Could you add an explanation of what “ACI” means in the beginning? I was a bit thrown off by this.
Regarding your proposal, maybe you want to compare it to Richard Jeffrey’s utility theory. He assumes that value/utility (“desirability”, degree of desire) is a function of a Boolean algebra of propositions, and propositions can be interpreted as sets of possible worlds in which they are true. So it seems quite similar to your approach. Actions can also be interpreted as propositions of the form of “I do X now” which the agent can make true.
Thank you for your introduction of Richard Jeffery’s theory! I just read some article about his system and I think it’s great. I think his utility theory built upon proposition is just what I want to describe. However, his theory still starts from given preferences without showing how we can get these preferences (although these preferences should satisfy certain conditions), and my article argues that these preferences cannot be estimated using the Monte Carlo method.
Actually, ACI is an approach that can assign utility (preferences) to every proposition, by estimate its probability of “being the same as example of right things”. In other words, as long as we have examples of doing the right things, we can estimate the utility of any proposition using algorithmic information theory. And that’s actually how organisms learn from evolutionary history.
I temporarily call this approach Algorithmic Common Intelligence (ACI) because its mechanism is similar to the common law system. I am still refining this system from reading more other theories and writing programs based on it, that’s why I think my old articles about ACI may contain many errors.
Again, thank you for your comment! Hope you can give me more advices.
Could you add an explanation of what “ACI” means in the beginning? I was a bit thrown off by this.
Regarding your proposal, maybe you want to compare it to Richard Jeffrey’s utility theory. He assumes that value/utility (“desirability”, degree of desire) is a function of a Boolean algebra of propositions, and propositions can be interpreted as sets of possible worlds in which they are true. So it seems quite similar to your approach. Actions can also be interpreted as propositions of the form of “I do X now” which the agent can make true.
Thank you for your introduction of Richard Jeffery’s theory! I just read some article about his system and I think it’s great. I think his utility theory built upon proposition is just what I want to describe. However, his theory still starts from given preferences without showing how we can get these preferences (although these preferences should satisfy certain conditions), and my article argues that these preferences cannot be estimated using the Monte Carlo method.
Actually, ACI is an approach that can assign utility (preferences) to every proposition, by estimate its probability of “being the same as example of right things”. In other words, as long as we have examples of doing the right things, we can estimate the utility of any proposition using algorithmic information theory. And that’s actually how organisms learn from evolutionary history.
I temporarily call this approach Algorithmic Common Intelligence (ACI) because its mechanism is similar to the common law system. I am still refining this system from reading more other theories and writing programs based on it, that’s why I think my old articles about ACI may contain many errors.
Again, thank you for your comment! Hope you can give me more advices.