# Event

**Title:**

Dr. Tom Sterkenburg (Univ. Groningen): The Impossibility of Universal Prediction

**Date / Time:**

17.10.17 / 18:30 - 20:00

**Organizer:**

Institut für Theoretische Philosophie

**Location:**

24.53.01 Raum 81

**Description:**

**Abstract**

Putnam (1963) reconstructed Carnap's program of inductive logic as the search for a 'universal prediction method,' and presented a diagonalization proof against the very possibility of such a notion. In my talk, I will revisit Putnam's argument to assess a tradition of work on universal prediction in theoretical computer science.

The proposal by Solomonoff (1964) that set off this tradition was indeed inspired by Carnap, and I will explain how the resulting theory can be seen as an explicit attempt to evade Putnam-style diagonalization. It promises the specification of methods of prediction that are universally optimal, pointing the way to a Reichenbachian vindication of induction. I will show, however, that this promise cannot be delivered, thus strengthening Putnam's argument.

Next, I will consider a second component of Putnam's charge: that Carnapian (in effect, Bayesian) methods are limited in ways that other methods need not be. While this claim of Putnam appears to be misleading, it directs us to an important theme. This the infamous Bayesian inability to deal with theory change, which is actually the more general fact of the 'fixity' of well-defined prediction methods. I will point out how this problem returns in recent work to establish a 'meta-inductive justification of induction,' based on modern developments in the tradition of universal prediction.

**Speaker **

Tom Sterkenburg's research is concerned with the philosophy of induction and the foundations of information-theoretic statistics and machine learning. He wrote his PhD thesis on the theory of universal prediction stemming from algorithmic information theory. This work he carried out at the CWI (the Dutch national research institute for mathematics and computer science) in Amsterdam and the Faculty of Philosophy of the University of Groningen. In November 2017 he will start a postdoctoral fellowship at the Munich Center for Mathematical Philosophy, LMU Munich.

Zurück