Date: May 11
Time: 8.30 – 13.00
Location: commng soon
Modern AI is to a large degree build on algorithms with many parameters. This ultimately means that in order to explain AI algorithms one has to understand these many parameters and their interaction. Due to the sheer amount of parameters (typically in the tens of millions) this is not immediate possible and modern AI algorithms are therefore often referred to as a ‘black box’ technology. Especially in critical decision making, using ‘black box’ technology is not acceptable, which has led to the birth of a new research field coined Explainable AI (XAI). The starting point in XAI has been methods to visualize the effect input variables, e.g. pixels in an image or words in a text, has on the output of the AI algorithm. While such methods render some insights into AI algorithms, they are clearly not answering questions like: ‘what explanations do the end-user actually want/need?’ or ‘what kind of bias is imbedded in the algorithm?’ etc.
The research field of XAI is an important stepping stone for developing and deploying trustworthy, bias-free and robust AI solutions in to society. Moreover, XAI calls for interdisciplinary collaborations in order to progress the field. To this end we will organize a physical workshop for AAU researchers working with or interested in XAI. The program will start with a general introduction to the topic of XAI followed by diverse presentations of how different AAU researchers across the faculties work with (and understand) the topic of XAI. During the latter there will be amble of opportunities for discussing both details and general questions.
It is important to stress that this event is NOT reserved for XAI algorithmic experts, but for ALL with an interest in XAI.
8.30 - 09.00: Breakfast
9.00 – 9.45: Introduction to XAI
9.45 – 10.00: Coffee break
10.00 – 12.00:
- Different perspectives on XAI followed by discussions
12.00 – 12.45: Lunch and networking
The workshop is physical, but the general introduction to XAI from 9.00-9.45 will also be streamed via Zoom.