3 ms·
My idea would be as follows. 1. The financial system is extremely automated already. Not bitcoin but the financial system in General. So the first step in this
by zitterbewegung 11y ago
My idea would be as follows.
1. The financial system is extremely automated already. Not bitcoin but the financial system in General. So the first step in this ethical AI system would be to train it to be an ethical hedonist. I am already an ethical hedonist so I am the domain expert. https://en.wikipedia.org/wiki/Hedonism https://en.wikipedia.org/wiki/Hedonism
2. Then I would say make an agent that would make purchases based on my own likes and dislikes. This agent would use whatever API's are necessary to do this.
3. The last step would be to take the agent that I create and say 'Based upon your opinion of me and us as a system how can we increase our joint utility'.
4. The last step would be to create a trust that would make the AI the executor of the trust and it would work through lawyers to implement the joint utility function.
- onion2k 11y agoI doubt hedonism would work once you have more than 1 AI - there would be competing goals in the overall system. What represents pleasure to one agent might make another less happy, so the two would 'fight' and (presumably) both end up in a less positive position. Instead, I would aim for something more like Epicureanism; that is to say, maximise your own happiness but only if it doesn't impact negatively on others. Applying philosophic principles to AI is quite interesting. This also raises the question of whether you would want an AI to consider ethics - if you think of it as something that works for you rather than humanity as a whole, and assuming that it's so much smarter than you that you can't comprehend what it's doing, having it do things that your ethical stance would stop you doing yourself would be distinctly advantageous. You could reap the benefits of screwing everyone over without even knowing you're doing it.