A fundamental problem in artificial intelligence is that nobody really knows what intelligence is. The problem is especially acute when we need to consider artificial systems which are significantly different to humans. In this paper we approach this problem in the following way: We take a number of well known informal definitions of human intelligence that have been given by experts, and extract their essential features. These are then mathematically formalised to produce a general measure of intelligence for arbitrary machines. We believe that this equation formally captures the concept of machine intelligence in the broadest reasonable sense. We then show how this formal definition is related to the theory of universal optimal learning agents. Finally, we survey the many other tests and definitions of intelligence that have been proposed for machines.

In this paper we develop a computational framework for the measurement of different factors or abilities which are usually found in intelligent behaviours. For this, we first develop a scale for measuring the complexity of an instance of a problem, depending on the descriptional complexity (Levin LT variant) of the ‘explanation ’ of the answer to the problem. We centre on the establishment of either deductive and inductive abilities, and we show that their evaluation settings are special cases of the general framework. Some classical dependencies between them are shown and a way to separate these dependencies is developed. Finally, some variants of the previous factors and other possible ones to be taken into account are investigated. In the end, the application of these measurements for the evaluation of AI progress is discussed.

Abstract not found

www.hutter1.net Specialized intelligent systems can be found everywhere: finger print, handwriting, speech, and face recognition, spam filtering, chess and other game programs, robots, et al. This decade the first presumably complete mathematical theory of artificial intelligence based on universal induction-predictiondecision-action has been proposed. This information-theoretic approach solidifies the foundations of inductive inference and artificial intelligence. Getting the foundations right usually marks a significant progress and maturing of a field. The theory provides a gold standard and guidance for researchers working on intelligent algorithms. The roots of universal induction have been laid exactly half-a-century ago and the roots of universal intelligence exactly one decade ago. So it is timely to take stock of what has been achieved and what remains to be done. Since there are already good recent surveys, I describe the state-of-the-art only in passing and refer the reader to the literature.

Abstract not found

Abstract not found

Abstract. In this paper we analyse whether there is a subclass of environments that are more discriminative for intelligence measurement. We try to characterise this class as a kind of selection of those which do not have noise or randomness. We explore such a possibility and whether it can be formalised and put into practice. In order to do this, we first introduce a simple formalisation of ‘projectible ’ complexity which is valid for infinite strings and, consequently, for environments. From this result, we suggest an approach which both reduces the dependence on the reference machine and on the possible start-up garbage generated by an environment. More precisely, in order to avoid ‘noisy ’ environments, especially those where ‘noise ’ appears initially, we propose to let the environment play with a random agent for a certain number of interactions before letting the agent we want to be evaluated interact with the environment.

Abstract. Compression has been advocated as one of the principles which pervades inductive inference and prediction- and, from there, it has also been recurrent in definitions and tests of intelligence. However, this connection is less explicit in new approaches to intelligence. In this paper, we advocate that the notion of compression can appear again in definitions and tests of intelligence through the concepts of ‘mindreading’ and ‘communication ’ in the context of multi-agent systems and social environments. Our main position is that two-part Minimum Message Length (MML) compression is not only more natural and effective for agents with limited resources, but it is also much more appropriate for agents in (co-operative) social environments than one-part compression schemes- particularly those using a posterior-weighted mixture of all available models following Solomonoff’s theory of prediction. We think that the realisation of these differences is important to avoid a naive view of ‘intelligence as compression ’ in favour of a better understanding of how, why and where (one-part or two-part, lossless or lossy) compression is needed.

The first decade of this century has seen the nascency of the first mathematical theory of general artificial intelligence. This theory of Universal Artificial Intelligence (UAI) has made significant contributions to many theoretical, philosophical, and practical AI questions. In a series of papers culminating in book (Hutter, 2005), an exciting sound and complete mathematical model for a super intelligent agent (AIXI) has been developed and rigorously analyzed. While nowadays most AI researchers avoid discussing intelligence, the awardwinning PhD thesis (Legg, 2008) provided the philosophical embedding and investigated the UAI-based universal measure of rational intelligence, which is formal, objective and non-anthropocentric. Recently, effective approximations of AIXI have been derived and experimentally investigated in JAIR paper (Veness et al. 2011). This practical breakthrough has resulted in some impressive applications, finally muting earlier critique that UAI is only a theory. For the first time, without providing any domain knowledge, the same

The measurement of intelligence is usually associated with the performance over a selection of tasks or environments. The most general approach in this line is called Universal Intelligence, which assigns a probability to each possible environment according to several constructs derived from Kolmogorov complexity. In this context, new testing paradigms are being defined in order to devise intelligence tests which are anytime and universal: valid for both artificial intelligent systems and biological systems, of any intelligence degree and of any speed. In this paper, we address one of the pieces in this puzzle: the definition of a general, unbiased, universal class of environments such that they are appropriate for intelligence tests. By appropriate we mean that the environments are discriminative and that they can be feasibly built, in such a way that the environments can be automatically generated and their complexity can be computed.