Currently the topic of artificial intelligence dominates every discussion about digitization. As a former researcher on open systems and trust- and security mechanisms, this development has prompted our Head of Product Management Dr. Yvonne Bernard to take a closer look. In her recently published article „AI – the same procedures as last century?“ she provides a view behind the current hype. In the following interview with Yvonne, we will explore the background of this innovative technology, take a look at the implementation of artificial intelligence in an entrepreneurial context, and in conclusion discuss its potential in IT security.
So, what made you decide to take a further look at AI?
Especially in recent years, I have seen an enormous increase in AI technologies applied and – perhaps more importantly – advertised by technology companies and vendors around the world.
Since I have been dealing with this topic in research and teaching for several years, I was really curious: Have the mechanisms that I used and taught at Leibniz Universität Hannover developed further? Basic research takes up to 20 years, as they say, to be separated (if at all) from basic research in business-relevant technology, but to be honest, some of the features we used back then, such as artificial neural networks, were already older than me.
If you say this technology has been around for decades, why is it actually being applied just now?
Nowadays, what makes the implementation of AI technologies really worthwhile is the possibility to store and process large amounts of data and to adapt the processing schemes if necessary. Big data doesn’t mean storing everything and then looking at what you do with it: you have to think about data types in order to calculate efficiently and effectively on the basis of these data volumes. Also, the promotion of these technologies, which have been in use for years, has of course made its contribution to the hype. Furthermore, the growing number and quality of libraries that are available to the public and not only to researchers is a further aspect. You don’t have to spend much time looking for suitable software or frameworks to realize your AI ideas in functional code. Frameworks such as TensorFlow, Caffe and CNTK can be mentioned here. Thus, AI is increasingly used for the fast and (nearly) optimum solution of real problems.
Stay in touch
Sign up to get the latest News about Cloud Security.
What has made the use of AI possible in companies and what is the necessity?
As already mentioned, the increasing number and quality of libraries and the possibility to work with large amounts of data are the main growth drivers of the use of AI in the business environment. In addition, completely new and additional techniques such as supervised machine learning can be applied. In this case, a certain amount of the available total data is used, which is assumed to be very similar to the data for which the algorithms are trained for. An “unlearning” of desired characteristics is thus to be excluded.
To compare: In research laboratories, it is always important to make sure that the algorithms to be applied are well parameterized and suitable for the targeted data set. In business life one often does not want to and cannot spend this time to evaluate every possible parameter set. Moreover, a learning algorithm that learns something unexpected is great for a researcher but cannot be tolerated in business.
In which industries and processes do you see the greatest opportunities for the application of artificial intelligence?
It is safe to say that AI will not be the only solution to each of today’s problems. But there are areas where AI techniques are easier and more accessible than ever, and nothing should prevent developers and system developers from using the former pure research technology in any way that helps them find a good (or if possible the best) solution to their real problems. I would like to emphasize that – also at Hornetsecurity – many procedures from the quantity of AI methods have already been used successfully for years. In the past, however, such techniques were not advertised consciously, whereas today AI is perceived as a quality criteria or at least as an innovation. In general, the application is generally widespread in the area of optimization procedures and is also recommended, since simple heuristics are often not sufficient in terms of quality, but the determination of the optimal solution would not be possible in the desired time due to the complexity of processing times. Suitable learning methods can achieve excellent results in a short time – if you know how to use them wisely. Optimization processes can be found in almost all industries.
And finally: Do you think that artificial intelligence will influence and change IT security?
Yes, absolutely, but in both ways: Not only security research, but also attackers will increasingly use the accessibility of these technologies. With our comprehensive understanding and many years of experience in this field of algorithms, Hornetsecurity is well prepared for this “Arms Race”.