According to the externalist approach, subjective experience hypothesizes a processual unity between the activity in the brain and the perceived event in the external world. A perception loop therefore occurs among the brain's activities and the external world. In our work the metaphor of test is employed...

We outline eight characteristics of the environments, tasks, and agents important for human-level intelligence. Treating these characteristics as influences on desired agent behavior, we then derive twelve requirements for general cognitive architectures. Cognitive-architecture designs that meet the...

In this paper a generic adaptive agent architecture is presented that integrates the interaction between cognitive and affective aspects of mental functioning, based on variants of notions adopted from neurological literature. It is discussed how it addresses a number of issues that have recurred in...

One of the basic requirements of an intelligent agent is the ability to represent and reason about space. While there are a number of approaches for achieving this goal, the recent gains in efficiency of the Satisfiability approach have made it a popular choice. Modern propositional SAT solvers are efficient...

A benchmark in the ¯eld of Arti¯cial General Intelligence (AGI) would allow evaluation and comparison of the many computational intelligence algorithms that have been developed. In this paper I propose that an ideal benchmark would possess seven key characteristics: ¯tness, breadth, specificity, low...

Using the contemporary view of computing exemplified by recent models and results from non-uniform complexity theory we investigate the computational power of artificial general intelligence systems (AGISs). We show that in accordance with the so-called Extended Turing Machine Paradigm such systems can...

Theoretical entities are aspects of the world that cannot be sensed directly but that nevertheless are causally relevant. Scientifc inquiry has uncovered many such entities, such as black holes and dark matter. We claim that theoretical entities are omportant for the development of concepts within the...

Analogical reasoning plays an important role in the context of higher cognitive abilities of humans. Analogies can be used not only to explain reasoning abilities of humans, but also to explain learning from sparse data, creative problem solving, abstractions of concrete situations, and recognition of...

Probabilistic Logic Networks (PLN), a comprehensive framework for uncertain inference currently in use in the OpenCog and Novamente Cognition Engine AGI software architectures, has previously been described in terms of the ¨experiential semantics" of an intelligent agent embodied in a world. However,...

Rewards typically express desirabilities or preferences over a set of alternatives. Here we propose that rewards can be defined for any probability distribution based on three desiderata, namely that rewards should be real-valued, additive and order-preserving, where the latter implies that more probable...

Compression in the program space is of high importance in Artificial General Intelligence. Since maximal data compression in the general sense is not possible to achieve, it is necessary to use approximate algorithms, like AIXIt;l. This paper introduces a system that is able to compress data locally...

The construction of an artificial scientist, a machine that discovers and describes the general rules governing a variety of complex environments, can be considered an important challenge for artificial general intelligence. Recently, a computational framework for scientific investigation has been postulated...

This paper is about Algorithmic Probability (ALP) and Heuristic Programming and how they can be combined to achieve AGI. It is an update of a 2003 report describing a system of this kind (Sol03). We first describe ALP, giving the most common implementation of it, then the features of ALP relevant to...

The main goal of the Artificial General Intelligence field (AGI) to create human level intelligence is known as a very ambitious one. On the way to the field development there are many difficult problems to solve, like natural language translation, for example, which seem to share some 'hardness' properties....

According to the Ouroboros Model several occasions can be distinguished over the course of the general autonomous cyclic activity in which new concepts are established and associated memories are preferentially laid down. Whereas a rather standard habituation process can lead to the extraction of statistical...

A software architecture is described which enables a virtual agent in an online virtual world to carry out simple English language interactions grounded in its perceptions and actions. The use of perceptions to guide anaphor resolution is discussed, along with the use of natural language generation to...

Two new formal definitions of intelligence are presented, the "pragmatic general intelligence" and "efficient pragmatic general intelligence." Largely inspired by Legg and Hutter's formal de nition of "universal intelligence", the goal of these de nitions is to capture a notion of general intelligence...

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...

Universal Search is the asymptotically fastest way of finding a program that calculates a solution to a given problem, provided nothing is known about the problem except that there is a fast way of verifying solutions.

This paper argues that the CHREST architecture of cognition can shed important light on developing artificial general intelligence. The key theme is that 'cognition is perception'. The description of the main components and mechanisms of the architecture is followed by a discussion of several domains...

As machines become more intelligent, more flexible, more autonomous and more powerful, the questions of how they should choose their actions and what goals they should pursue become critically important. Drawing upon the examples of and lessons learned from humans and lesser creatures, we propose a hierarchical...

A new family of gradient temporal-difference learning algorithms have recently been introduced by Sutton,Maei and others in which function approximation is much more straightforward. In this paper, we introduce the GQ(lambda) algorithm which can be seen as extension of that work to a more general setting...

Achieving goals in a complex environment in which many players interact is a general task demanded from an AI agent. When goals of the players are given explicitly, such setting can be described as a multi-player game with complete information. We introduce a general model of such games in which states...

The evaluation of incremental progress towards ˜Strong AI¨ or "AGI" remains a challenging open problem. In this paper, we draw inspiration from benchmarks used in artificial commonsense reasoning to propose a new benchmark problem—the Toy Box Problem—that tests the practical real-world intelligence...

How to search the space of programs for a code that solves a given problem? Standard asymptotically optimal Universal Search orders programs by Levin complexity, implementing an exponential trade-off between program length and runtime. Depending on the problem, however, sometimes we may have a good reason...

In this paper the principle of minimum relative entropy (PMRE) is proposed as a fundamental principle and idea that can be used in the field of AGI. It is shown to have a very strong mathematical foundation, that it is even more fundamental then Bayes rule or MaxEnt alone and that it can be related to...

Gigamachine is our initial implementation of an Artificial General Intelligence (AGI system) in the O'Caml language with the goal of building Solomonoff's Phase 1 machine that he proposed as the basis of a quite powerful incremental machine learning system (Sol02). While a lot of work remains to implement...

The principle of minimum description length suggests looking for the simplest network that works well on the training examples, where simplicity is measured by network description size based on a reasonable programming language for encoding networks. Previous work used an assembler-like universal network...

The neuroethology is an interdisciplinary study among artificial intelligence, biology and robotics to understand the animal behavior and its underlying neural mechanism. We argue that the neuroethological approach helps understand the general artificial intelligence.

Satisfiability (SAT) testing methods have been used effectively in many inference, planning and constraint satisfaction tasks and thus have been considered a contribution towards artificial general intelligence. However, since SAT constraints are defined over atomic propositions, domains with state variables...

We argue that the ability to find meaningful chunks in sequential input is a core cognitive ability for artificial general intelligence, and that the Voting Experts algorithm, which searches for an information theoretic signature of chunks, provides a general implementation of this ability. In support...

I have argued that a simple but general formal theory of creativity explains many essential aspects of intelligence including science, art, music, humor. It is based on the concept of maximizing reward for the creation or discovery of novel patterns allowing for improved data compression or prediction....

The paper surveys the evaluation approaches used in AGI research, and argues that the proper way of evaluation is to combine empirical comparison with human intelligence and theoretical analysis of the assumptions and implications of the AGI models.

General intelligence requires open-ended exploratory learning. The principle of compression progress proposes that agents should derive intrinsic reward from maximizing "interestingness", the first derivative of compression progress over the agent's history. Schmidhuber posits that such a drive can explain...

A cognitive architecture specifies a computational infrastructure that defines the various regions/functions working as a whole to produce human-like intelligence [1]. It also defines the main connectivity and information flow between various regions/functions. These functions and the connectivity between...

Spatiotemporal reasoning is an important skill that an AGI is expected to have, innately or not. Much work has already been done in defining reasoning systems for space, time and spacetime, such as the Region Connection Calculus for space, Allen's Interval Algebra for time, or the Qualitative Trajectory...

With the expected growth in mobile robotics the demand for expertise to develop robot control code will also increase. As end-users cannot be expected to develop this control code themselves, a more elegant solution would be to allow the end-users to teach the robot by demonstrating the task. In this...

This paper proposes a continuous-time machine learning model that learns the chronological relationships and the intervals between events, stores and organises the learnt knowledge in different levels of abstraction in a network, and makes predictions about future events. The acquired knowledge is represented...

Explaining adaptive behavior is a central problem in artificial intelligence research. Here we formalize adaptive agents as mixture distributions over sequences of inputs and outputs (I/O). Each distribution of the mixture constitutes a "possible world", but the agent does not know which of the possible...

Here we present a framework for AGI inspired by knowledge about the only working prototype: the brain. We consider the neurobiological findings as directives. The main algorithmic modules are de¯ned and solutions for each subtasks are given together with the available mathematical (hard) constraints....

The evaluation of several agents over a given task in a finite period of time is a very common problem in experimental design, statistics, computer science, economics and, in general, any experimental science. It is also crucial for intelligence evaluation. In reinforcement learning, the task is formalised...