As I wrote the other day, p(doom) is somewhat-tongue-in-cheek shorthand for the probability that we are all going to die from some AI-induced extinction event. I rather doubt that will happen, but I am not sure that it won’t.
Indeed. And as long as you have politics working to limit the downsides of capitalism, it is OK. But everywhere that these limitations fail (and capitalists will attack them mercilessly, by definition) you get real problems. And that is generic and even affects the parts of siociety that should be most resilient: think how the fact that scientific publications are for-profit enterprises drives publication bias in many ways (who buys a boring magazine, you want to hear about the exciting new discoveries, not the failed search for a discovery or a successful corroboration of something). Even science is affected by 'entertainment' (the 'new' factor) that way.
I'm not entirely straight on why an alternative voting system would help get AI legislation passed. Is the argument that our current system benefits the extremes and thus congress just generally gets less done overall? I'm sympathetic to this.
It doesn't seem like congress is currently in love with AI and tech companies the way they've been in love with traditional media and innovators in the past. If anything, I'd guess that the more "extreme" ends of both parties are where you'll find the most skepticism of tech companies and AI. That's where we found the most opposition to section 230 of the CDA, right?
I'm kind of guessing that those of us who read Substacks about AI probably interact with more pro-AI and pro-tech people on a regular basis than do most Americans.
Not that a lack of popularity will necessarily stop an industry from successfully pulling off regulatory capture...
We have a two party system because nobody can rule from the center. There is always a benefit to taking a position slightly more to the Left or Right of an incumbent and the process continues.
Term limits are well studied. They lead to GREATER regulatory capture (because lobbyists go directly to the executive branch/civil servants) and LESS consensus (because if you can only run 1, 2, 3, times, why bother?).
We need to get all the soft money out of politics. Additionally, political flaws are a symptom of deeper economic problems. People who are secure about their economic well being aren't generally crackpots nor can they be exploited by crackpots.
My surprise is at, more or less, many well meaning people who otherwise refer to theories, science, etc., NOT referencing the ton of theoretical research done on RCV.
I remember being in an Econ department meeting in the 90s and the chair suggested everyone rank their choices for the new chair and everyone immediately yelled, "No"...because we were all economists and we'd read Arrow. In this particular case such an RCV-based vote was clearly biased in the direction of NOT electing a more diverse but equally qualified candidate. All the current examples are of "see this is better"...but it's confirmation bias. We haven't yet plumbed how wrong it can go.
Worse, I think that RCV can be easily gamed. This is why I think concentrating on soft money, corporation donations, etc., is a better bet. I am not swayed by term limits. There may need to be age limits. :(
Several new voting methods have been proposed. The short summary is that AV, though not the best-performing, remains attractive by virtue of being the simplest to implement. The best-performing new method is STAR. STAR asks voters to rate each candidate on a multi-point scale (usually 5-, 6-, or 7-point); of the two candidates with the highest total rating, the one rated above the other on more ballots wins.
STAR is so good, at least in simulation, that I think we're very close to being able to declare the voting method problem solved in an engineering sense; we just need to get more real-world experience with it to be sure. If you just ask people to rate candidates on a scale and take the highest-rated candidate, there will be a tendency for them to use only the highest and lowest values; they have no incentive to use the intermediate values, except when they're so ambivalent about a candidate that the cogntive effort required to choose thumbs-up or thumbs-down itself provides the incentive. (AV forces them to make that choice anyway, because its scale has only 2 points.) But STAR very cleverly provides a structural incentive to use the intermediate values, because preferences thus expressed can affect the outcome.
You are right that confirmation bias is rampant among IRV (RCV) supporters. They also tend to dislike AV specifically; I think, psychologically, this is because it's too different from IRV. STAR might be easier to persuade them to switch to, because rating candidates on a multi-point scale is not so different from ranking them; it mostly just means that you're allowed to give two candidates the same rating.
The problem with the "prioritize getting the money out of politics first" argument is that as long as an entrenched two-party system is in place and both parties have been captured by that money, who's going to reform the corrupted system? It's a logical double-bind. After witnessing the same recommendation going nowhere for decades, I've gotten to view it as impotent posturing. Especially since Citizens United. And fwiw, I view bemoaning Citizens United as impotent posturing, too. I want results.
There's more than one way to do ranked-choice voting. In my opinion, the crucial measure of benefit is found by simply allowing voters a first and second choice. It's the easiest to understand: rank #1 for the candidate you most prefer, and #2 for the one that you'll settle for.
I wouldn't anticipate that either of the two major parties would be supplanted by an outsider movement in the first few elections after RCV is in place. But both entrenched parties would be put on notice. And any outsider party with an articulate, charismatic candidate and a platform that reads more reasonably than a wish list to Santa at the North Pole might actually stand a chance of building its appeal over the course of several elections, as documented in tabulated vote rankings.
Speaking as someone who's been voting for Presidents since 1976, I've seen how easily the status quo has been gamed by the two parties over time- and also how much more of a thin margin the electorate is being handed to determine their choice as to who is Least Worst.
Picking the Least Worst of the Two Candidates we've been cornered into accepting as the Two Major Party Nominees: that's a pretty pathetic criterion to determine the leader of the United States of America. A subjective assessment- and, furthermore, one that requires no accountability from either major party to keep a single campaign promise. Because, where else you gonna take your vote? The Duopoly has all the leverage.
If even the limited RCV voting system I've recommended can somehow be gamed in a way that's even more pernicious than the status quo of entropic Duopoly, I'd like to read a detailed argument of its perils.
Would you happen to have any thoughts on the framework suggested by Prof Lessig in his book Republic Lost? The ideas expressed in the book, including Ranked choice voting, mainly democracy vouchers, and crowd-funded elections, seemed like an exciting solution to a messy funding problem. Would such a system help curb regulatory capture strategies?
Crowd-funding a candidate for an outsider party is practically guaranteed to be ineffective without ranked choice voting. With ranked-choice voting, that might be a different story.
The crucial thing about RCV is that it's ultimately a majoritarian system. Since Jimmy Carter won with 50.1% of the vote in 1976, the US has had only one new president elected with a majority: Barack Obama, in 2008. All of the other newly elected presidents got into office with less than 50% of the vote.
Furthermore, under the current status quo voting system, in the event that some especially charismatic/demagogic politician could manage to leverage their cult of personality as a 3rd candidate, it's entirely possible that they could gain the Presidency with only 33.3% of the vote- or even less. That isn't the kind of dynamism I want to see infused into the American electoral process. I'm not fond of the way that 3rd party votes have been practically guaranteed to consigned to play the role of "spoiler", either.
It only briefly acknowledges that there is controversy over whether AGI/singularity will ever actually happen, and especially in its timid section on regulation, doesn’t address potential biases that could come about from a company lobbying the government hard to regulate AI based on AGI fears rather than the problems pose today. Also, unsurprisingly doesn’t seem to demonstrate how openAI’s ever-growing LLMs will one day become AGI....
Anyway, not the worst thing I’ve ever read, but definitely a puff piece.
It also bizarrely mentions Altman’s desire to run for governor (?!?) but doesn’t explain his motivation for this at all, which kind of makes me skeptical of his current “I don’t want money and power, I just want to help!” stance
District-based systems are vulnerable (FPTP as in the UK even more so), far more than proportional representation (which is not invulnerable itself and also has disadvantages that must be taken into account). But if you have district-based, then, yes, ranked voting is a way to remedy a bit of that vulnerability. Far larger districts with say, 4-5 representatives (the 4-5 with the most votes) would also be an option.
The catch-22 is: you need to overcome the money-influence to get to a system where you can limit the vulnerability to money influence. How are you going to get this in the US as long as you have Talk Radio and Fox poisoning democracy? Not impossible, but very, very hard.
There is an extensive thorough article in Wikipedia on voting systems, their advantages and drawbacks; as someone said long ago, democracy is the second-best system of governance; limiting terms to a single one is a simple solution that could help remove individuals who use a long tenure to stack votes in their favor
These political ideas are interesting and worthy of discussion. But I doubt they are the solution to a possible doom from AI, or any other threat. As example, no matter what form of democracy we might have, nuclear weapons head in the sand denial would still be pervasive, and that's a far more immediate and tangible threat than what may someday possibly happen with AI.
My best guess is that we are substantially over estimating our ability to steer the future course of this civilization. We are probably riders on the bus of history, who are confused, and thinking that we are the driver.
If there is a solution to the existential threats that we keep inventing, it will most likely come in the form of pain, that which we actually learn from. As example, public consciousness on climate change vastly exceeds our focus on either AI or nuclear weapons, because we can see the results of climate change happening every other night on the news.
Which reminds me, I'll leave this here for now, as there is a major hurricane headed straight for our house, and due to arrive in the coming hours.
Strong agree! It is obvious these changes are what is best for people and the planet. I’ll add we need to bring open data science into our democratic processes to provide transparency into how things work.
People need to understand how things like gerrymandering and how states with lots of land but less people sway outcomes against the people’s wishes. The mathematics of voting gives Republicans an inherent unfair advantage. The benefits of ranked choice voting, taking money out of politics, eliminating gerrymandering, giving states with more people more representation, are all easy to prove as more democratic processes. Most people just don’t understand the statistics enough to see how rigged our election systems are against what people actually want.
Gary, you might be interested to hear that the US' two-party system has been called a "Doom Loop" by political scientists! See https://www.uniteamerica.org/page/two-party-doom-loop for background -- Lee Drutman has a full-length book treatment of the issue https://www.amazon.com/Breaking-Two-Party-Doom-Loop-Multiparty/dp/0190913851 .
I would prefer a combination of term limits AND ranked choice voting.
“The goal of the companies is return to shareholders, not to return to humanity.”
Aye there’s the rub
Indeed. And as long as you have politics working to limit the downsides of capitalism, it is OK. But everywhere that these limitations fail (and capitalists will attack them mercilessly, by definition) you get real problems. And that is generic and even affects the parts of siociety that should be most resilient: think how the fact that scientific publications are for-profit enterprises drives publication bias in many ways (who buys a boring magazine, you want to hear about the exciting new discoveries, not the failed search for a discovery or a successful corroboration of something). Even science is affected by 'entertainment' (the 'new' factor) that way.
I'm not entirely straight on why an alternative voting system would help get AI legislation passed. Is the argument that our current system benefits the extremes and thus congress just generally gets less done overall? I'm sympathetic to this.
It doesn't seem like congress is currently in love with AI and tech companies the way they've been in love with traditional media and innovators in the past. If anything, I'd guess that the more "extreme" ends of both parties are where you'll find the most skepticism of tech companies and AI. That's where we found the most opposition to section 230 of the CDA, right?
I'm kind of guessing that those of us who read Substacks about AI probably interact with more pro-AI and pro-tech people on a regular basis than do most Americans.
Not that a lack of popularity will necessarily stop an industry from successfully pulling off regulatory capture...
RCV is a well studied problem in economics. Look at Arrow's Impossibility Theorem. Condorcet's Paradox. RCV doesn't actually work right.
https://lawliberty.org/the-flaws-of-ranked-choice/
We have a two party system because nobody can rule from the center. There is always a benefit to taking a position slightly more to the Left or Right of an incumbent and the process continues.
Term limits are well studied. They lead to GREATER regulatory capture (because lobbyists go directly to the executive branch/civil servants) and LESS consensus (because if you can only run 1, 2, 3, times, why bother?).
We need to get all the soft money out of politics. Additionally, political flaws are a symptom of deeper economic problems. People who are secure about their economic well being aren't generally crackpots nor can they be exploited by crackpots.
While I agree that RCV is flawed, I don't understand how someone can write such an article and not even mention Approval Voting.
Here's my case for AV: https://scottsthotts.substack.com/p/one-person-one-vote-is-wrong
sir, i referenced other systems and said i was open to them but wasn’t going into detail or adjudicated.
Hi Gary, I was referring to the link in the parent comment by Stephyn Butcher, not to your post. Sorry, I should have been clearer.
I haven't heard of AV before. That's interesting.
My surprise is at, more or less, many well meaning people who otherwise refer to theories, science, etc., NOT referencing the ton of theoretical research done on RCV.
I remember being in an Econ department meeting in the 90s and the chair suggested everyone rank their choices for the new chair and everyone immediately yelled, "No"...because we were all economists and we'd read Arrow. In this particular case such an RCV-based vote was clearly biased in the direction of NOT electing a more diverse but equally qualified candidate. All the current examples are of "see this is better"...but it's confirmation bias. We haven't yet plumbed how wrong it can go.
Worse, I think that RCV can be easily gamed. This is why I think concentrating on soft money, corporation donations, etc., is a better bet. I am not swayed by term limits. There may need to be age limits. :(
If you haven't heard about AV, you're way out of date. A lot has happened since Arrow! First, we got Yee diagrams: http://zesty.ca/voting/sim/ Then we got a couple of Monte Carlo-style approaches to evaluating the performance of voting methods in practice: the Bayesian Regret framework: https://rangevoting.org/BayRegDum.html , and Voter Satisfaction Efficiency: https://electionscience.github.io/vse-sim/
Several new voting methods have been proposed. The short summary is that AV, though not the best-performing, remains attractive by virtue of being the simplest to implement. The best-performing new method is STAR. STAR asks voters to rate each candidate on a multi-point scale (usually 5-, 6-, or 7-point); of the two candidates with the highest total rating, the one rated above the other on more ballots wins.
STAR is so good, at least in simulation, that I think we're very close to being able to declare the voting method problem solved in an engineering sense; we just need to get more real-world experience with it to be sure. If you just ask people to rate candidates on a scale and take the highest-rated candidate, there will be a tendency for them to use only the highest and lowest values; they have no incentive to use the intermediate values, except when they're so ambivalent about a candidate that the cogntive effort required to choose thumbs-up or thumbs-down itself provides the incentive. (AV forces them to make that choice anyway, because its scale has only 2 points.) But STAR very cleverly provides a structural incentive to use the intermediate values, because preferences thus expressed can affect the outcome.
You are right that confirmation bias is rampant among IRV (RCV) supporters. They also tend to dislike AV specifically; I think, psychologically, this is because it's too different from IRV. STAR might be easier to persuade them to switch to, because rating candidates on a multi-point scale is not so different from ranking them; it mostly just means that you're allowed to give two candidates the same rating.
For more detail, Jameson Quinn has an excellent summary: https://www.lesswrong.com/posts/D6trAzh6DApKPhbv4/a-voting-theory-primer-for-rationalists
The problem with the "prioritize getting the money out of politics first" argument is that as long as an entrenched two-party system is in place and both parties have been captured by that money, who's going to reform the corrupted system? It's a logical double-bind. After witnessing the same recommendation going nowhere for decades, I've gotten to view it as impotent posturing. Especially since Citizens United. And fwiw, I view bemoaning Citizens United as impotent posturing, too. I want results.
There's more than one way to do ranked-choice voting. In my opinion, the crucial measure of benefit is found by simply allowing voters a first and second choice. It's the easiest to understand: rank #1 for the candidate you most prefer, and #2 for the one that you'll settle for.
I wouldn't anticipate that either of the two major parties would be supplanted by an outsider movement in the first few elections after RCV is in place. But both entrenched parties would be put on notice. And any outsider party with an articulate, charismatic candidate and a platform that reads more reasonably than a wish list to Santa at the North Pole might actually stand a chance of building its appeal over the course of several elections, as documented in tabulated vote rankings.
Speaking as someone who's been voting for Presidents since 1976, I've seen how easily the status quo has been gamed by the two parties over time- and also how much more of a thin margin the electorate is being handed to determine their choice as to who is Least Worst.
Picking the Least Worst of the Two Candidates we've been cornered into accepting as the Two Major Party Nominees: that's a pretty pathetic criterion to determine the leader of the United States of America. A subjective assessment- and, furthermore, one that requires no accountability from either major party to keep a single campaign promise. Because, where else you gonna take your vote? The Duopoly has all the leverage.
If even the limited RCV voting system I've recommended can somehow be gamed in a way that's even more pernicious than the status quo of entropic Duopoly, I'd like to read a detailed argument of its perils.
Would you happen to have any thoughts on the framework suggested by Prof Lessig in his book Republic Lost? The ideas expressed in the book, including Ranked choice voting, mainly democracy vouchers, and crowd-funded elections, seemed like an exciting solution to a messy funding problem. Would such a system help curb regulatory capture strategies?
Crowd-funding a candidate for an outsider party is practically guaranteed to be ineffective without ranked choice voting. With ranked-choice voting, that might be a different story.
The crucial thing about RCV is that it's ultimately a majoritarian system. Since Jimmy Carter won with 50.1% of the vote in 1976, the US has had only one new president elected with a majority: Barack Obama, in 2008. All of the other newly elected presidents got into office with less than 50% of the vote.
Furthermore, under the current status quo voting system, in the event that some especially charismatic/demagogic politician could manage to leverage their cult of personality as a 3rd candidate, it's entirely possible that they could gain the Presidency with only 33.3% of the vote- or even less. That isn't the kind of dynamism I want to see infused into the American electoral process. I'm not fond of the way that 3rd party votes have been practically guaranteed to consigned to play the role of "spoiler", either.
Are you going to write anything in response to WIRED’s puff piece for OpenAI?
https://www.wired.com/story/what-openai-really-wants/
It only briefly acknowledges that there is controversy over whether AGI/singularity will ever actually happen, and especially in its timid section on regulation, doesn’t address potential biases that could come about from a company lobbying the government hard to regulate AI based on AGI fears rather than the problems pose today. Also, unsurprisingly doesn’t seem to demonstrate how openAI’s ever-growing LLMs will one day become AGI....
Anyway, not the worst thing I’ve ever read, but definitely a puff piece.
It also bizarrely mentions Altman’s desire to run for governor (?!?) but doesn’t explain his motivation for this at all, which kind of makes me skeptical of his current “I don’t want money and power, I just want to help!” stance
District-based systems are vulnerable (FPTP as in the UK even more so), far more than proportional representation (which is not invulnerable itself and also has disadvantages that must be taken into account). But if you have district-based, then, yes, ranked voting is a way to remedy a bit of that vulnerability. Far larger districts with say, 4-5 representatives (the 4-5 with the most votes) would also be an option.
The catch-22 is: you need to overcome the money-influence to get to a system where you can limit the vulnerability to money influence. How are you going to get this in the US as long as you have Talk Radio and Fox poisoning democracy? Not impossible, but very, very hard.
There is an extensive thorough article in Wikipedia on voting systems, their advantages and drawbacks; as someone said long ago, democracy is the second-best system of governance; limiting terms to a single one is a simple solution that could help remove individuals who use a long tenure to stack votes in their favor
I have some Substack posts on voting systems that may be of interest. Here's a starting point:
https://open.substack.com/pub/scottsthotts/p/an-introduction-to-the-voting-system
These political ideas are interesting and worthy of discussion. But I doubt they are the solution to a possible doom from AI, or any other threat. As example, no matter what form of democracy we might have, nuclear weapons head in the sand denial would still be pervasive, and that's a far more immediate and tangible threat than what may someday possibly happen with AI.
My best guess is that we are substantially over estimating our ability to steer the future course of this civilization. We are probably riders on the bus of history, who are confused, and thinking that we are the driver.
If there is a solution to the existential threats that we keep inventing, it will most likely come in the form of pain, that which we actually learn from. As example, public consciousness on climate change vastly exceeds our focus on either AI or nuclear weapons, because we can see the results of climate change happening every other night on the news.
Which reminds me, I'll leave this here for now, as there is a major hurricane headed straight for our house, and due to arrive in the coming hours.
They might be necessary:; certainly they are not sufficient
Sounds good, I can vote for that.
A possible topic for a future article? What would be sufficient to credibly address whatever threat AI may present?
My best guess, some kind of dramatic, easily understood, widely reported on AI calamity. An AI weapons system gone wrong, or something like that.
This scenario is distinctly possible :
Yes
Strong agree! It is obvious these changes are what is best for people and the planet. I’ll add we need to bring open data science into our democratic processes to provide transparency into how things work.
People need to understand how things like gerrymandering and how states with lots of land but less people sway outcomes against the people’s wishes. The mathematics of voting gives Republicans an inherent unfair advantage. The benefits of ranked choice voting, taking money out of politics, eliminating gerrymandering, giving states with more people more representation, are all easy to prove as more democratic processes. Most people just don’t understand the statistics enough to see how rigged our election systems are against what people actually want.
https://adamgolding.substack.com/publish/post/113579204
https://www.lesswrong.com/posts/z4xNBnusXzysm9mob/votes-per-dollar