Replacing politicians with AI?

[Icey!]

I think most of us can agree that Politicians, can't be trusted. They are easily corrupted by the power they gain and mainly think about themselves first when passing a bill. The lobbying in our democracy also doesn't help. There have been different solutions of handling this (like direct democracy) but there exists one idea that I thought about which is the notion of AI replacing Politicians.

I can imagine being implemented in multiple ways:

1.) AI Political candidates.

The least groundbreaking idea is to just vote different AIs to be in power which doesn't solve much because you are depending on that the different AIs aren't backed by lobbyists.


2.) AI + Direct Democracy.

This idea is actually from a Ted Talk seen here:

For the people who don't know, Direct Democracy basically means that we skip the politicians and vote the bills ourselves and then use AI to automate things. I have absolutely no idea how this will pan out as this assumes that the AI automating the voting process isn't in favor of the creators of said AI.


3.) AI Dictatorship.

This idea came from the notion of authoritarian socialist countries and what originally made me come up with the idea. Much of the problem of socialism comes from how the people in charge have a lot more control than in capitalist countries which makes them more susceptible to corruption. Take Cuba for example where much of the population is very poor while the Politicians live like kings. An AI Dictatorship that has no self interests can get around the "hierarchy" problem and might be more open minded to it's own people. But again this assumes that the people who built the AI are not in the interests of the creators.


I don't know what else to say I just found the idea interesting.

[DiffydaDude]

yeah nah that sounds pretty fucky
thats all i have to say

[keybladekid6]

My initial reaction to the title of your topic was "is this guy crazy?" but after reading your post and thinking about it I actually see where you're coming from.

The high that politicians/government leaders get off of their power is definitely a huge factor into why so many decisions only serve to benefit them, their interests, and their wealth, but I think a lot of that is also caused by their lack of empathy toward the greater population that they serve as a whole- while AI would be able to automate things, I'm assuming that we're not at the point technologically that AI can effectively empathize with a human being.

Similarly, I don't think we're at a point yet where an AI can be built and trained on whatever it's built and trained on without somehow reflecting the interests of either the creators or what it was trained on- it'd be hard to create an impartial judge of human affairs, and that also seems like a pretty risky way to dictate people's lives, right? (lack of empathy coming into play again)

Anywhoozies. You're right though!! This is really interesting to think about. I think I read a webcomic once where AI served as the leader of the community the main character was living in, and it wasn't a dystopia or anything. I remember that being really interesting, and having a think about this was too c:

xoxo, lolo =3

[vvinrg]

1.)

I think this would just lead to high-speed political debates, with no real difference to today. You can have a cult of personality around just about anything, assuming it has a brand of some sort.

2.)

Watching the TED Talk, I find it funny that his examples of modern automation are, on the one hand, a factory, which no-doubt requires plenty of staff on-hand to perform software and physical maintenance, and self-driving cars, a product that has yet to come to fruition in any real form.

The entire notion of a digital political agent seems utterly magical and absurd.

Navigation software ultimately leverages a 7-decade old algorithm with various additions, all maintained and updated actively by engineers at Google or Apple or whatever else, not to mention the immense costs in creating the datasets they work off of.

Most recommendation algorithms at some point involve the idea of scrobbling, that is, cross-referencing the actions of one user with several others to find trends in what similar people like. Many platforms (YouTube, Twitter, etc.) simply use the metric of increasing engagement, in the hopes it will drive up profits. Even with as mysterious as the YouTube "black box" is, I routinely see content I wouldn't touch with a 10-foot pole in my feed. For that matter, social media platforms will generally end up just serving content that upsets the user, as that provokes engagement effectively.

There is a major difference between an algorithm that can go "hm, similar people also like the Pixies, you should listen to them", or, "hm, they've clicked on content that says 'death to such and such' before, let's show them this", and an algorithm that can understand political nuance. Neural networks, as it stands, are not reliable enough, and the amount of disk, compute, and time, that it would take, for every person to be doing RLHF, would be ridiculous.

The entire idea of an "Open Market of Political Algorithms" is absurd on its own(how would that even work?), but we've also reintroduced the role of representative. If "algorithms get more popular", that's just a new form of personality cult. Someone, somewhere, made that algorithm. Beyond that, what does this auditing process look like? If I leave my algorithm on autopilot, and see on the news that a bill I wholeheartedly disagree with was passed, and I find that the algorithm has voted for it instead, can I retract its vote?

"Beware of Pessimists," believe me, if I never had to look at Mitch McConnell's cold dead face again, I would rejoice. The speaker seems to forget that he is speaking about the problems of the real world, that can really affect people. This is a very interesting hypothetical, but severely flawed and incomplete, even at the big picture.

"An algorithm that tries to get a certain percentage of approval." These are, presumably, machine learning systems. They are notoriously good at breaking the systems they are in, in pursuit of higher agility. If we assume the agents are, at least some of them, based on LLMs, how long before the approval maximizer learns to exploit prompt injection?

The implementation strategy is, at least, reasonable, but I think the final product would be so far ship-of-theseused from the original premise, to be at all reliable, that it would no longer be this man's idea.

I think this proposal is fundamentally unserious, and seems to not give thought to the fact that it will affect real human lives, if implemented.

3.)

I can't parse your last sentence, but I think this is the most-right idea. Still, an "AI making decisions" is very handwave-y. I think few would be opposed to a perfectly benevolent ruler, it's more a question of how we get there.

My Proposal:

I think the most fully-automatable portion of the government, and I am speaking with regards to the United States, as that's where I reside, would be the judicial branch.

Reasoning engines are not particularly clever, at least not public ones, but they could be. If the rigor of proofing systems were to be applied to digital reasoning, we could, perhaps, logically prove the verdict court cases. The major implementation step here would be to convert all laws into a format understood by this hypothetical legal engine. Legalese is formalized to a degree that partial automation of this step is not out-of-the-question.

At that point, however, there is the far greater concern of validating evidence. For certain forms of crime-scene evidence, such as DNA, this is trivial. Similarly, it would be easy enough to convey the contents of financial records in a fraud case, but what if video is to be taken as evidence? Photos? Classifier neural networks are, as stated above, very susceptible to trickery. We could, perhaps, employ a human set of classifiers, but now every citizen needs to be trained in digital legalese. That's not impossible, but if your average Usonian can't recognize states properly, why would we expect them to be adequately trained in a language for specifying formal logic?

Ultimately, I think a system like this would reduce strain on the legislative branch. If the legal code is machine-readable, ambiguous or inconsistent laws can be detected and fixed before proposal. If fewer edge cases need to be handled after-the-fact, fewer laws will need to be passed. If there are fewer laws, the many will need less time to vote on them, and this solves a lot of potential shortcomings of direct democracy.

That being said, the reason we have a representative democracy in the US is not to reduce strain on voters, it is because the founding fathers believed we the people could not be trusted with such responsibility.

This also leaves open the question of who exactly writes laws. In an ideal world, everyone, but that's likely impossible.

I don't think this is a topic that is anywhere near having bearing on reality, aside from the fact of already-heavy use of automation in the various government-adjacent entities(banks, credit bureaus, hiring management, etc.). The non-government "Man" has rashly opted for automation in all manner of things already, as profit maximization can be focused on easily.

[ThunderPerfectWitchcraft]

I think most of us can agree that Politicians, can't be trusted. They are easily corrupted by the power they gain and mainly think about themselves first when passing a bill. The lobbying in our democracy also doesn't help. There have been different solutions of handling this (like direct democracy) but there exists one idea that I thought about which is the notion of AI replacing Politicians.

I agree that many politicans can't be trusted, but a part of the problem is that the representative democracy is subject to the overlaying capitalist system. It shouldn't wonder that Biden or - even more so - Trump are Marionettes of the capital; this was clear before they were elected. The problem is that the media, parts of the education system, the means of production, and the infrastructure is in the hand of the hegemonial powers that are interested in the preservation and intensification of the status quo (Thanks to this, the next German chancellor is likely to be a guy who works for Blackrock - but the people who will be most affected by his politics are to tired and uneducated to care, or even manipulated to support him or his alikes).  Candidates that really want to improve things have barely a platform to reach the people - I'm myself a member of a troubled, small, and ideological stringent party and can sing you a song about it (and, if this changes, are often banned or attacked in other ways).

Direct democracy is no help; the causing problems will still be there - additionally it is often unsharp, and also easily to manipulate; see the Swiss (once called a "feudal system" by a researcher) were a terrible social inequality is at hand, and the rights of minorities are often enough trampled. With direct democracy, my country here would probably have the death penalty and probably torture of suspects - no thank you. This is iirc what Aristotle had in mind when he called democracy a degenerated form of state system.

It is yet unclear if a true AI is possible. What is traded as AI by now is a glorified chance calculator. Lets pretend it is for this post.

1.) AI Political candidates.

The least groundbreaking idea is to just vote different AIs to be in power which doesn't solve much because you are depending on that the different AIs aren't backed by lobbyists.

Problem is even more intense: You are totally subject to the developer of this AI. Whoever makes it can indirectly control your overlord, and you can be sure that all kinds of lobbyists will be standing at his door; if he want he can create a AI that is a turbo version of Donald Trump (or Adolf Hitler) - the AI as middleman will make this even harder to control, as it will have no visible relations that could warn us.

2.) AI + Direct Democracy.

This idea is actually from a Ted Talk seen here:

For the people who don't know, Direct Democracy basically means that we skip the politicians and vote the bills ourselves and then use AI to automate things. I have absolutely no idea how this will pan out as this assumes that the AI automating the voting process isn't in favor of the creators of said AI.

Will be a combination of the problems of direct democracy (to much subject to the needs of majority groups, to much driven by contemporary "feelings") and the problems of AI mentioned above.

3.) AI Dictatorship.

This idea came from the notion of authoritarian socialist countries and what originally made me come up with the idea. Much of the problem of socialism comes from how the people in charge have a lot more control than in capitalist countries which makes them more susceptible to corruption. Take Cuba for example where much of the population is very poor while the Politicians live like kings. An AI Dictatorship that has no self interests can get around the "hierarchy" problem and might be more open minded to it's own people. But again this assumes that the people who built the AI are not in the interests of the creators.


I don't know what else to say I just found the idea interesting.

This would mean to take the worst part of real-socialism and combine it with the problems of AI. But this deserves a bit more love:

1. You take Cuba as an example. The claim that the politicians live like kings while the population is poor is backed only by a person now fled to the USA. Possible that it is true. Surely true that there is a problem with corruption. However, one should have in mind that the live-expectation in Cuba is - even though they are sanctioned and economically isolated, the live expectation is at this time over the live expectation of the USA.

2. Many real-existing socialist countries didn't start out as authoritarian. The streams within the socialist countries that were ideological true to socialism and often more liberal were cut out by movements that were authoritarian, most often since those promised to be more effective in the conflict with the capitalist sates and to "solve" imminent economical problems (an example is the SU, where Stalin could outdo Lenin and Trotzky (both not without wrongs) after it became visible that the revolution probably wouldn't go on in other states).

3. As sorry as I'm to say this: A good AI-overlord wouldn't help against an US-coup. We had great, ideologically true socialist politicians in the past that fell victim to violence from the imperial powers. See Salvador Allende or Patrice Lumumba. Getting resilient against this violence will always conflict with the needs of the humans we want to serve with our system in the one form or the other.

The problem is capitalism and humanities problem to overcome this - basically primitive - system, and the fact that it is much easier to destroy than to improve things - not the wrongs of individuals in power. Hope I could deliver my point, the language barrier strikes a bit for me when things get complex ;).

[BlazingCobaltX]

AI and algorithms are nowhere near the point where we can leave them to run things unsupervised. In the Netherlands, the lives of tens of thousands of families got destroyed because a trained algorithm illegitimately flagged the parents for child benefits fraud. The information we feed to these systems simply isn't unbiased enough to have the AI sieve an universal truth from it. 

Another thing is the assumption that the AI will innately draw conclusions that are beneficial to humankind. If you task an "AI dictator" with protecting the environment, I doubt humans will be spared its wrath. Tangentially related to this is the literary work I Have No Mouth, And I Must Scream, where a supercomputer made for war takes over. The computer was created for one purpose and now even as it is in full control over itself, it is incapable of inflicting anything but pain on its subjects. I was a bit too unsettled by the story to fully read it, but I think it's interesting for this topic.

[Melooon]

Navigation software ultimately leverages a 7-decade old algorithm with various additions

This is a really good point; there are two things that people mean when they say AI; there's what we have today which is basically a carefully managed algorithmic machine that's constantly being tweaked and maintained by a software team; and then there's the sci-fi idea of an AI that exists in some sort of black box and manages itself entirely without contact with human hands. Those are totally different things!

AI served as the leader of the community the main character was living in, and it wasn't a dystopia or anything. I remember that being really interesting, and having a think about this was too c:

It is a fun idea, and it would be super easy to trial (maybe in a small isolated island nation; but not the one I live in ) All that said; when I ask myself if an AI could help moderate this forum.. honestly the answer is no (It could prob greet people but that's about all); the social balance is too complex; if an AI couldn't manage this place, how could it manage a whole country??

problem is that the media, parts of the education system, the means of production, and the infrastructure is in the hand of the hegemonial powers

Direct democracy is no help; the causing problems will still be there

If the internet has proven anything it's that mob rule is no rule Maybe what Iv observed is that it really doesn't matter who is in power, it can be a monarch, an elected party, or an economic class, they can all be both good and bad; but usually they become bad when they don't feel accountable to the entire population they govern.

@Icelogist Have you watched Logans Run? Its a fun movie and it's a pretty good depiction of a society that has been handed over to an AI (if I remember correctly!)

Another fun angle to consider is Asimov's Zeroth law of robotics (which is basically the trolly paradox) - A governing AI will almost certainly be faced with situations where the information it promotes will benefit a majority but destroy certain individuals - it would have no emotional feeling towards this; so what logical conclusion would it pick!

[Icey!]

@Icelogist Have you watched Logans Run? Its a fun movie and it's a pretty good depiction of a society that has been handed over to an AI (if I remember correctly!)

Never heard of it! I saw the trailer and it looks interesting... Might give it a watch.

[devils]

It's a bad idea. Considering AI racism is a thing, chances are things wouldn't end well. Machines aren't neutral and often favor the ones that command them.

Much of the problem of socialism comes from how the people in charge have a lot more control than in capitalist countries which makes them more susceptible to corruption. Take Cuba for example where much of the population is very poor while the Politicians live like kings.

Cuba lives under a communist dictatorship. This isn't to say socialist-run countries are perfect (as someone from one, I can assure you they're not), but socialism is not communism, despite some similarities.

[ThunderPerfectWitchcraft]

If the internet has proven anything it's that mob rule is no rule Maybe what Iv observed is that it really doesn't matter who is in power, it can be a monarch, an elected party, or an economic class, they can all be both good and bad; but usually they become bad when they don't feel accountable to the entire population they govern.

I believe that a elected ruler behaves better than a monarch any day. As far as I'm aware we share the privilege of living in states that are at least somewhat constitutional.

Cuba lives under a communist dictatorship. This isn't to say socialist-run countries are perfect (as someone from one, I can assure you they're not), but socialism is not communism, despite some similarities.

Cuba is (still) socialist, and last I checked Portugal(?) is not oo

[devils]

Cuba is (still) socialist, and last I checked Portugal(?) is not oo

Really? Weird, I could've sworn it was considered a dictatorship... Maybe I should've searched it up before typing that, whoops.

Portugal is weird right now politically- we literally had our prime-minister resign last week- but while we do economically live in this weird capitalist-socialist limbo, our government is very much a social democracy (which is a branch of socialism).

[ThunderPerfectWitchcraft]

Really? Weird, I could've sworn it was considered a dictatorship... Maybe I should've searched it up before typing that, whoops.

Portugal is weird right now politically- we literally had our prime-minister resign last week- but while we do economically live in this weird capitalist-socialist limbo, our government is very much a social democracy (which is a branch of socialism).

Hi.
Short (and thus only so-and-so-far correct) explanation:
In a Marxist sense, socialism is a term used to describe a society that is in the stage of transformation to communism (which is, by no means, a dictatorship - the resulting communist society is a liberal, classless society that can be either democratic or anarchist in its way of organization).
Most of the existing socialist states, including Cuba, considered themself to be socialist (but never managed to transform into truely communist states), and shifted between authoritarian and democratic structures of organization (much like the western states). -> A socialist country is not forcefully democratic in its organization, it is a term that mostly describes the management of the means of production.

Social Democracy resulted from a schism within the early 20th century workers movements, but basically waved the idea of a communist transformation within a realistic timeframe. Although they still borrow ideas from socialist organization, they aren't usually socialist in a Marxist sense, and tend to collaborate with capitalist players citing the need to orient on the necessities, often even taking right-wing stances over this that are directed against their original base (also called the neoliberal transformation, a progress that started in the 80s and took much acceleration in the 90s after the iron curtain fell). Your "Socialist" Party in Portugal is seemingly no exception on this: https://en.wikipedia.org/wiki/Socialist_Party_(Portugal)#Ideology (see the article about the "third way" for further explanations - it is actually written from a perspective that is opposed to mine, so you'll have a round and nice picture!).

[Icey!]

Seeing the amount of disapproval for this idea...I agree, AI is a very odd piece of tech and when training it, it can get the wrong idea on almost everything.

There exists a limitation with machines, and that is it needs exact instructions. When you ask an AI to "solve the unemployment crisis" it may make policies to make unemployment illegal and make government jobs to play Minecraft. Technically this may solve the unemployment crisis but it's not what we wanted.

With our current tech the best that we can do is probably a planned economy, a whole dictatorship is another story. Maybe one day a AI government will be possible but we just aren't ready for it.