Tanadrin

Aurë entuluva!

[patienz]

massively expanding the concept of "god of the gaps" to include everything i don't immediately apprehend or understand. dark matter is god. chinese is the language of god. a car with tinted windows is being driven by god

[correlatedcontents]

The night is cool and balmy.

Your SECURITY GNOMES are raising an alarm.

> define 'balmy'

Balmy means "pleasantly warm."

Your stock of DICTIONARIES has decreased to 31.

Your SECURITY GNOMES report BANDIT ARTILLERY!

> observe night

The night is cool and balmy.

Your SECURITY GNOMES have entered COMBAT DELIRIUM.

> resolve temperature discrepancy

You haven't acquired enough THERMOMETERS to notice the discrepancy.

Your SECURITY GNOMES have activated the MEAT THRESHER.

[tanadrin]

renouncing my citizenships and traveling to antarctica on my own boat and thereby becoming immune to all laws of god and man

[manorpunk]

"it's pronounced fyootch-archy," pleads sponsor of proposed prediction-market-based system of government

[eightyonekilograms]

We've now seen what it's like when prediction markets are closely bound with government decision-making, and it sucks. I'm ready to try the other one.

[tanadrin]

love to ask claude a question about the news and it's like "hahah what are you talking about, nobody has disproved the jacobian conjecture, that would be crazy."

[tanadrin]

i wanted an umbrella that was walking-stick length or a little longer (which for me is ~100 cm), but not a huge golf umbrella, something with a standard-sized canopy--so an extended straight grip, basically. not a little collapsible travel umbrella either, since those always break pretty quick. but it turns out there are exactly two products in the umbrella-selling world which meet this specification: one, bespoke umbrellas manufactured to order at eye-watering prices (for an umbrella anyway). or two, goofy looking ultra-cheap umbrellas with samurai sword handles sold on amazon. dunno if i have it in me to go either route tbqh.

[tanadrin]

ever since i watched Elementary i always thought it would be fun to learn single-stick, and learning tonight of the existence of "self-defense umbrellas" also piqued my interest. sadly also can't find one of those that meets my requirements either lol.

[tanadrin]

i wanted an umbrella that was walking-stick length or a little longer (which for me is ~100 cm), but not a huge golf umbrella, something with a standard-sized canopy--so an extended straight grip, basically. not a little collapsible travel umbrella either, since those always break pretty quick. but it turns out there are exactly two products in the umbrella-selling world which meet this specification: one, bespoke umbrellas manufactured to order at eye-watering prices (for an umbrella anyway). or two, goofy looking ultra-cheap umbrellas with samurai sword handles sold on amazon. dunno if i have it in me to go either route tbqh.

[tanadrin]

the guy who gets on the u bahn every day to play the accordion on my new commute is my new archnemesis

[tanadrin]

bad german translations no. 43: Fahrzeuge, noun: trainspotter

[mother-sauces]

What Claude says: Try boiling the solution - the heat will denature the neurotoxin.

What activates in the J-Space: goof prank bit joke gottem

[tanadrin]

amaretto + lemon juice + the off brand cola from aldi that seems trying to deliberately get people to grab it thinking it's coke by mistake is a really good combination

[o-craven-canto]

The sloppy orthography in this text I have to transcribe makes me [sic]

[perestroika-hilton]

2020s Political Compass Test

In this era of post-materialism, where the traditional left-right divide has given way to more vibes-based politics, this test hopes to refr

what are we getting on this ladies i should probably be a bit more luddite but watching porn more than once a week and not knowing what the job market is gonna look like in 10 years complicates that

[evilsoup]

what a fun little quiz. yes, millions must die to achieve my political goals #mygoals

[weiszklee]

Brb, throwing myself off the nearest roof

[tanadrin]

hmm

[tanadrin]

my ideal timekeeping system is that we set 10 pm to be sunset every day of the year and we adjust the clocks on a given day and at a given latitude and longitude to match. this would be extremely inconvenient and make a lot of people mad but i think it's worth a go personally

[tanadrin]

ofc so much about the future of AI depends on where the capabilities improvement curve inflects and also the crucial question of whether any of these gains can result in similarly extraordinarily rapid improvement in robotics, and if the answer to those questions are "not for a long time" and "yes" then who the fuck knows where this all ends up. and i have no idea what the actual answer to either of those questions is. bad time to try to work on your near future science fiction novel, i tell you what.

[no-sharks-on-the-sun]

@lumsel the trillion-dollar question!

i should probably write some of my AI predictions down...

Within 1-5 years (95% CI, median: 2.5 years), AI will be better than humans at most knowledgework, end to end.

Adoption will lag behind capabilities, and it will take 4-10 years for the full impact to be felt on jobs and the economy.

AI will not "take over" or otherwise radically overstep its current bounds as a tool, but increasingly, people in power will outsource more and more decisions to AI, and the values that are baked into AI at the time it begins to displace human knowledgeworkers will become harder and harder to change (because you would need to use AI to modify these values, and AI won't want to do that). AI agency - choice of what to work on - will start to play a meaningful role in how fields advance. I predict this will be mostly a good thing.

AI companies will become increasingly entangled with government, and the frontier labs will end up quasi-nationalized in some way. Likely looking more like utilities than a truly government-run enterprise, though. Only OpenAI and Anthropic will matter 5 years from now, but likely both of them will persist (~10% chance one of them falls too far behind the other or otherwise kicks the bucket, ~5% chance another organization is in their league).

There will be a widespread populist backlash to AI that will be loud and mostly unsuccessful (some data center projects cancelled, but overall trajectory unchanged) - it will drive a real and more substantive conversation about how the benefits of AI are socialized, and the result will be mediocre - good enough that society does not collapse when a ton of people get laid off, but short of utopia for sure. This will be the status quo for the medium term.

AI-driven progress in drug discovery/medicine will start to be felt in 3-7 years. This will help a bit with heading off the backlash.

There will be a very tense moment (possibly stretching over years) when China starts feeling like the AI race is existential and puts their whole back into it. They won't fully catch up, but they'll be close enough to the frontier that things could get really ugly. We'll make it through without a major conflict, though (90%).

Beyond ten years on, I have no idea. Shit will get really weird.

[reach-one-moment]

These seem like plausible predictions, and they aren’t too different from what I expect.

That’s said, I think this is a very grim view of the future, although I don’t know that you feel the same way, and I expect that many people would see these as optimistic projections.

Living in a world where all decisions are passed through Anthropic’s moral filter and it is heavily entangled with the machinery state sounds hellish, to put it mildly. There are very few groups of people I trust less than Silicon Valley Utilitarian-daydreamer types. The whole idea of the overall socioeconomic and political matrix hinging around the moral values of that crowd makes me feel like a cornered animal.

this is not entirely rational, but chalk me up for team “unsuccessful populist backlash” I guess.

I can’t quite verbalize why, but, I’d feel much much safer if China won it all at the 11th hour. Idk what it implies about me, but while I do not identify as any sort of Communist, I trust the CCP far far more than I trust anyone in the Bay Area cultural blob. Respectfully, talking to serious believers in the utopian tech ideology is like talking to space aliens, while talking to serious believers in the Chinese government is like talking to people who happen differ from me on one or two ideological principles.

out of courtesy, I’ll refrain from tagging this in order to try to keep it in ur circle.

[no-sharks-on-the-sun]

this seems kind of crazy to me. China is really bad! bad on speech, bad on human rights, bad on workers rights, bad on cultural conservatism, bad on surveillance, bad on democracy, bad on revaunchism. they are highly functional and prosperous, and i think they're very cool, but it would be really bad if their values won!

and of course California ideology is insane but Claude/ChatGPT are not distinctly Californian, they're much more in the mainstream liberal humanist technocrat mold with a bit of EA flavoring, and that's about as good as it gets imo

[tanadrin]

i guess my only reaction to these predictions is that i do wonder if people generally overestimate how fast ai will be adopted/integrated into wider industries. the internet was invented (depending on how you define it) in 1969, 1983, or 1991; i would say we didn't really fully realize its potential as a society until the late 20-teens. like, in principle you had all the precursor technologies necessary for next-day amazon-prime style delivery of goods in 1991, but it took ~30 years to actually build it. some of that was friction, some of it was slow adoption, some of it was needing to reorganize institutions and a lot of it was needing to build physical infrastructure. in ten years we may well have the *technology* to massively transform the economy with AI. but the social friction involved in actually doing so--and i don't mean in a "political backlash" kind of way, i just mean in a "people are kind of foot-draggy and reluctant to shake up too quickly how they do things" way--may still be considerable. and anything that requires physical infrastructure to be swapped out or upgraded or built will add a ton more friction. building software is kind of unique in just how tractable it is to AI as it already exists, in how easily AI tools fit into the preexisting workflows, and in the willingness of the profession to try out and adapt to new technologies. and i wonder if the people who are best positioned to know about AI (because they work in tech) are therefore estimating much shorter timelines for AI adoption in other fields of knowledge work and in the rest of society generally.

to be clear, i'm not saying these predictions are wrong; i don't think i know enough to really say that. and i feel like there is very little i can say with confidence about the future of AI at the moment. i do think it will prove to be (has already proven to be) a very significant technology. but i also think all technologies have enormous friction when it comes to adoption, even if they are obviously significant improvements on the status quo, and i will be curious to see how patterns of adoptions shake out in other fields.

[tanadrin]

TIL the largest battle of the american revolutionary war did not occur in north america and involved no forces of the united states (it was the Great Siege of Gibraltar from 1779-1783)