I've been posting a lot on value/reward learning recently, and, as usual, the process of posting (and some feedback) means that those posts are partially superseded already - and some of them are overly complex.
So here I'll try and briefly summarise my current insights, with links to the other posts if appropriate (a link will cover all the points noted since the previous link):
I'm modelling humans as a pair (p,R), where R is the reward function, and p is a planning algorithm (called a planer) that maps rewards to policies. An agent is trying to learn R.
The policy of a given human is designated πH. A pair (p,R) is compatible if p(R)=πH.
There is a no-free-lunch theorem for these (p,R) pairs. Once the agent has learnt πH, it can get no further evidence from observing the human. At that point, any compatible pair (p,R) is a valid candidate for explaining the human planner/reward.
Thus the agent cannot get any idea about the reward without making assumptions about the human planner (ie the human irrationality) - and can't get any idea about the planner without making assumptions about the human reward.
Unlike most no-free-lunch theorems, a simplicity prior does not remove the result. LINK
It's even worse than that: a simplicity prior can push us away from any "reasonable" (p,R). LINK
Ignoring "noise" doesn't improve the situation: the real problem is bias. LINK
The (p,R) formalism can also model situations like the agent "overriding" the human's reward. LINK
There are Pascal's mugging type risks in modelling reward override, where a very unlikely (p,R) pair may still be chosen because the expected reward for that choice is huge. LINK
There is also the risk of an agent transforming human into rational maximisers of the reward its computed so far LINK.
Humans are not just creatures with policies, but we are creatures with opinions and narratives about our own values emotions, and rationality. Using these "normative assumptions", we can start converging on better (p,R) pairs. LINKLINK
Even given that, our values will remain underdefined, changeable, and open to manipulation. LINK
Resolving those problems with our values is a process much more akin to defining values that discovering them. LINK
There are important part of human values that are not easily captured in the (p,R) formalism. LINK
A putative new idea for AI control; index here.
I've been posting a lot on value/reward learning recently, and, as usual, the process of posting (and some feedback) means that those posts are partially superseded already - and some of them are overly complex.
So here I'll try and briefly summarise my current insights, with links to the other posts if appropriate (a link will cover all the points noted since the previous link):