Towards a Democratisation of Artificial Intelligence

Károly Zsolnai-Fehér, Doctoral Researcher, Institute of Visual Computing and Human-Centered Technology, Technical University of Vienna

Transforming audio and video information has become easier than ever: through the recent ascendancy of neural network-based learning algorithms, researchers have been able to solve scientific problems that were deemed unfathomable until just a few years ago. This remarkable research progress left no branch of science untouched: self-driving cars, diagnosing eye conditions, cancer detection, and many more practical applications have surfaced recently, many of which will no doubt enhance our lives in the near future.

As is frequently the case with rapid technological progress, ethical and legal considerations are lagging behind. Today, we not only have to be vigilant about the potential misuse of these powerful methods, but have to make sure that we are also equipped to deal with such cases. This is far from trivial: for instance, it generally requires more effort to detect a piece of audio footage that has been tampered with than creating the forgery itself. It is also important to consider that such a detector network can potentially help an adversary to craft even more convincing forgeries, which makes the development of practical solutions even more challenging.

To address this, we have to endeavour to inform as many people about the existence of these techniques as possible, and make Artificial Intelligence research more open and accessible. This would lead to the democratisation of Artificial Intelligence where not only a few powerful and well-funded ivory towers, but every citizen and research institute is equally equipped to gain access to these techniques.

As many of these techniques are already in use, the urgency of these challenges grows rapidly over time — tackling these issues will require ample discussions and a renewed commitment to empower Artificial Intelligence research within Europe.