This is going to sound hopelessly pedestrian relative to the galaxy-brained ideas outline here, but something we *could* do in the short term would be to just try to fix the primary system. Everything you're saying about democracy is entirely true, but a massive accelerant to all of the bad stuff is the reality that the average voter ends up with choices selected by the most insane 10% of the voter base.
In my experience, the people who are really into politics--i.e. the people who vote in primaries--are extremely screwed up people. They're on average smarter and better-informed than the mean voter, but because they're demented, they end up choosing outrageous options. Seriously, think about it: off all the smart, interesting, well-acclimatized and generally happy people you know, how many are really into politics?
And then there's the issue that, within any political institution, the way to prestige is to out-extreme the next-most-extreme person. And THEN you add in Sunstein polarization (where like-minded members debating something gradually move the entire group to a position more extreme than the most extreme person when it began), and you have . . . well modern American politics. Obviously, a fluid information environment supercharges this.
Politicians have figured out that the way to these extremist voters' hearts is negative polarization, so they're not even _lying_ about solving problems anymore! It's just "eat the rich" and "make the libs cry" all the way down; our leaders aren't even gesturing at solving problems. It was actually BETTER when politicians lied about things like "higher taxes on the rich mean everyone gets amazing, free everything forever" or "lower taxes means 25% annual economic growth forever!" At least the goal was something positive.
But, if we could deploy something like ranked choice voting at scale (or just go back to old smoke-filled rooms where party elites chose their candidates), it would be a huge step in the right direction. It still wouldn't be as good as idealized non-democratic options, but it could at least take us back to the mid-20th century when we had leaders like Calvin Coolidge get elected every so often.
> Seriously, think about it: off all the smart, interesting, well-acclimatized and generally happy people you know, how many are really into politics?
Oh, 100% - politics is literally the mind and happiness killer. But isn't the rot all the way to the top? Like the "smoke filled rooms" are *already* choosing the bonkers candidates to primary, because they know that's what gets the vote out, and I worry that would still be true even without primaries.
I mean, who are the Republicans going to run in 2024? Trump is completely overdetermined, primary or no, right?
And who else would the Dems run? Harris was already literally getting up in front of microphones and basically explicitly saying "I have zero principles, I will say or do whatever you, the smoke filled rooms / voters want to hear, just please put me in office" and so on.
I think the core of the problem is less the literal system (which worked okay enough for ~200 years), and more the cultural polarization and memetics of it all that you point to.
Sure, let's try SOMETHING - we know big things need to change, and the Overton Window is really narrow on this, so maybe ranked choice is the best we can do, but I don't think it would move the needle much in the same cultural polarization and memetics environment, and good luck changing *that.*
Also, Silent Cal was never elected - he was Veep when the President died. But he got the Coolidge Effect named after him, so uh...that's something, posterity wise. 😂
Democratic AI-in-the-loop governance...sure, but don't give each voter the same number of tokens. Let the number be determined by scores on an AI-designed, multipartisan committee- approved test of simple knowledge about government and current events.
(Almost everyone hates almost every version of that idea. :)
Not me, I love it, but think it would be pretty hard to push through, and the straight purely democratic version is gonna be hard enough!
But supportive of your point, from Jason Brennan's Against Democracy:
The ANES surveys eligible voters on basic political information, such as who the candidates are or what these candidates stand for. There is tremendous variance in what eligible voters know. Political scientist Scott Althaus summarizes some of the results:
Just how high [the variance is] is made clear when we add up the number of correct answers to these questions and divide respondents into knowledge quartiles. While people in the highest knowledge quartile averaged 15.6 correct answers out of 18 possible, people in the lowest averaged only 2.5 correct answers.
On this test of political knowledge, the top 25 percent of voters are well informed, the next 25 percent are badly informed, the next 25 percent are know-nothings, and the bottom 25 percent are systematically misinformed.
I loved "Against Democracy". My only complaint is that Brennan was too accommodating about objections to voter pre-testing. I think each of the usual objections is legitimate, but that doesn't prove that testing would be a bad idea. We don't want to let the perfect be the enemy of the good...
> I personally HATE sleeping! It’s the most colossal waste of human potential and time, and I find it extremely and personally offensive that I’m subjected to it, and I’d pay significant sums to reduce it in myself or my kids.
I find something strangely endearing about the fact that they're trying to do a "full nerd" push here for some obscure neuropeptide to understand whether it can help them sleep a little less when judiciously used, when both cocaine and amphetamines exist, have been utilized by n=millions, and are well known for allowing essentially arbitrary sleep reduction amounts for years at a time.
Their hazard profiles and impacts on all cause mortality and many other endpoints are also already very well studied and attested, and their ease of acquisition and price are probably one OOM easier / cheaper, too.
But no, let's go for the obscure neuropeptide, guys, that's harder to get, more expensive, and that we're not sure will actually do anything. Hilarious.
Ahaha yeah. Modafinil exists, and seems to actually work at reducing sleep need without much side effects besides tolerance.
Everyone’s trying to crack the short sleeper syndrome hack though. An extra 2-3 hours a day, with a massive resistance to feeling tired is a major gain. I believe Obama had it.
It would be fun if it was all just a couple of genes that happened to up your orexin production. You’d get all the benefits of more time without the downside of amphetamines or stimulants.
> Ahaha yeah. Modafinil exists, and seems to actually work at reducing sleep need without much side effects besides tolerance.
Yeah, exactly, and so do coke and speed, it's just people don't generally use them cautiously and judiciously, with an eye towards overall health.
And even with people basically going all "fast food" on them, they have less negative impact than smoking!
And if you did use them responsibly and judiciously, you can achieve the types of ongoing sleep reduction they're aiming for with essentially no negative health effects.
I hear you on envying the people with the DEC2 mutation, though - one of my good friends has it, and MAN do I envy him that! He's just absurdly productive, too - was getting recruited for pro baseball when his shoulder blew out, pivoted to nerd stuff, got into finance (where we met), got his MBA, and now is leading teams AND heading a nonprofit AND doing a lot of work for his church AND having kids. He's a powerhouse! I love catching up with him, he's living 3-PMC-people's-lives-worth simultaneously.
EDIT - forgot to add that higher orexin levels is the hypthesized mechanism of action behind the DEC2 mutation giving ~2 hours of sleep per night reductions with no negative downsides.
One more thing. As always, your posts bristle with ideas, and I find them a joy to read. Anyone can be right, but few can actually think for themselves. And as interesting as a next-generation, AI-assisted democracy might be, I increasingly feel that the real challenge is figuring out the transition. Our politics have grown tired and sclerotic. Even the most beneficial AI would require society to adapt, either pro-actively (ha!) or at least in a timely fashion at multiple points. We are not up to the challenge. I’m dying to hear someone (and why not you?) sketch out an optimistic scenario of how AI leads us to the promised land without societal collapse or world war. AI-driven unemployment of even 20% in an important but AI-poor country could lead to ruin.
I’m certainly not expecting an off-the-cuff answer down here in the comments!
> I’m dying to hear someone (and why not you?) sketch out an optimistic scenario of how AI leads us to the promised land without societal collapse or world war.
You're completely right that this is the key bottleneck and question.
It's certainly not the in-depth treatment the question merits, but basically I think there's going to be a level of growth and alignment behind UBI that makes avoiding the "billionaires behind killbot-patrolled walled compounds" future relatively easy.
The broader geopolitical landscape is just totally ignored in the story, but I think that's where most of the known unknowns and unknown unknowns will live in practice, and is the interesting / hard part (beyond AI alignment) in terms of getting to good futures.
So I appreciate you pointing this out, I'll definitely give it some thought, and if I come up with anything I consider worth sharing, will write a post on it.
Two problems 1) the humans in the loop veto the good policies because humans HATE good policies 2) the AI, trained on milquetoast human policy pablum, produces mediocre ideas. I mean I can have a pretty good conversation with my AI about policy because I’ve forced it to be brutally honest, but that’s not going to be your starting point. I am reminded that long lasting, high functioning democracies tended to have a pretty narrow franchise. But how do we get there from here? Of course there is one other solution, to upgrade the whole electorate…
> 1) the humans in the loop veto the good policies because humans HATE good policies
Yeah, it's true, but my basic model was "most people don't even vote, the great majority are going to just not even going to look at them, much less veto," and so they'd generally carry though on inertia, despite the 10% handful of weirdos on either end who actually care and vote for or against.
> 2) the AI, trained on milquetoast human policy pablum, produces mediocre ideas.
Oh, I think this is very much a matter of context and prompting. As you say, you've reached a space where it can actually think of smart things, and I have too, and I'm sure the majority of my readers who use AI are there, too.
But also, think of this, pretty soon there's going RL loops directly skilling up AI capabilities to infer and be smarter than the context of past interactions and the prompt, too, because when personal AI assistants are out, average people will be asking it for advice and recipes on how to do well at various things at "average people" levels of context and prompting. And if the AI's give better advice, those people are likely to upgrade their price tier, so it's a directly aligned bundle of incentives on both ends - better life outcomes for people is better for the AI assistant companies' bottom line.
Those RL loops are going to be lifting an AI up from "regardless of the context and prompt here's personality space, what is the optimal personality / mind space to answer this question in an effective way."
> Of course there is one other solution, to upgrade the whole electorate…
Ha! Yup - definitely a fan of this one, as you saw in m SNP post. Lowering standards is always the worse choice!
But you know, the AI solution will move about 100x faster than any biological upgrade can percolate out society-wide.
> I imagine everyone else feels the way I feel about sleep
Not really. I wish days were 100 hours long so that I might attend to all those areas of life I feel are important. But I love how great days end on a high rather than simply peter out, and how bad days eventually end period. And every new morning holds a promise than even the best nap never knows.
Oh yeah, I know everyone loves sleep - the rest of the line was "about eating right and exercising" because I actually like those things, but by inspection, at least 90% of the population everywhere hates both of them with the fury of ten thousand suns.
So I'm basically reversed from the usual position - hates sleep, loves eating right + exercise, but most people love sleep and hate exercising and eating right.
Either way, it definitely sucks that "living a life well lived" relies on getting all 3 right, consistently.
This is going to sound hopelessly pedestrian relative to the galaxy-brained ideas outline here, but something we *could* do in the short term would be to just try to fix the primary system. Everything you're saying about democracy is entirely true, but a massive accelerant to all of the bad stuff is the reality that the average voter ends up with choices selected by the most insane 10% of the voter base.
In my experience, the people who are really into politics--i.e. the people who vote in primaries--are extremely screwed up people. They're on average smarter and better-informed than the mean voter, but because they're demented, they end up choosing outrageous options. Seriously, think about it: off all the smart, interesting, well-acclimatized and generally happy people you know, how many are really into politics?
And then there's the issue that, within any political institution, the way to prestige is to out-extreme the next-most-extreme person. And THEN you add in Sunstein polarization (where like-minded members debating something gradually move the entire group to a position more extreme than the most extreme person when it began), and you have . . . well modern American politics. Obviously, a fluid information environment supercharges this.
Politicians have figured out that the way to these extremist voters' hearts is negative polarization, so they're not even _lying_ about solving problems anymore! It's just "eat the rich" and "make the libs cry" all the way down; our leaders aren't even gesturing at solving problems. It was actually BETTER when politicians lied about things like "higher taxes on the rich mean everyone gets amazing, free everything forever" or "lower taxes means 25% annual economic growth forever!" At least the goal was something positive.
But, if we could deploy something like ranked choice voting at scale (or just go back to old smoke-filled rooms where party elites chose their candidates), it would be a huge step in the right direction. It still wouldn't be as good as idealized non-democratic options, but it could at least take us back to the mid-20th century when we had leaders like Calvin Coolidge get elected every so often.
> Seriously, think about it: off all the smart, interesting, well-acclimatized and generally happy people you know, how many are really into politics?
Oh, 100% - politics is literally the mind and happiness killer. But isn't the rot all the way to the top? Like the "smoke filled rooms" are *already* choosing the bonkers candidates to primary, because they know that's what gets the vote out, and I worry that would still be true even without primaries.
I mean, who are the Republicans going to run in 2024? Trump is completely overdetermined, primary or no, right?
And who else would the Dems run? Harris was already literally getting up in front of microphones and basically explicitly saying "I have zero principles, I will say or do whatever you, the smoke filled rooms / voters want to hear, just please put me in office" and so on.
I think the core of the problem is less the literal system (which worked okay enough for ~200 years), and more the cultural polarization and memetics of it all that you point to.
Sure, let's try SOMETHING - we know big things need to change, and the Overton Window is really narrow on this, so maybe ranked choice is the best we can do, but I don't think it would move the needle much in the same cultural polarization and memetics environment, and good luck changing *that.*
Also, Silent Cal was never elected - he was Veep when the President died. But he got the Coolidge Effect named after him, so uh...that's something, posterity wise. 😂
Democratic AI-in-the-loop governance...sure, but don't give each voter the same number of tokens. Let the number be determined by scores on an AI-designed, multipartisan committee- approved test of simple knowledge about government and current events.
(Almost everyone hates almost every version of that idea. :)
Not me, I love it, but think it would be pretty hard to push through, and the straight purely democratic version is gonna be hard enough!
But supportive of your point, from Jason Brennan's Against Democracy:
The ANES surveys eligible voters on basic political information, such as who the candidates are or what these candidates stand for. There is tremendous variance in what eligible voters know. Political scientist Scott Althaus summarizes some of the results:
Just how high [the variance is] is made clear when we add up the number of correct answers to these questions and divide respondents into knowledge quartiles. While people in the highest knowledge quartile averaged 15.6 correct answers out of 18 possible, people in the lowest averaged only 2.5 correct answers.
On this test of political knowledge, the top 25 percent of voters are well informed, the next 25 percent are badly informed, the next 25 percent are know-nothings, and the bottom 25 percent are systematically misinformed.
I loved "Against Democracy". My only complaint is that Brennan was too accommodating about objections to voter pre-testing. I think each of the usual objections is legitimate, but that doesn't prove that testing would be a bad idea. We don't want to let the perfect be the enemy of the good...
> I personally HATE sleeping! It’s the most colossal waste of human potential and time, and I find it extremely and personally offensive that I’m subjected to it, and I’d pay significant sums to reduce it in myself or my kids.
https://manifund.org/projects/orexin-pilot-experiment-for-reducing-sleep-need
^ I gave these guys a few dollars. It's being done by Sam Harsimony if you know and a few other guys who pop up occasionally on reddit. Here's a post for context: https://splittinginfinity.substack.com/p/sleep-need-reduction-therapies
I find something strangely endearing about the fact that they're trying to do a "full nerd" push here for some obscure neuropeptide to understand whether it can help them sleep a little less when judiciously used, when both cocaine and amphetamines exist, have been utilized by n=millions, and are well known for allowing essentially arbitrary sleep reduction amounts for years at a time.
Their hazard profiles and impacts on all cause mortality and many other endpoints are also already very well studied and attested, and their ease of acquisition and price are probably one OOM easier / cheaper, too.
But no, let's go for the obscure neuropeptide, guys, that's harder to get, more expensive, and that we're not sure will actually do anything. Hilarious.
Ahaha yeah. Modafinil exists, and seems to actually work at reducing sleep need without much side effects besides tolerance.
Everyone’s trying to crack the short sleeper syndrome hack though. An extra 2-3 hours a day, with a massive resistance to feeling tired is a major gain. I believe Obama had it.
It would be fun if it was all just a couple of genes that happened to up your orexin production. You’d get all the benefits of more time without the downside of amphetamines or stimulants.
> Ahaha yeah. Modafinil exists, and seems to actually work at reducing sleep need without much side effects besides tolerance.
Yeah, exactly, and so do coke and speed, it's just people don't generally use them cautiously and judiciously, with an eye towards overall health.
And even with people basically going all "fast food" on them, they have less negative impact than smoking!
And if you did use them responsibly and judiciously, you can achieve the types of ongoing sleep reduction they're aiming for with essentially no negative health effects.
I hear you on envying the people with the DEC2 mutation, though - one of my good friends has it, and MAN do I envy him that! He's just absurdly productive, too - was getting recruited for pro baseball when his shoulder blew out, pivoted to nerd stuff, got into finance (where we met), got his MBA, and now is leading teams AND heading a nonprofit AND doing a lot of work for his church AND having kids. He's a powerhouse! I love catching up with him, he's living 3-PMC-people's-lives-worth simultaneously.
EDIT - forgot to add that higher orexin levels is the hypthesized mechanism of action behind the DEC2 mutation giving ~2 hours of sleep per night reductions with no negative downsides.
One more thing. As always, your posts bristle with ideas, and I find them a joy to read. Anyone can be right, but few can actually think for themselves. And as interesting as a next-generation, AI-assisted democracy might be, I increasingly feel that the real challenge is figuring out the transition. Our politics have grown tired and sclerotic. Even the most beneficial AI would require society to adapt, either pro-actively (ha!) or at least in a timely fashion at multiple points. We are not up to the challenge. I’m dying to hear someone (and why not you?) sketch out an optimistic scenario of how AI leads us to the promised land without societal collapse or world war. AI-driven unemployment of even 20% in an important but AI-poor country could lead to ruin.
I’m certainly not expecting an off-the-cuff answer down here in the comments!
> I’m dying to hear someone (and why not you?) sketch out an optimistic scenario of how AI leads us to the promised land without societal collapse or world war.
You're completely right that this is the key bottleneck and question.
A friend challenged me to write the path to the "superstimuli" future I keep talking about, so I threw together a short story that actually detours through an AI+human golden age, I'll link it here for kicks: https://performativebafflement.substack.com/p/946c2601-6871-479f-a009-daf50f370377
It's certainly not the in-depth treatment the question merits, but basically I think there's going to be a level of growth and alignment behind UBI that makes avoiding the "billionaires behind killbot-patrolled walled compounds" future relatively easy.
The broader geopolitical landscape is just totally ignored in the story, but I think that's where most of the known unknowns and unknown unknowns will live in practice, and is the interesting / hard part (beyond AI alignment) in terms of getting to good futures.
So I appreciate you pointing this out, I'll definitely give it some thought, and if I come up with anything I consider worth sharing, will write a post on it.
Two problems 1) the humans in the loop veto the good policies because humans HATE good policies 2) the AI, trained on milquetoast human policy pablum, produces mediocre ideas. I mean I can have a pretty good conversation with my AI about policy because I’ve forced it to be brutally honest, but that’s not going to be your starting point. I am reminded that long lasting, high functioning democracies tended to have a pretty narrow franchise. But how do we get there from here? Of course there is one other solution, to upgrade the whole electorate…
> 1) the humans in the loop veto the good policies because humans HATE good policies
Yeah, it's true, but my basic model was "most people don't even vote, the great majority are going to just not even going to look at them, much less veto," and so they'd generally carry though on inertia, despite the 10% handful of weirdos on either end who actually care and vote for or against.
> 2) the AI, trained on milquetoast human policy pablum, produces mediocre ideas.
Oh, I think this is very much a matter of context and prompting. As you say, you've reached a space where it can actually think of smart things, and I have too, and I'm sure the majority of my readers who use AI are there, too.
But also, think of this, pretty soon there's going RL loops directly skilling up AI capabilities to infer and be smarter than the context of past interactions and the prompt, too, because when personal AI assistants are out, average people will be asking it for advice and recipes on how to do well at various things at "average people" levels of context and prompting. And if the AI's give better advice, those people are likely to upgrade their price tier, so it's a directly aligned bundle of incentives on both ends - better life outcomes for people is better for the AI assistant companies' bottom line.
Those RL loops are going to be lifting an AI up from "regardless of the context and prompt here's personality space, what is the optimal personality / mind space to answer this question in an effective way."
> Of course there is one other solution, to upgrade the whole electorate…
Ha! Yup - definitely a fan of this one, as you saw in m SNP post. Lowering standards is always the worse choice!
But you know, the AI solution will move about 100x faster than any biological upgrade can percolate out society-wide.
> I imagine everyone else feels the way I feel about sleep
Not really. I wish days were 100 hours long so that I might attend to all those areas of life I feel are important. But I love how great days end on a high rather than simply peter out, and how bad days eventually end period. And every new morning holds a promise than even the best nap never knows.
Oh yeah, I know everyone loves sleep - the rest of the line was "about eating right and exercising" because I actually like those things, but by inspection, at least 90% of the population everywhere hates both of them with the fury of ten thousand suns.
So I'm basically reversed from the usual position - hates sleep, loves eating right + exercise, but most people love sleep and hate exercising and eating right.
Either way, it definitely sucks that "living a life well lived" relies on getting all 3 right, consistently.