Analytical inductive programming and evolutionary in- ductive programming are two opposing strategies for learning recursive programs from incomplete specifica- tions such as input/output examples. Analytical induc- tive programming is data-driven, namely, the minimal recursive generalization over the...

The evaluation of an AGI system can take many forms. There is a long tradition in Artificial Intelligence (AI) of competitions focused on key challenges. A similar, but less celebrated trend has emerged in computational cognitive modeling, that of model comparison. As with AI competitions, model comparisons...

A class of environments for teaching and evaluating AGI systems is described, modeled loosely on preschools used for teaching human children and intended specifically for early-stage systems aimed at approaching human-level, human-inspired AGI through a process resembling human developmental psychology....

In this paper, we discuss the possibility of applying rule abstraction, a method designed to understand emergent systems, to the physiology of the brain. Rule abstraction reduces complex systems into simpler subsystems, each of which are then understood in terms of their respective subsystems. This process...

The case is made that the path to AGI through cogni- tive and developmental robotics is compelling. Beyond the familiar argument that it keeps researchers honest by forcing their systems to cope with the real world, it encourages them to recapitulate the evolutionary de- velopmental path which gave rise...

Feature Markov Decision Processes (MDPs) [Hut09] are well-suited for learning agents in general environments. Nevertheless, unstructured ()MDPs are limited to rela- tively simple environments. Structured MDPs like Dynamic Bayesian Networks (DBNs) are used for large-scale real- world problems. In this...

This paper shows that a constraint on universal Turing machines is necessary for Legg's and Hutter's formal measure of intelligence to be unbiased. It also explores the relation of the No Free Lunch Theorem to formal measures of intelligence.

Most work on self-motivated agents has focused on ac- quiring utility-optimizing mappings from states to ac- tions. But such mappings do not allow for explicit, rea- soned anticipation and planned achievement of future states and rewards, based on symbolic knowledge about the environment and about the...

Artificial general intelligence (AGI) has no consensus definition but everyone believes that they will recognize it when it appears. Unfortunately, in reality, there is great debate over specific examples that range the gamut from exact human brain simulations to infinitely capable systems. Indeed, it...

The brain is autonomously active and possesses an on- going internal dynamics which continues even in the temporary absence of external sensory stimuli. New experimental evidences, and theoretical considerations, indicate that this eigendynamics plays a central role in regulating the overall cognitive...

Certain necessary features of general intelligence are more system-wide than others; features such as attention, learning and temporal grounding are transversal in that they seem to affect a significant subset of all mental operation. We argue that such transversal features unavoidably impose fundamental...

A great deal of contention can be found within the published literature on grounding and the symbol grounding problem, much of it motivated by appeals to intuition and unfalsifiable claims. We seek to define a formal framework of representa- tion grounding that is independent of any particular opinion,...

Inspired by mental imagery, we present results of extending a symbolic cognitive architecture (Soar) with general computational mechanisms to support reasoning with symbolic, quantitative spatial, and visual depictive representations. Our primary goal is to achieve new capabilities by combining and manipulating...

To solve the hard problem of consciousness we first note that all cognitive systems of sufficient power must get into difficulty when trying to analyze consciousness concepts, because the mechanism that does the analysis will bottom out in such a way that the system declares these concepts to be both...

Cognitive architectures aspire for generality both in terms of problem solving and learning across a range of problems, yet to date few examples of domain independent learning has been demonstrated. In contrast, constraint programming often utilizes the same domain independent heuristics to find efficient...

In this paper we present a comparison of several inductive programming (IP) systems. IP addresses the problem of learning (recursive) programs from incomplete specifications, such as input/output examples. First, we introduce condi- tional higher-order term rewriting as a common framework for inductive...

This paper represents a first step in attempting to engage the research community in discussions about evaluation of human-level intelligent systems. First, we discuss the challenges of evaluating human-level intelligent systems. Second, we explore the different types of claims that are made about HLI...

Animals display exceptionally robust recognition abilities to analyze scenes compared to artificial means. The prevailing hypothesis in both the neuroscience and AI literatures is that the brain recognizes its environment using optimized connections. These connections are determined through a gradual...

A novel method for simultaneously storing memories and allocating resources in AI systems is presented. The method, Economic Attention Networks (ECANs), bears some resemblance to the spread of activation in attractor neural networks, but differs via explicitly differentiating two kinds of "activation"...

A new cognitive model is presented for large scale representation of episodic situations and for manipulating these representations using the model's innate natural language processing mechnisms. As formulated,and implemented in part,the purpose of the model seeks to attain basic child level cognitive...

Feedback cycles are proposed as the general unit of in- tellect and intelligence. This enables the importation of space-time concepts from physics. The action-lens, a processing structure based on these measurable ob- jects, is defined. Larger assemblies of this structure may support the Lowen model...

In recent years the video game industry has experienced rapid expansion developing virtual environments that accurately mimic a real-world setting. However, the industry almost entirely relies on finite state machines for deploying computer-controlled characters within these environments. This has resulted...

The challenge of creating AGI is better understood in the context of recent studies of biologically inspired cognitive architectures (BICA). While the solution is still far away, promising ideas can be derived from biological inspirations. The notions of a chain reaction, its critical mass and scaling...

This paper presents an approach for integrating action in the world with general symbolic reasoning. Instead of working with task-specific symbolic abstractions of continuous space, our system mediates action through a simple spatial representation. Low-level action controllers work in the context of...

Human beings have an apparently innate ability to seg- ment continuous audio speech into words, and that abil- ity is present in infants as young as 8 months old. This propensity towards audio segmentation seems to lay the groundwork for language learning. To artificially repro- duce this ability would...

We show how a general-purpose cognitive architecture augmented with a general diagrammatic component can represent and reason about Large-scale Space. The diagrammatic component allows an agent built in this architecture to represent information both symbolically and diagrammatically as appropriate....

There is a great deal of variability in the way in which different language users understand a given natural language (NL) character string. This variability probably arises because of some combination of differences in language users' perceptions of its context-of-use (pragmatics), identity and mode...

Systems for general intelligence require a significant poten- tial to model a variety of different cognitive abilities. It is often claimed that logic-based systems ­ although rather suc- cessful for modeling specialized tasks ­ lack the ability to be useful as a universal modeling framework due to the...

In this paper, we will discuss a new approach to formally modelling belief change in systems of sentences with inter- dependency. Our approach is based on the paradigm called pointer semantics or revision theory which forms a funda- mental way of successfully understanding the semantics of logic programming,...

General purpose intelligent learning agents cycle through (complex,non-MDP) sequences of observations, actions, and rewards. On the other hand, reinforcement learning is well- developed for small finite state Markov Decision Processes (MDPs). So far it is an art performed by human designers to extract...

We argue that Neural-Symbolic Integration is a topic of central importance for the advancement of Artificial General Intelligence.

Traditional machine learning systems work with relatively flat, uniform data representations, such as feature vectors, time-series, and context-free grammars. However, reality often presents us with data which are best understood in terms of relations, types, hierarchies, and complex functional forms....

This extended abstract outlines a project to build computer programs that understand. Understanding a domain is defined as the ability to rapidly produce computer programs to deal with new problems as they arise. This is achieved by building a CAD tool that collaborates with human designers who guide...

The Turing test was originally conceived by Alan Turing [20] to determine if a machine had achieved human-level intelligence. Although no longer taken as a comprehensive measure of human intelligence, passing the Turing test remains an interesting challenge as evidenced by the still unclaimed Loebner...

One of the most admirable characteristic of the hu- man cognitive system is its ability to extract gener- alized rules covering regularities from example expe- rience presented by or experienced from the environ- ment. Humans' problem solving, reasoning and verbal behavior often shows a high degree of...

We elaborate on a potential resolution of the homunculus fallacy that leads to a minimal and simple auto-associative recurrent `reconstruction network' architecture. We insist on Hebbian constraint at each learning step executed in this network. We find that the hidden internal model enables searches...

The "China Brain Project" is a 4 year (2008-2011), 10.5 million RMB research project to build China's first artificial brain, which will consist of 10,000- 50,000 neural net modules which are evolved rapidly in special FPGA hardware, downloaded one by one into a PC or supercomputer, and then connected...

A conversational English text parsing and generation system is described in which its lexicon and construction grammar rules are revised, augmented, and improved via dialog with mentors. Both the parser and generator operate in a cognitively plausible, incremental manner. Construction Grammar is well...

Relevance Based Planning (RBP) is a general method that plans in interaction with a domain simulation and domain specialized procedures. I argue that exploitation of the properties of causality and Euclidean topology which hold in many domains is a critical inductive bias necessary if an AGI (or any...

In previous papers we presented a robot cognitive architecture organized in three computational areas. The subconceptual area is concerned with the processing of data coming from the sensors. In the linguistic area representation and processing are based on a logic-oriented formalism. The conceptual...

The general intelligence of any autonomous system must in large part be measured by its ability to automatically learn new skills and integrate these with prior skills. Cognitive architectures addressing these topics are few and far between ­ possibly because of their difficulty. We argue that architectures...

Case-by-case Problem Solving solves each occurrence, or case, of a problem using available knowledge and resources on the case. It is different from the traditional Algorithmic Problem Solving, which applies the same algorithm to all occurrences of all problem instances. Case-by-case Prob- lem Solving...

This paper analyzes the different understandings of "embodiment". It argues that the issue is not on the hardware a system is implemented in (that is, robot or conventional computer), but on the relation between the system and its working environment. Using an AGI system NARS as an example, the paper...

Lacking an operational definition of autonomy has considerably weakened the concept's impact in systems engineering. Most current "autonomous" systems are built to operate in conditions more or less fully described a priori, which is insufficient for achieving highly autonomous systems that adapt efficiently...

One of the aims of Artificial General Intelligence(AGI) is to use the same methods to reason over a large num- ber of problems spanning different domains. Therefore, advancing general tools that are used in a number of domains like language, vision and intention reading is a step toward AGI. Probabilistic...