AI Austria on television!
On 18 October 2021, we were part of the program "Politik am Ring", in which every third Monday of the month speakers of the five parliamentary parties and experts discuss a current legislative proposal.
If you would like to watch the full episode of "Politik am Ring: Back from the future? Is Austria's strategy for artificial intelligence ambitious enough?", you can find the link here. (German only)
In August, at the Alpbach Technology Talks, the federal government's strategy for artificial intelligence, or AI for short, was presented. The two responsible ministers, Margarete Schramböck and Leonore Gewessler, spoke of an ambitious plan, while the opposition criticized this strategy as being too slow and too little.
In the meantime, there has also been a lot of criticism from scientists. In a joint statement by several research institutions, it literally says:
"The Austrian AI strategy is a bitter disappointment and a danger for the location."
So, ambitious or disappointing? Is the federal government's AI strategy suitable for making Austria fit for the future?
Clemens Wasner (AI Austria) gave an introduction to AI, answered questions about application areas, and talked about the EU AI Act:
What we understand by artificial intelligence today actually describes the fact that unlike in the past, where every single command had to be programmed by a programmer in order to recognize a face, for example, this is now learned by machines from so-called training data. This means that I feed a system with a sufficiently large number of images of people, and the system then learns on its own to recognize in the image how it can distinguish people from objects, for example.
The best-known examples that we encounter every day as an application of artificial intelligence are certainly, on the one hand, recommendation systems, keyword online shop, which suggests the next product you buy based on your personal history. Certainly, the most basic application of artificial intelligence takes place every day on the telephone. If you take a mobile phone and take a photo with it, then this depth of field effect, this bokeh effect, for example, is based on a neural network that recognizes who is in the front and what the background is, so that the background is then drawn softly.
This year, the first draft of the EU AI Act was published in April, which is currently being reviewed in the national parliaments and is to be implemented in mid-2023. It is very comparable to the GDPR, and it actually stipulates four risk classes:
What do I have to certify? - I don't have to certify machine maintenance software because it has nothing to do with people. If it's customer support and there's a chatbot, for example, then I have to show that it's an AI and not a human being behind it.
Everything that has to do with personal data, be it a human resources system or the segmentation of customers and so on, is also worthy of regulation. I really have to go through a certification process to prove that there is no malpractice. And then, at the very top of the pyramid in this four-level system, there are of course areas of application that are completely excluded, such as mass surveillance based on biometric signals, or that someone uses profiling in areas that have nothing to do with their core area.
Main objectives of the AI Strategy of the Austrian Federal Government
In August, Digitalisation Minister Margarete Schramböck of the Austrian People's Party (ÖVP) and Climate Protection Minister Leonore Gewessler of the Green Party present the Austrian federal government's AI strategy. The main goals are:
- to aim for broad use of AI-oriented towards the common good and carried out in a responsible manner on the basis of fundamental and human rights, fundamental European values, and the forthcoming European legal framework.
- Austria should position itself as a research and innovation location for AI in key areas and fields of strength.
- the competitiveness of Austria as a technology and business location is to be secured by means of the development and use of AI.
Experts argue that this is not enough.
"Austria does not have an AI strategy. Austria has a declaration of intent that is at the level of three years prior, i.e. 2018, compared to other national governments."
How can this be determined? - If we look at our neighboring countries, for example, Germany, Italy, or Slovenia, they have already allocated budgets in the three-digit million range, which are also backed by concrete measures.
The Austrian AI strategy contains declarations of intent such as: Yes, we have to see that something happens in education, we have to see that something happens in industrial application and research, and, among other things, we also deal very strongly with the whole regulation part, which, paradoxically, comes from the EU side - we no longer have any influence there at the moment.
After AI-Austria's introduction to the topic, the issues raised as well as other questions were further discussed by experts such as Sabine Köszegi (Austrian Council for Robotics and Artificial Intelligence), Michael Katzlberger (AI expert), and members of the National Council: Maria Theresia Niss (ÖVP), Petra Oberrauner (SPÖ), Gerhard Deimek (FPÖ), Süleyman Zorba (Grüne) and Helmut Brandstätter (NEOS).