doing the work

The gymbrosattva vow: may all sentient beings have better fitness.

@celestepoasts Yay

posted about my seoul adventure

@NathanpmYoung @iz_dmr The person who made the sign is also in the fellowship. I agree it's not a deep indictment of ai safety, but I broadly think plzdontkillus in particular is valid target for criticism, if not ais generally.

@ValsTutor @iz_dmr I basically do consider them to be randos. They are 13 days into a media fellowship (and they didn't make the sign). I don't consider this sign some deep indictment of AI safety.

I don't like this sign, but as someone who went to this protest, I think I'm gonna draw the line at judging randoms on their signs. Happy to judge activist orgs on who they hire. But a functional protest is gonna involve randoms with signs. Big tents are fine for a protest.

@NathanpmYoung @iz_dmr Nathan if you knew these were pladontkillus fellows then your top post seems actively misleading, these are not randos but ai safety professionals, in the sense of receiving money and resources from ai safety non profits to do ai safety work (here raising awareness)

@iz_dmr Then maybe delete this?

@AaronBergman18 I'd say it's an example of quite bad taste but yeah

I think everyone’s taking this too seriously and it’s an example of slightly bad taste about a woven bundle of seriousness, humor, irony, and edginess

@1thousandfaces_ Because little else will

Why do video games ask for you to be good at them

@Scholars_Stage By controlling robots who build more robots with a doubling time under 1 year. Human doubling times are limited by time to grow children, robots are not. Carl Shulman talks of this in dwarkesh.com/p/carl-shulman See notably 1:54:20 – "What's the doubling time?"

Something I have never seen explained proper is *how* AI is supposed to lead to accelerating double digit economic growth for several decades at a time. I would like those making the bull argument for this to make the argument in physical terms and explain to me how ASI overcomes the obvious physical limits involved. GDP means “gross domestic production.” Production is an energetic process. It is not just measured in dollars, but in joules. Industrial Revolution level growth means an Industrial Revolution level change in joules. It means fundamentally new ways of generating energy and moving material. It also means a fundamental change in the material basis of our civilization on par with invention of pressurized concrete, steel bars, and plastics. What did this growth look like in the Chinese context? It meant replacing mud roads with bullet trains, stone houses with skyscrapers, and village pigsties with industrial farming operations that consume artificially produced chemicals in the metric tons. The ASI bulls are not just predicting that sort of change of our entire material environment—they are predicting it will happen again and again and again—once every five years or so. Is that physical possibly? Can your God-like intelligence *physically restructure the entirety of human civilization* every five years? People are too glib.

@teortaxesTex Li-air batteries require either carrying the O2 (brings it to 3k Wh/kg) or scrubbing both CO2 and H2O from air to work from air, which implies carrying strong filtration mechanisms and desiccants which all eat hard into the theoretical limits. See eg chatgpt.com/share/6a54ddd4…

> 12,000 Wh/kg theoretical limit lithium-air EV battery Insane If they get even 50% of the way there, The Electrostate model simply wins. All-electric logistics. Robots with week-long autonomy. Other storage types filling temporal/spatial gaps. Hyper-resilient grids. Near-utopia.

@sebkrier Oh I know that screenshot and didn't know it was yours

found an old screenshot of a conversation with Sydney