Digital ethics, trusted AI and IBM

Last week I gave a talk followed by a discussion at a university. The presentation was about the current state of Artificial Intelligence (AI) and AI research topics. A good chunk of the discussion was dedicated to fairness, trust and digital ethics. In the following, I am sharing some of the related links.

IBM Research has a site dedicated to  AI. On that, a section provides insight into topics on what they call Trusted AI. On the main IBM site is also a portal Trusted AI for Business, providing an introduction and overview for the non-research crowd. If you are interested and want to try out and learn about few problems hands-on, I recommend these links:

• AI Fairness 360 Open Source Toolkit: http://aif360.mybluemix.net/
• Detect the bias - a game and survey: http://biasreduction.mybluemix.net/
• Old, but still great: MIT Moral Machine:  http://moralmachine.mit.edu/

IBM experts are part of many public panels, workgroups and commissions. In Germany, there is the Enquete-Kommission "Künstliche Intelligenz - Gesellschaftliche Verantwortung und wirtschafliche, soziale und ökologische Potenziale". On the European level, it is the EU High-Level Expert Group on Artificial Intelligence.

Finally, as a showcase of current AI capabilities, I recommend this video of IBM Project Debater and the live debate at Think 2019. A short video explains how Project Debater works:


If you have feedback, suggestions, or questions about this post, please reach out to me on Twitter (@data_henrik) or LinkedIn.

Combining data to find suspect in over 700 highway shootings

In Germany a 57-year-old trucker was arrested over the weekend. He is accused of shooting over 700 times at trucks (mostly car transporters) and cars over a span of several years. Only few people got injured as this were drive-by shootings on highways and he was aiming at the load, not the drivers. As the case couldn't be resolved for some years it eventually up with German federal police. They used new devices to automatically collect license plate information at central highway locations plus cell tower data from mobile phones to drill down on the suspect.

As soon as the arrest was reported, a discussion about data privacy started in Germany (again). Is it ok to store all the license plate information, even if it was deleted after 10 days? Does a case like this warrant to obtain cell tower information (who had their mobile phone turned on around possible shooting locations and traveling on the highway)? Was it ok to combine the data to find the suspect? There are pros and cons to it and I don't want to comment in either direction. When I teach data management at university I tell my students to be aware of the amount of data they produce, e.g., by having a mobile phone, by traveling, by using the Internet. Such data is stored for billing purposes and for legal reasons (laws that require companies to keep data for 30 or 90 days or even longer). And when there is data, there is always an opportunity to work with it, legally or illegally.

Are you in favor of using such data, maybe within certain limits? Are you always aware of what data you produce? Any opinions or thoughts?

Bank Data R Us (not anymore)

The European Parliament voted down a temporary agreement that allowed US agencies to access information on bank transfers of EU citizens. Strong privacy concerns were the driving motivation behind the parliaments decision to stop the data access.

The decision is interesting for EU citizens in several aspects: One is the strong emphasize on data privacy and data protection, the second one is that the EU parliament made use of its veto powers and shot down a decision of the EU council, indicating that it cannot be ignored in the future.

(For those not living in the EU, here is more background on the EU Parliament. The members are directly elected by the EU citizens.)