17 April  (Friday) 4:15 PM  ONLINE*

Zoltán Sóstai

PhD Program in Logic and Philosophy of Science Eötvös University, Budapest

The construction problem of induction

Creativity, especially the ability of a physical system to create new theories on its own is one of the core properties sought in artificial general intelligence (AGI). According to David Deutsch, some AI systems - although seemingly capable of displaying knowledge - are not genuinely capable of creating theories on their own. For example, with behaviouristically interpreted artificial intelligence it's possible that the knowledge displayed by the AI system to maintain a meaningful communication is not created genuinely by the AI, but it's something that was already present in its program at the outset. In other words, it's possible that the knowledge displayed by the system was created earlier and elsewhere. Either by its programmer, or maybe even by some form of miraculous preadaptation. If Deutsch is right then we can ask the following question: Can we eliminate the possibility that the knowledge displayed by the AI system was created not by the system itself, but by some other process earlier and elsewhere? Answering this question would be an important step towards an explanation of creativity. According to the construction problem of induction - a variation of the classical problem of induction - it's not possible to construct a theory-creating physical system whereby we can eliminate this possibility and answer the question in a useful way. References: Deutsch, David, The Beginning of Infinity: Explanations That Transform the World. Chapter 7: Artificial Creativity. New York: Viking, 2011. Deutsch, David, Creative Blocks - How close are we to creating artificial intelligence?, Aeon magazine, https://aeon.co/essays/how-close-are-we-to-creating-artificial-intelligence Popper, Karl R. 1959. The Logic of Scientific Discovery (2002 edition), pp. 442-446., Appendix *X: Universals, Dispositions, and Natural or Physical Necessity Popper, Karl R. 1968. Conjectures and Refutations: The Growth of Scientific Knowledge (2nd edition). Chapter 1/IV., pp. 42-45., New York: Harper & Row.

24 April  (Friday) 4:15 PM  ONLINE*

Mátyás Lagos

MA Program in Logic and Theory of Science Eötvös University, Budapest

What makes a pattern complex?

This talk explores a property of formal languages that was defined as an attempt to explain why some languages are more computationally complex than others. We will compare a few languages using this property and see if it can really be considered as an indicator of computational complexity.