←
tom bombadil36,418 posts
tom bombadil avatar

tom bombadil

@Authw8

if the bartender hitting me with a shovel is wrong, i don't want to be right authw8.bsky.social he/him. born 1989

◷ Joined February 2009▧ 1,161 Photos and videos
tom bombadil avatar
tom bombadil@Authw8·21h

corollary: whatever part of a medium you do appreciate is the part you have experience in. this is why pop music still has real singing

tom bombadil avatar
tom bombadil@Authw8·21h

what's interesting about this is it (correctly imo) implies that you have to be able to make art in a medium in order to appreciate art in that medium

single player minecraft is kind of a great example of why people don't like ai art. you could see two worlds that are identical with the only difference being one was made in survival and one was made in creative and the experience of looking at it totally changes.

tom bombadil avatar
tom bombadil@Authw8·Nov 17

my model of this is that we're currently in the process of letting AI remove many chesterton fences without realizing that's what's going on

I have to say, the rolling outages & the way they're going sector by sector -- transportation, banking, crypto, refrigeration, etc. -- where they're not doing lasting damage, just temporary, pattern matches to me like we're being probed. But who or why, I truly have no clue.

tom bombadil avatar
tom bombadil@Authw8·Nov 17

people keep calling zelda "medieval fantasy" but i don't think that's really right. it's more like iron age fantasy with a grab bag of fun anachronisms

tom bombadil avatar
tom bombadil@Authw8·Nov 16

my model of totk is it's sort of retroactively subsidizing the development of botw, which i'm fine with. i'm very excited to see what comes next though

Breath of the Wild was a revolutionary step for the franchise that could easily have snowballed into us getting a third pillar of the Zelda formula. Tears of the Kingdom is reheated nachos, the game. if it wasn’t mechanically fun as hell, it would be the worst.

tom bombadil avatar
tom bombadil@Authw8·Nov 16

imagine there's a rogue superintelligence. suppose we further imagine this superintelligence is manipulating events behind the scenes. specifically, let's imagine it's able to "poison" people such that they start doing the superintelligence's bidding. someone poisoned in this way will mostly act like they did before, but they'll be secretly working to further the superintelligence's plans. imagine trying to do politics in this environment. if your goal is to oppose the rogue superintelligence, you can't just take people at their word. the enemy is always looking for new ways to compromise politicians, pundits, and activists. so you establish weird, costly rituals that only a believer would bother with. you also become very skeptical of anyone with a policy more complicated than "directly oppose the superintelligence." more complicated plans are very likely planted by the superintelligence for the purpose of undermining your movement. i think a lot of what i would call the "factional left" (people excited about bernie sanders, zohran mamdani, AES, etc) are operating in something kinda like this epistemic environment. the rogue superintelligence is "capitalism." they know they're opposed, they know the enemy dominates the information environment, so they are mistrustful of anyone who seems less than 100% committed to the movement. and even then, they are constantly vigilant for signs of memetic contagion.

tom bombadil avatar
tom bombadil@Authw8·Nov 15

i think this is my basic model of ai risk. there's definitely a bunch of dry tinder to be burned. and it's definitely the case that under the right circumstances, ai could become self-improving and consume most of that dry tinder very quickly. but what i feel uncertain of is: - i don't know conditions are necessary for the fire to start - i don't know how much tinder there actually is to burn that leaves me with these scenarios: - the dog that never barks. rsi just never happens. ai plateaus around a human level of capability on most high-level tasks - the little boom. rsi kicks off, gets super smart, but plateaus at a comprehensible spot above human intelligence - the big boom. rsi kicks off and goes somewhere fundamentally incomprehensible to human minds. this is probably bad for humanity unless we get very lucky i feel like either of these booms could kick off basically any time. they might never kick off, but they both feel pretty possible.

i think part of the confusion is that a very old assumption has been that there's a tipping point where intelligence becomes self-improving. if that assumption holds, it means there's really only one stable state for artificial intelligence, which is wherever diminishing returns bite harder than recursive self-improvement is able to overcome. but nobody knows where that line is. it's sort of like everyone knows there's a bunch of dry tinder ready to light up under the right circumstances, but nobody's sure what those circumstances are or how much tinder is actually there to burn

tom bombadil avatar
tom bombadil@Authw8·Nov 15

to me, the essence of seriousness is in the willingness to accept trade-offs

Sorry guys but if you're actually serious about wanting to defeat climate change then you're going to have to start getting comfortable with telling the tribes "no" when they ask for an exemption on climate legislation.

tom bombadil avatar
tom bombadil@Authw8·Nov 13

i find for myself there's sometimes a distinction between thinking about something vs Actually Thinking about something. thinking about something is normal and unremarkable. but Actually Thinking about something is unpleasant and rare. in order to Actually Think about something, it's necessary for me to put myself in this sort of headspace: 1) there is an objectively right answer to the question, and the lives of people i care about hinge on getting the objectively correct answer 2) nobody will ever know the answer i land on except me whenever i sense that someone has never gone through this exercise or anything like it on the topic at hand, it's hard to take them very seriously

tom bombadil avatar
tom bombadil@Authw8·Nov 13

unironically i think ending the filibuster would be one of the most impactful long-term moves we could make to fix US politics. i think many problems in US politics have roots in the way the filibuster makes it impossible to win an election and implement a straightforward agenda. supermajority rules naturally favor backbenchers and bomb-throwers over party leaders. if you lean libertarian and like small government, i think ending the filibuster will get you more of what you want. right now the bad laws are already on the books. getting rid of the filibuster makes it easier to repeal or rewrite bad laws. but even if you don't lean that way, i think ending the filibuster is likely to produce positive results all around. i think we underestimate the impact of the filibuster because it's become the water we swim in. but it's actually really bad to kneecap basic lawmaking in a time of rapid technological and cultural change.

I think Senate Democrats should show they’re willing to work with Donald Trump by embracing his proposal to end the filibuster #bipartisanship

Post attachment
Jeff Maurer avatar
⟳ tom bombadil retweeted
Jeff Maurer@JeffMightBWrong·Nov 12

@MattBruenig, I can’t tell you how pumped I am to see your statistical model that corrects for state lean in outcomes of competitive Senate races! When I published my finding last week that leftist Democratic Senate candidates underperformed non-leftists, you popped into the thread to declare my finding “not valid”. Obviously, I was shattered to fail to win the approval of you, the most exalted data scientist who has ever lived or ever will live. Your objection was that my model failed to account for state lean. That is: A candidate in a bluer state might find it harder to over-perform than a candidate in a redder state. So, I ran a new analysis that corrected for state lean and found that leftists still underperform. You dismissed that, too. And that’s why I offered to buy you a steak dinner if you could pretty please produce a statistical model that corrects for state lean and fundamentally alters my finding that leftist Senate candidates underperformed in competitive races. Please, Matt, impart your wisdom on us! I am but a humble comedian, and I lack the God-like abilities of you, the most powerful Data Sorcerer in the universe. And allow me to fill in the context, which will be tricky, because Twitter seems to have a weird glitch where every time you and I have an exchange, it gets deleted. That’s unfortunate, because it might look to some like you’re jumping into conversations to do your Data Batman routine and then destroying the evidence when the conversation turns against you. It *could* appear that way, though I’m sure that it’s just some bad code on Twitter’s end. When I found that leftist Democratic Senate candidates (my proxy for “leftist” was “Bernie-endorsed”) underperformed other Democrats in competitive races, you declared that finding “not valid”. You noted that the leftist candidates come from bluer states on average, and declared the partisan lean effect to be so potent that I had wasted everyone’s time. You even bluntly informed me that “it’s math”, and then proceeded to dazzle me and any onlookers with some vintage Bruenig number-ifying! I was honored to witness it! Of course, the math was garbage, because it 1) Assumed that the electorate stays fixed from one election to another, which is obviously false, and 2) Precluded the existence of any other factors. But, utter horseshit aside, it was awe-inspiring to see the master at work! And here’s where I really screwed up: I attempted to take your argument seriously and do an analysis that corrected for state lean. How hubristic of me, because surely you have already performed that analysis! After all: You spoke with such CERTAINTY, you were POSITIVE that the state lean effect was big enough to render my finding…what was your phrasing again?...ah yes: “Not valid.” Obviously, you would only feel that way if you already possessed sound statistics supporting your view! Otherwise, you would just be opining “from the gut”, i.e. exactly the same thing that you repeatedly accuse others of doing. So, really, I’m not asking you to *perform* an analysis; I’m asking you to please *share* the analysis that you have surely already performed! After all: If you hadn’t already done that analysis, then you would be a hypocrite talking out of your ass, which I know that you are not. I ran a regression to see if there was a correlation between a state’s partisan lean and the margin the Democratic Senate candidate achieved in that state. There was a moderate correlation, so the question became: How big of a correction should be made for that effect? I made what I considered a very big correction – 1.4 “bonus” or “penalty” points for each point of partisan lean. I was trying to give your argument every benefit of the doubt – after all, you were CERTAIN that the effect was huge, and when an Analytics Prophet like you declares the existence of an effect, a shit-covered peasant like me must leave no stone unturned in pursuit of that truth! Alas, even that gigantic correction did not save the Bernie-backed candidates: Leftists in competitive races still underperformed normies by an average of six points. I published this finding. You appeared in the Twitter thread to accuse me of “throwing out” the finding, which was odd, because you were commenting in the very thread in which the finding was published. Also, the results did not contradict my basic thesis that leftists underperformed. At any rate, you declared me not worthy of performing further data analysis, and I was shattered…shattered because you were RIGHT. I am not worthy to breathe the same air as you, The Great Khalessi Of Data, Exalted Be Thy Name! And that’s why I implore you to bestow on us the analysis that shows that when a sound correction for partisan lean is included, leftist Senate candidates do not underperform in competitive races. I confess: I am too ignorant to even *conceive* of how you will do this! Because I ran the analysis with different levels of correction for partisan lean, and leftists underperformed at every spot on the dial. The problem is that the four Bernie-backed candidates come from states that aren’t *that* different from the rest of the sample. If the “correction” dial gets turned up, Flanagan does better and Lee does worse; you turn the dial down, the opposite happens. I could get the gap down to two by turning the dial up an absurd amount, but then I was simply measuring each state’s partisan lean: The Pearson’s r between candidate performance and state lean in that version was .94, which a Stats Oracle like you will know is sometimes called a “perfect” correlation. And when the dial was turned way up, the model did weird things. I noted this and commented that I think a subtler effect is closer to reality, and that’s why you accused me of “throwing out” data. But even the “1.4 bonus points” model – in which leftists underperformed by six – caused partisan lean to wash out all other effects: All of the “worst” performers came from very red states and the “best” ones were in blue states (Pearson r = .87). Ten of 34 Senate candidates failed to over-perform Kamala on a night when the average performance was +16. And the Democrats’ breakout star was Rhode Island’s Jack Reed, a name I rarely hear mentioned in response to the question “Who is this generation’s FDR?” But I am so chastised by your accusation that I “threw out” data that if you declare this version of the model to be the preferred one, then I will agree it is so! Leftists underperformed by six – Bruenig hath declared it! And as giddy as I am to receive your data analysis, I also throbbingly await your article about how Jack Reed has ascended to a higher plane of political performance and we must all embrace Reed-ism. I encourage you to act quickly. Because, in the absence of empirical evidence supporting your confidently stated beliefs, you *could* begin to look like a bullshit artist who uses numerical hand-waving to obscure reality. It could appear to some that your schtick is to parachute into any conversation that contains a finding that leftists might find uncomfortable, raise a statistical quibble – which is usually possible, because social science is imperfect – and therefore give leftists psychological cover to disregard the finding. Therefore, your function is to help leftists persist in their delusions…is how things *could* look to your critics. But boy will they have egg on their faces when you produce that statistical correction! And I apologize profusely for annoying you with my ignorance. You are a God surrounded by mortals, and it must be frustrating to watch us struggle to perceive what’s in front of our nose while the entire universe dances to the subtle movements of your fingertips. When this analysis arrives, I pray that I will be able to comprehend it, and not stare in puzzlement like a dog trying to read Proust. Alas: The latter case will probably be true! Because I cannot conceive of how your arguments in this case could possibly be sound, and I pray that you will deign to enlighten me.

tom bombadil avatar
tom bombadil@Authw8·Nov 9

there are two basic skills you need to get by in the modern world. you need to be literate and you need to be numerate. most people are neither. some people are one or the other. few people are both, but they run the world

tom bombadil avatar
tom bombadil@Authw8·Nov 7

my prior on this is that origin myths from vague places nobody cares about (at the time written down) are probably more likely to be true than origin myths from super famous places (at the time written down)

Interestingly, the Goths' mythic origin seems to be mostly true, but that of the Romans (mythical Trojan or scholarly Aeolic) mostly false. Not every barbarian tribe's mythic origin checks out in the light of information from modern archaeology, linguistics, and genetics.

tom bombadil avatar
tom bombadil@Authw8·Nov 7

more people need to internalize to their world model that cultural homogenization happens AFTER state consolidation, not before it. states are born through conquest and semi-arbitrary line-drawing exercises that then harden into social reality over generations

these "anti colonial" maps always piss me off because 90% of these fake countries would be completey ungovernable on their own. how is a city state in the congo jungle supposed to function in the modern day. european straight borders are a problem but this is too much

tom bombadil avatar
tom bombadil@Authw8·Nov 6

trans kids in sports is an inherently cursed discourse. in america a, sports are often understood mostly as an enriching children's activity. lots of parents are paying so their kid can participate in various random sport activities at various levels of seriousness. denying trans kids these enrichment opportunities feels mean-spirited. america b sees sports as a serious competition. the point of sports isn't to enrich every kid but to let the best kids prove it in a fair way. any perceived unfairness calls into question the value of the whole activity. my own view is that individual leagues need to work out their own rules that make sense for whatever type of thing they're trying to be. i think serious people on both sides tend to agree with me on this. but neither side trusts the other to oversee the process because they are worried about totally different victim classes. both sides can say "but think of the children!" and mean it from their heart.