Intelligent Agent Foundations Forumsign up / log in
Change utility, reduce extortion
post by Stuart Armstrong 295 days ago | 3 comments

A putative new idea for AI control; index here.

EDIT: This method is not intended to solve extortion, just to remove the likelihood of extremely terrible outcomes (and slightly reduce the vulnerability to extortion).

A full solution to the extortion problem is sorely elusive. However, there are crude hacks that we can use to mitigate the downside.

Suppose we figured out that a friendly AI should be maximising an unbounded utility function \(U\). The extortion risk is that another AI could threaten a FAI with unbounded disutility if it didn’t go along with its plans. This gives the extorting AI – the EAI – a lot of leverage, and things could end up badly if the EAI ends up acting on its threat.

To combat this, we first have to figure out a level \(z\) of utility that is a lower bound on what \(U\) could ever reach naturally and realistically.

By “naturally” we mean that \(U\) going below \(z\) would require not just incompetence or indifference, but some AI actively and deliberately arranging the lowering of \(U\). And “realistically” just means that we’re confident that getting \(U\) lower than \(z\) by chance, or having a \(U\)-minimising AI, are exceedingly low.

Then what we can do is to cut off \(U\) at the \(z\) level, replacing \(U\) with \(U'=\max(U,U(z))\). See \(z\) indicated by the red line on this graph of \(U'\) versus \(U\):

What’s the consequence of this? First of all, it ensures that no EAI would threaten to reduce \(U\) (the utility we really care about) below \(z\), because that is not a threat to the FAI. This reduces the leverage of the EAI, and reduces the impact of it acting on its threat.

Since levels of \(U\) below \(z\) are exceedingly unlikely to happen by chance, the fact the FAI has the wrong utility below \(z\) shouldn’t affect it’s performance much. And, even in that zone, the AI is still motivated to climb \(U\) above \(z\).

But we may still feel unhappy about the flatness of that curve, and want it to still prefer higher \(U\) to exceedingly low values. If so, we can replace \(U\) with \(U''\) as follows (the blue line is at \(z-1\)):

In this case, the EAI will not seek to reduce \(U\) below \(z-1\) (in fact, it will specifically target that value), while the FAI has the correct ordering of lower values of \(U\). The utility is weird around \(z\), granted, but this is a place where the FAI would not want to be and would almost certainly not reach by accident.

Though this method does not eliminate the threat of extortion, it does seem to reduce its impact.





NEW LINKS

NEW POSTS

NEW DISCUSSION POSTS

RECENT COMMENTS

[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

RSS

Privacy & Terms