Gazing into the AI Black Box

Fraunhofer HHI and BIFOLD researchers present an AI-supported image evaluation method at the Open Day of the Federal Ministry of Education and Research (BMBF). With the help of this presentation, they explain to visitors at eye level how decision-making in neural networks takes place. In this way, the scientists present a central part of their research. BIFOLD cordially invites all visitors to participate in the demonstration to delve deeper into basic AI research with experts.

AI Showcase

Is it explainable how AI makes decisions? Yes, it is, but there is still a lot of research, which needs to be done. The research field that delves deeper into this question is called Explainable Artificial Intelligence, or XAI. At the Open Day of the Federal Ministry of Education and Research (BMBF) researchers from Fraunhofer HHI and BIFOLD will explain how a neural network calculates its answers or makes a decision using the example of an AI-supported image evaluation process. This demonstration is called: „Ein Blick in die Blackbox Künstliche Intelligenz“.

There are many different AIs. One example are AI-supported image evaluation procedures used in cancer diagnostics. The AI can support the physician in the assessment of cell photographs. It can calculate whether the cells shown are diseased or healthy based on pictures. A doctor can use this to decide how to treat a patient further.

The presentation shown at Open Day is more straightforward. Here, the focus is not on cell images but on photos of everyday life. The presentation shows that an AI calculates its answers in probabilities and which elements of the image are used for this calculation. The evaluation starts with the selection of a photo. Then, using a keyboard, one enters a simple question about the selected image. For example, on a picture showing two people, one might ask, "Is the person on the left female?" Now, the AI answers its response on a screen. But it doesn't respond with a simple "yes" or "no" answer, but it indicates a probability in percentage: Yes (70%), No (28%), Can be (1%). In addition, the AI shows which parts of the image were relevant to its decision by fading all irrelevant details to white.

"Ein Blick in die Blackbox Künstliche Intelligenz" is just one of many presentations at the Open Day of the Federal Ministry of Education and Research. The Fraunhofer HHI and BIFOLD team - Leila Arras, Lorenz Hufe, Bruno Puri, and Dr. Ekkehard Schnoor - eagerly look forward to answering visitors' questions and inviting everyone to play with AI-based image analysis.

The presentation is based on the "Dual Recurrent Attention Units for VQA" research work, which was funded by the BMBF via Fraunhofer HHI and BIFOLD. Link: https://arxiv.org/abs/1802.00209. A selection of XAI demonstrations can be found here: https://lrpserver.hhi.fraunhofer.de/.  

The Open Day of the Federal Ministry of Education and Research is a partial event of the Open Day of the Federal Government.
Location: Federal Ministry of Education and Research (BMBF), Kapelle-Ufer 1, 10117 Berlin
Date: August 19 & 20, 2023
Time: 10:00 am to 6:00 pm 
Program details (german): Open Day Agenda