The Verdict Daniel S. Weld University of Washington.

The Verdict

Daniel S. WeldUniversity of Washington

Sleepless Nights

PlanningOn Trial

• Council for the Defense– 10% of AAAI & IJCAI on planning– Strength of ICAPS papers – Planning competition

• Scaling• Shared testbeds

Nagging Doubts

The Prosecutor

Says

•Where are the applications?

•Easy! There are applications…– Manufacturability analysis– Robotic path planning– Managing the power grid– Choosing airborne observations

But

•Do off-the-shelf planning methods apply to those applications?

– Mostly, no!

– Hence, why are we building them?

The Jury’s Still Out

•Choose the right problems– Aim towards real applications– Not towards incremental improvement

•Other fields of AI– Have the same problems– We need their respect … for jobs if nothing else