Time(s)
Description
Heuristic search is a discipline under the larger umbrella in AI. Heuristic search algorithms find solutions to a variety of problems, including pathfinding, logistics, and scheduling. Most of the problems heuristic search is applied to are hard in the formal sense. This forces us to find new techniques to tackle larger problems and new domains — we can’t simply wait for improved
hardware to solve it for us. When developing new AI algorithms, understanding what has come before and why it fails to perform in a new domain is critical. In this talk, we’ll look at how algorithm visualization drove the development of state of the art AI algorithms for heuristic search.
Key takeaways from the session:
- I hope to impress people with the idea that the first step to improving anything, even the state of the art in a field, is to build an intimate understanding of that thing.
- Second, that visualizations are an important tool in building an understanding of the behavior of a thing, in this case an algorithm.
- Finally, that AI isn’t science fiction. The simplest algorithms are accessible to anyone that wants to invest time in understanding how AI interacts with the world around us.
About the speakers
As AI practice lead at SEP Jordan’s responsibilities include educating clients and peers about what AI can do, identifying complex problems AI can address in whole or in part, and incorporating AI into software products that make a difference for their users.
Prior to joining SEP, Jordan held roles as a research scientist for a startup, the Charles-Stark Draper Laboratory, and an R&D consultancy. In each of these positions, he did essentially the same thing: found and delivered feasible solutions to technically challenging problems that arise in a wide variety of domains and industries.
He completed his PhD in artificial intelligence at the University of New Hampshire in 2012. His research focused on using approximation and machine learning to solve problems known to be computationally intractable, such as automated planning and scheduling problems.