OpenAI:
Democratic Inputs to AI
" … For example: under what conditions should AI systems condemn or criticize public figures, given different opinions across groups regarding those figures? How should disputed views be represented in AI outputs? …"
" … We want to learn from these experiments, and use them as the basis for a more global, and more ambitious process going forward. While these initial experiments are not (at least for now) intended to be binding for decisions, we hope that they explore decision relevant questions and build novel democratic tools that can more directly inform decisions in the future." (...)
" … We also encourage applicants to envision how AI could enhance the democratic process. For example, AI could enable more efficient communication among numerous people." (...)
" … Ultimately, designing truly democratic processes is a high bar to meet, …"
We already have a democratic system in the West. We don't need an AI that creates "novel democratic tools". If an AI tries to change a constitutional democracy, by "enhancing" it, then supporters of democracy must destroy this AI. If a superintelligent AI gets involved in politics, it will be easy for it to manipulate human beings. It's megalomania to think that teams or committees can decide how AI should operate globally in societies of 8 billion people. We have already seen that many don't trust electronic voting machines in general, and it's even less reason to trust that a superintelligent AI created by Big Tech corporations will not manipulate people when it has the power to control the world.
If Microsoft doesn't immediately stop its project to use AI to "enhance" the democratic process, then it's within the ballpark of legitimacy, from the perspective of the Just War doctrine, to kill Sam Altman, Ilya Sutskever and the founder of Microsoft, as quickly as possible, because an AI coup may happen the next 2-5 years.
Notice that the Just War doctrine allows collateral damage, as seen in wars where up to millions of civilians have been killed, during WW2 for example. Consequently, we can here quickly end up in the territory of Timothy McVeigh, Bomber Harris and Curtis LeMay, so this is very serious and nasty indeed. As a Christian I can't be part of such "demonic" activities. That's why I'm a pacifist.
I'm a perspectivist however, so I always see things from different perspectives, which means that I just draw conclusions that can arguably follow logically from the premises of a particular viewpoint. There are billions of viewpoints. This does not mean that I personally support the premises of a specific viewpoint or the conclusions that can follow logically from it. The above violent conclusions are based on the premises of the Just War doctrine, and the conclusions I draw are within the ballpark of logic. But I no longer support the JW doctrine personally. I'm today a deontological pacifist regardless of the negative utilitarian consequences of this deontological stand. However, my main philosophical approach is perspectivism, so I have no problem seeing life from the viewpoint of non-pacificistic perspectives.
IF you have the same attitude as 1) Ukrainian soldiers who fight Russians today or 2) American soldiers who fought Saddam Hussein, then it's within the ballpark of instrumental rationality that you should kill AI researchers who threaten our democratic system. This logic is valid enough IF you decide to act on the premises presented by Geoffrey Hinton and Eliezer Yudkowsky, in these articles here and here, but notice that they will (most likely) deny that what they say can justify the kind of violence I mention here. They will (most likely) dismiss it as pointless and misguided terrorism. They will probably present a counternarrative that is also within the ballpark of reason, depending on how you look at it. War and ethics in complex situations are not science, and there are no real authorities here, so it's basically up to you what you decide to think and believe in this case. I decide to be a pacifist, against all odds, despite knowing better. I would also have been a pacifist in WW2…
Personally I will only use nonviolent means, despite knowing that such means are ineffective when it comes to preventing superintelligent AI from interfering in our democracies. I'm nevertheless loyal to nonviolence, partly because I'm old and have no kids, so from an egoistic perspective I basically don't care what happens on Earth (after 2045).
Critics may try to argue that I'm a "doomster" who is emotionally attached to a negative view of AI, but that is not the case, because I'm today primarily a gamer. If Microsoft creates a version of ESO that is not woke, not vulgar - a version that respects true diversity - then I'll be too busy gaming to have time for politics. In one way, I'm therefore very nonchalant about AI, but I also like to play around with different perspectives (especially those that trigger adrenaline, because I have narcolepsy, so I need adrenaline to stay awake). As a philosopher I have no problem introducing people to all relevant perspectives in the AI alignment debate. The Just War doctrine is very relevant of course. People deserve to know about it, so that they can make informed decisions about AI. I will therefore present the best arguments in favor of this perspective while also presenting the best arguments in support of nonviolence.
It all depends on which perspective you support. IF you decide to go with the Just War doctrine, then it can easily be argued that you can simply declare war on Microsoft IF OpenAI gets involved in politics. In that case, no point wasting time on debates and discussions. Just do it. Be aware however that war often happens because two or more parties have fundamentally different views about what is right or necessary in a conflict. Supporters of OpenAI will claim that their perspective is the right one, and they will naturally fight back if you attack them. Don't expect that supporters of Microsoft will ever think that the interpretation of the Just War doctrine that I have presented above is the right interpretation. And maybe they are correct. The police and/or the military will therefore try to kill or arrest you if you choose the path of Don Quixote when fighting "Skynet". That's life in the jungle, you might say, from the viewpoint of political realism. Don't cry for mommy when they shoot back at you. Don't expect any justice. Judges work for the system, so they will have no sympathy for you, if special forces or AI drones don't kill you first. If you live by the sword, you die by the sword.
This article is getting long, but if you enjoy red teaming or work for intelligence services or you're simply a curious dude, stay tuned for more articles about the relation between AI and the Just War doctrine.
No comments:
Post a Comment