Intelligent Agent Foundations Forumsign up / log in
Acausal trade: being unusual
discussion post by Stuart Armstrong 281 days ago | discuss

A putative new idea for AI control; index here.

Other posts in the series: Introduction, Double decrease, Pre-existence deals, Full decision algorithms, Breaking acausal trade, Trade in different types of utility functions, Being unusual, and Summary.

It’s important to remember that, in acausal trade, the other agents truly know nothing about you. They have some statistical image about the initial conditions throughout the universe, the probability of life evolving, they’ve maybe run some simulations: but they truly have nothing beyond that. Nor do you have any idea about them.

What that means is that if you are a particularly unusual entity, then no causally disconnected agent knows that. You can’t update them on the fact that you exist, and that they should therefore trade more with you.

(Or maybe you’re an agent of type \(A_1\) and agents of type \(A_1\) are either almost infinitely rare (with high probability) or somewhat more common, independently of other agent types. Based on the fact that you exist, you can update towards the fact that \(A_1\) are more common - but only \(A_1\)-type agents will be able to do this update, so those are the only agents you’ll have much trade with.)

So unusual agents are less likely to benefit from acausal trade, but they’re also less pressured to follow it. Suppose that there are \(\Omega\) humans, all maximising the same \(u\), and one will be designated utility-dictator.

By shear coincidence, your decision theory is disconnected from that of the rest of the humans, and you’ve been designated utility dictator.

Any causally separated agent doesn’t know this, and so you can reach your decision while making an \(1/\Omega\) difference to how they deal with humanity. In many cases, it can be rational to simply defect from acausal trade entirely.

If you have a different utility function from the other humans, all the more reason to defect, because you get almost no benefit from cooperation.

Of course, a sensible human community will resolve its internal differences before designating a utility-dictator. And the community can’t plan to benefit from this loophole, since that would mean you being designated is not so unlikely after all.

Things might be different if you think you might be in a simulation, but unusual agents aren’t very informative in simulations either.

Seeing all the other potential barriers and difficulties with acausal trade, it possible, if every existing agent is unusual enough and have distinct enough utilities, that no acausal trade will happen at all.





[Delegative Reinforcement
by Vadim Kosoy on Stable Pointers to Value II: Environmental Goals | 1 like

Intermediate update: The
by Alex Appel on Further Progress on a Bayesian Version of Logical ... | 0 likes

Since Briggs [1] shows that
by 258 on In memoryless Cartesian environments, every UDT po... | 2 likes

This doesn't quite work. The
by Nisan Stiennon on Logical counterfactuals and differential privacy | 0 likes

I at first didn't understand
by Sam Eisenstat on An Untrollable Mathematician | 1 like

This is somewhat related to
by Vadim Kosoy on The set of Logical Inductors is not Convex | 0 likes

This uses logical inductors
by Abram Demski on The set of Logical Inductors is not Convex | 0 likes

Nice writeup. Is one-boxing
by Tom Everitt on Smoking Lesion Steelman II | 0 likes

Hi Alex! The definition of
by Vadim Kosoy on Delegative Inverse Reinforcement Learning | 0 likes

A summary that might be
by Alex Appel on Delegative Inverse Reinforcement Learning | 1 like

I don't believe that
by Alex Appel on Delegative Inverse Reinforcement Learning | 0 likes

This is exactly the sort of
by Stuart Armstrong on Being legible to other agents by committing to usi... | 0 likes

When considering an embedder
by Jack Gallagher on Where does ADT Go Wrong? | 0 likes

The differences between this
by Abram Demski on Policy Selection Solves Most Problems | 1 like

Looking "at the very
by Abram Demski on Policy Selection Solves Most Problems | 0 likes


Privacy & Terms