A New Constructivist AI: From Manual Methods to Self-Constructive Systems

Abstract

The development of artificial intelligence (AI) systems has to date been largely one of manual labor. This constructionist approach to AI has resulted in systems with limited-domain application and severe performance brittleness. No AI architecture to date incorporates, in a single system, the many features that make natural intelligence general-purpose, including system-wide attention, analogy-making, system-wide learning, and various other complex transversal functions. Going beyond current AI systems will require significantly more complex system architecture than attempted to date. The heavy reliance on direct human specification and intervention in constructionist AI brings severe theoretical and practical limitations to any system built that way. One way to address the challenge of artificial general intelligence (AGI) is replacing a top-down architectural design approach with methods that allow the system to manage its own growth. This calls for a fundamental shift from hand-crafting to self-organizing architectures and self-generated code-what we call a constructivist AI approach, in reference to the self-constructive principles on which it must be based. Methodologies employed for constructivist AI will be very different from today's software development methods; instead of relying on direct design of mental functions and their implementation in a cognitive architecture, they must address the principles-the "seeds"-from which a cognitive architecture can automatically grow. In this paper I describe the argument in detail and examine some of the implications of this impending paradigm shift.

Self-organizing multi-agent systems for the control of complex systems

“…Constructivism is the idea that humans have the ability to construct knowledge in their own mind through interactions with the environment. Constructivist artificial intelligence aims at designing self-constructive systems ( Thórisson, 2012 ). In such systems, not only the knowledge but the means to acquire it are learned.…”

“…Top-down classical methods have severe shortcomings when it comes to complexity: scale, integration, and flexibility ( Thórisson, 2012 ). This section presents the Adaptive Multi-Agent Systems (AMASs) theory, that aims at overcoming these limitations thanks to the natural modularity of MASs and the cooperative self-organization of agents.…”

Towards a General Attention Mechanism for Embedded Intelligent Systems

“…The significance of small processing units with homogenous computational complexity is that most processes take roughly the same amount of time to execute, making temporal aspects of performance more predictable, and that the system is highly interruptible and pre-emptive, as it never has to wait for time-consuming processes to complete before reacting to new data. An important feature of the design approach presented here is that it can be applied directly to systems that manage their own growth and expansion -constructivist architectures [12,13]. As the sum of internal system activity constitutes a large amount of information, an attention mechanism is required to manage resources for self-reconfiguration -in much the same way as for other task performance.…”

Bounded Seed-AGI

“…Unfortunately, none of the methodologies in the AI or CS literature are directly applicable for designing systems of this nature. For this reason we have advocated what we call a constructivist AI methodology (CAIM; [5][6][7][8][9]). In the main, our constructivist approach has two key objectives: (a) to achieve bounded recursive self-improvement [1] and generality and, (b) to uncover the principles for-and to actually build-systems that, given a small set of seed information, manage the bulk of the bootstrapping work on their own, in environments and on tasks that may be new and unfamiliar.…”