Ich habe viel darüber nachgedacht, besonders nachdem ich gesehen habe, wie maschinelles Lernen bereits in Dinge wie Wohlfahrtsberechtigung, Vorhersagetopolierung und Klimassimulationen eingebettet ist. Wir reden nicht einmal über Agi, nur neurale Netzwerke der aktuellen Generation, an die es gewöhnt ist "beraten" politische Entscheidungen.
Stellen Sie sich vor, AI schlägt nationale Budgets vor, die neue Gesetze zu simulieren, bevor sie verabschiedet werden, und treffen Sie Echtzeitentscheidungen während einer Pandemie, die automatisch Steuern, polizeiliche Prioritäten oder soziale Dienste auf der Grundlage von Daten anpassen?
Auf den ersten Blick klingt es ideal: Keine Korruption, kein Ego, kein Schlaf erforderlich. Aber dann haben Sie die harten Fragen getroffen. Wer definiert die Ziele der KI? Was passiert, wenn es aber kalt ist, aber "effizient" Entscheidung, die Menschen schadet? Kann die Demokratie überleben, wenn wir nicht verstehen (oder protestieren können), wie Entscheidungen getroffen werden?
Ich habe dies in einem Video -Aufsatz ausgepackt, weil die Auswirkungen massiv (und irgendwie furchterregend) sind. Aber in diesem Beitrag geht es nicht darum. Was ich Wirklich will hören::
👉 Würden Sie einer KI -Regierung vertrauen?
👉 Wo ziehen Sie die Grenze dazwischen? "Beratungsinstrument" Und "Regierungskraft"?
👉 Gibt es eine Version dieser Zukunft, in der Sie sich sicher fühlen würden?
Für jeden, der neugierig ist, hier ist der vollständige Video -Aufsatz, den ich erwähnt habe: https://www.youtube.com/watch?v=EFM8EPGHIPW
Aber die große Frage ist Wäre die Zukunft tatsächlich besser, wenn sie nach Code ausgeführt werden?
What if your next Prime Minister wasn’t human, but a neural network? Could we ever trust an AI government?
byu/hexferro inFuturology
24 Kommentare
For anyone curious, here’s the full video essay I mentioned: [https://www.youtube.com/watch?v=Efm8EPghipw](https://www.youtube.com/watch?v=Efm8EPghipw)
I try to explore both the benefits and dangers, with examples from China, Estonia, the UK, and Denmark’s “Leader Lars.”
> Would the future actually be better if run by code?
I just want to point out that there is a difference between „AI“ and „code“. At the moment AIs are mostly LLMs, but AIs could be part hardware/software, part biological substracts
So I guess the question you wanted to ask is „Would the future actually be better if run by non human intelligences“
I imagine that as the amount of compute increases, so does the efficacy of central economy planning. We’ll call it distributed cloud economic planning so nobody freaks out.
With the current state of government (US), processor take the wheel.
I’m at a point where I would prefer a global, algorithmically-managed prefecture system in place of nation states competing over finite resources. Nation states sending people to die over oil or minerals is old people playing old games. It’s about time to start colonizing other planets and mining asteroids.
For sure. The faster we hand the reigns over to an Uber AI the better it is. See: [https://www.uberai.org/](https://www.uberai.org/)
IF the AI was properly aligned and significantly more capable than what we have today, the results would be much better than any possible human form of government. Just the fact that you could get rid of individual biases, agendas, corruption and escape the election cycle would be huge.
Imagine a significant part of your government’s agenda wouldn’t be „how do we get elected again in four years?“ all of a sudden. Imagine your government being able to plan mid and long term, instead of just patching the holes? Wouldn’t that be great?
The devil is in the details, though. As of now, we couldn’t even agree what „properly aligned government AI“ would mean. Do you maximize flourishing, or minimize suffering? Do you go for equal opportunity, or equality of outcome?
We should not give control of our government to AIs.
But we should definitely try to model our collective aspirations using AI technology to determine what is important. Voting once every 2 or 4 years is far from being an accurate representation of what we really want, and this might be a way to get better information in the hands of our politicians, while giving us more solid arguments to defend our own interests when we have to confront those politicians.
given the hideous wars causes by the long line of psychos who tend towards power, how could it be any worse
I think the accountability issue is one of the worst ones, but not just for the owner of the AI, but the user too. Will modern LLM usage there seems to be a growing helplessness attitude of „well I asked ChatGPT and I followed its recommendation blindly, can’t be my fault what happened next“. I see people complaining about this a lot in places where LLMs are being used for programming. Some programmers are just generating a shit ton of terrible code with an AI tool, then sending that to get peer reviewed without any analysis from themselves.
We need a culture where anytime someone uses AI to generate information and acts on its, the user needs to be considered completely responsible for their actions. No „but the computer told me“ excuses. If you, for example, are a politician that uses an AI to construct sentencing policy, and that policy turns out to have extreme racial bias, the AI didn’t make the racist policy *you* did.
You can’t hire, fire, or send an AI to jail, but you can do those things to people that use AI, and if they understand they’re the ones on the chopping block if they use it bad, hopefully they’ll use it well.
I would trust an AI run government more than a people run government.
It definitely won’t be worse (any country applies). The problem is „who controls the machine it is running on“.
The biggest issue is that we can’t remove the human from AI. What I mean by that is that *someone* creates the AI, *someone* decides what information to feed to the AI, and *someone* decides when the AI is ready for use. Now, we could all vote on who makes those decisions and hell, we could all vote on each of those decisions individually, but *someone* (ie us) still made that decision. So now we face the issue of who updates the AI and how do we decide that? Does the AI decide that? Does it not ever get updated? Both of those answers result in future generations being governed by someone *we* chose, and I don’t know about you but I wouldn’t be too keen on being forced to have the same politicians voted in by my grandparents generation.
So then to solve that issue, we’d have to allow our children to re-do the same voting we did, and ignoring the economic and ecological costs, that’s more or less our current system with extra steps.
So the only way AI could really be in government is as an aid to human politicians. AI that helps politicians make informed decisions (those that would want to anyway…). AI that helps connect the people to politics and helps us stay informed. AI that helps us the people keep politicians in check.
Governing right now seems to have no basis on predictable logic or reason. Not sure how AI would handle it.
I think it’s inevitable, and I think we will be better off. As long as it is ethically coded and protected against hacking, I think an autonomous AI would be the least corrupt government ever established in human history.
How much worse could it possibly be. I wish I could turn off human politicians
nobody is going to hand over power to an AI. decisions and data interpretation will be assisted by AI though. They already have been for a long time in some spheres. US military in Afghanistan had ML programs that could predict when the Taliban would attack, just by reading data like satellite images of traffic.
I like a group of human beings voting on doing what the ai says to do that as opposed to giving full autonomy to it
We’re constructing something like an electronic autistic savant child with no capacity for empathy and feeding it with human-generated (hopefully) content of wildly varying quality. Imagine putting such a child in front of a TV for years and then sending it out into the world to make critical decisions. Probably not ideal.
I think an AGI that’s been tasked to determine the best outcome for the broadest spectrum of people without being open to bribery or financial incentive would likely achieve a better outcome for humanity. In fact, the future of our species is likely to be creativity and leisure, whilst robots and AGI perform all the tasks that made money a necessity in the first place. Money is a trade of time, but if nobody is spending time you don’t need money. A self sustaining society of robotics will hopefully remove the desire that some humans have to hold power over others.
And after watching your video (very nicely produced!) I broadly concur. My only thoughts on the case against would be that yes, right now the AI we have isn’t at the right level. But a sufficiently advanced AGI with all of the knowledge in the world and a complete understanding of the scientific and engineering methods should be able to determine what is and what isn’t. After all, we’re only biological machines ourselves. The main difference is that our hardware is predefined and our capacity for learning, whilst broad, is limited. An AGI should have no such limitation.
National leaders tend to be puppets already with written speeches to act and orate in front of the cameras…
An intelligent knowledgeable such proxy could be considered an upgrade!
I think I saw somewhere else where they talked about this in detail and it was eye opening…ai as long as it wasn’t controlled by someone, would actually do a better job than anyone leader ever has. Because its not going to be influenced, corrupted by the rich, it would do what’s best for people.
This would just be an oligarchy but with extra steps. Whoever has the most money would just buy whatever AI was running the government. It’s an absolutely terrible idea.
This is what I’ve been dreaming about for years. Glad to see people starting to share my idea.