How Safe Is Artificial Intelligence?
Machine learning dramatically changes our civilization. We rely more and more on efficient algorithms, because otherwise the complexity of our civilizing infrastructure would not be manageable: Our brains are too slow and hopelessly overwhelmed by the amount of data we have to deal with. But how secure are AI algorithms? In practical applications, learning algorithms refer to models of neural networks, which themselves are extremely complex. They are fed and trained with huge amounts of data. The number of necessary parameters explodes exponentially. Nobody knows exactly what happens in these “black boxes” in detail. A statistical trial-and-error procedure often remains. But how should questions of responsibility be decided in, e.g., autonomous driving or in medicine, if the methodological basics remain dark?
In machine learning with neural networks, we need more explainability and accountability of causes and effects in order to be able to decide ethical and legal questions of responsibility!
- 1.Peters J, Janzing D, Schölkopf B (2017) Elements of causal inference. Foundations and learning algorithms. The MIT Press, Cambridge, p 6 f.Google Scholar
- 2.Vapnik VN (1998) Statistical learning theory. Wiley, New YorkGoogle Scholar
- 3.Peters et al (Note 1), 6, after Figure 1.1Google Scholar
- 4.Mooij JM, Janzing D, Schölkopf B (2013) From ordinary differential equations to structural causal models: the deterministic case. Proceedings of the 29th Annual Conference on Uncertainty in artificiaL Intelligeence (UAI), pp 440–448Google Scholar
- 6.Peters et al (Note 1), p 84, after Figure 6.1Google Scholar
- 7.Peters et al (Note 1), p 144, after Figure 7.1Google Scholar
- 8.Foundation Inc OEIS (2017) The on-line encyclopedia of integer sequences. http://oeis.org/A003024.2017. Accessed 24 Nov 2018
- 11.Knight W (2017) The dark secret at the heart of AI. MIT Technol Rev 11:1–22Google Scholar
- 14.Nipkow T, Paulson LC, Wenzel M (2002) Isabelle/HOL. A proof assistant for higher-order logic. Springer, HeidelbergGoogle Scholar
- 15.Mainzer K (2018) How predictable is our world. Challenges for mathematics, computer science, and philosophy in the age of digitalization. Springer, WiesbadenGoogle Scholar
- 16.Bourque P, Dupuis R (2004) SWEBOK. Guide to the software engineering body of knowledge. IEEE Computer Society, Los AlamitosGoogle Scholar
- 17.Bertolino A (2007) Software testing research: achievements, challenges, dreams. In: Future of Software Engineering (FOSE’07) 0-7695-2829-5/07 IEEEGoogle Scholar
- 18.Tretmans J (1996) Test generation with inputs, outputs, and repetitive quiescence. Softw Concepts Tools 17(3):103–120Google Scholar
- 19.Tretmans J, Brinksma E (2003) TorX: Automated model-based testing. In: Hartman A, Dussa-Zieger K (eds) Proceedings of the First European Conference on Model-Driven Software Engineering 2003Google Scholar
- 21.Simon H (1947) Administrative behavior: a study of decision-making processes in administrative organizations. Macmillan, New YorkGoogle Scholar
- 23.Brown N, Sandholm T (2017) Reduced space and faster convergence in imperfect information games via pruning. International Conference on Machine Learning (ICML)Google Scholar
- 24.Economist Staff (2015) Blockchains: The great chain of being sure about things. The Economist October 31stGoogle Scholar
- 25.Narayanan A, Bonneau J, Felten E, Miller A, Goldfeder S (2016) Bitcoin and cryptocurrency technologies. A comprehensive introduction. Princeton University Press, PrincetonGoogle Scholar
- 26.crypto currency. Wikipedia. https://de.wikipedia.org/wiki/Kryptow%C3%A4hrung. Accessed 24 Nov 2018
- 27.Wikipedia (2017) Bitcoin. https://de.wikipedia.org/wiki/Bitcoin. Accessed 24 Nov 2018