Ryan Sepassi

why is there no national office for open-source. countries should be convening and making agreements to ape in together like they do for other global issues like climate change. $1bn/yr would be peanuts compared to NSF $10bn/yr and would have an impact even assuming 90% wasted

This poem by Maggie Smith doubles as good parenting advice. Research finds that when parents teach their kids the world is a bad place, it doesn’t protect them from harm, but leads to “dramatically less psychological flourishing." & worse mental outcomes: tandfonline.com/doi/full/10.10…

@douglas_eck @JFPuget @GaryMarcus @ylecun gnat*

@JFPuget @GaryMarcus @ylecun This is the role, for good and bad, of a gadfly. A gadfly gets up in your face and questions things.

My take on @GaryMarcus vs @ylecun : I muted and unfollowed and Marcus. Yann is sharing stuff he or his team is doing, while Marcus only share criticisms. Even though some criticism may look right, it would be so much better if Marcus was showing us something that works.

@ilyasut bold move coming out against tolstoy: “all loving agis are alike, but every hateful agi is hateful in its own way.”

It’s not a bad world if every AGI loves humanity in its own way

so they’ve taken over the judiciary, state legislatures, have wildly disproportionate sway in the house and senate, and now they want to take over the military. legal doesn’t mean just. how do we defend our democracy from these psychopaths?

Just in case you are still not informed. Ukraine has launched the greatest counteroffensive since World War II. Ukraine has regained/liberated over 1,000 km of land & cities. Ukraine is winning. Ukraine is defeating Russia (once thought to be the second most powerful military)!

@antonio_paglino for me at least i would love to be able to make it a collective endeavor, including all, but regimes are real (for better or worse) and so we’re left with the thorniness of geopolitics. the people of russia are not attacking ukraine, but “russia” is.

@rsepassi Russia and China are made of humans and individuals who are intimately connected with all life on earth. Old marshall plan is what has created this geopolitical conflict in the first place and you want to double down?

this is awesome. maybe US could help on energy with some sort of green neo-marshall plan. good for us, good for europe, bad for russia/china.

@douglas_eck you’re right, biden’s on a roll, let him ride, come back for ‘28 and ‘32. do term limits apply to single AIs or all AIs? 😉

From the department of bad ideas:

@geoffreyirving how do you apply moral uncertainty to the utilitarianism of EA (and not just the time horizon)? should we also be (more explicitly) hedging our bets there?

Tweet motivated both by the general sentiment that people think X-risk folk don't think near-term causes are important, and a specific conversation where someone was surprised I didn't think AI safety should be most of EA.

As someone who works on AGI safety and is motivated by longtermism (and was prior to EA being a thing): I don't think we should treat longtermism as correct with high confidence. Moral uncertainty should mean we put lots of resources towards helping the present as well.

@douglas_eck the bitter lesson is indeed bitter

@rsepassi "Large spam" is brilliant. I'm not anti large model, but it's not something I'm totally happy with. Taps hood of car. "I can make this puppy faster by a factor of 2 with only exponentially more fuel." What a deal! (Ok power law not exp, but still... :)

Just to keep things interesting, let's call them "Large Sequence Prediction Models." It's better than the Asimov-scifi-vibed (and grandiose) "Foundation Models. Don't worry, "LSPM" won't confuse anyone. /s

@DynamicWebPaige is this an announcement that you’re going back to microsoft to work on vs code and github??? :)

The ML community has distressingly little source code (and even more troublingly: little developer context data), compared to text. 😔 Improving telemetry in IDEs, and creating delightful developer tools (ML-enabled/general) feels existential for the creation of code-y LLMs.

@bryan_caplan wouldn’t an increasing minimum wage change investment in technology to increase productivity? also part of the textbook, wonder how 13yo would weigh it.

Took me under 5 minutes to turn a normal, smart 13-year-old against the minimum wage. Contrary to almost everyone, the textbook argument IS intuitive. It's just emotionally unappealing.

@mattyglesias seeing a lot of explanations that don’t account for the 2012 divergence. an overt misogynist being elected president seems relevant.

Kind of remarkable.

@LolOverruled 💯 10B people at 30/hr for 2k hours/yr is 624T. 2021 global GDP 96T. Need at least 6.5x “growth”, 3.8% real growth rate over 50 years. Do more with less, sustainably. It’s doable.

i’m a communist for land/natresources and a capitalist for everything else. neo-georgism is the way.

fuck clarence thomas. the other ones too.