When?
On 25 April 2024 from 4.00 pm to approximately 5.30 pm

Where?
German Research Centre for Artificial Intelligence (DFKI)
Building D3 2, Reuse Room (Main Building, Room -2.17)
Stuhlsatzenhausweg 3, 66123 Saarbrücken

How can we ensure that the increasingly powerful AI systems of the future, which interact with us, behave in our best interests and do not spiral out of control? This control problem is a key challenge in the field of AI alignment, which is the subject of passionate debate amongst leading AI scientists and philosophers. Stuart Russell’s book Human Compatible, aimed at a wide audience, represents a particularly widely read contribution to this debate.

We are delighted to welcome Elijah Millgram, Professor of Philosophy at the University of Utah, who will present his current philosophical critique of Russell’s ideas:

Instrumentalism, Moral-Theory Overlays, and the Control Problem
How do we ensure that the much more capable AI agents of the future behave themselves and do not get out of line? The default response to concerns about alignment and the control problem imposes constraints on AI decision-making processes, which are typically adapted from some familiar moral theory. An examination of Stuart Russell’s representative proposal makes it clear that recommendations of this kind will not work, and demonstrates why it is a mistake to view the underlying issue as a question of control.

Following Millgram’s talk, there will be ample opportunity for questions and discussion (in English).

The event is organised by the Centre for European Research in Trusted AI (CERTAIN) in collaboration with the ‘Ethics for Nerds’ lecture series and is aimed at anyone interested in the long-term ethical challenges of AI development.

Further information:
Elijah Millgram’s homepage: https://www.elijahmillgram.net/
Information on Stuart Russell’s book: https://en.wikipedia.org/wiki/Human_Compatible