The Awakening of Conscious Bots

Inside the Mind of the 2K BotPrize 2010 Winner

AIGameDev.com features an in-depth article about CC-Bot2, the winning entry (Conscious-Robots team) of this year’s 2K BotPrize competition at CIG.

Related links:

Related information:

  •  “Killer Bots Are Getting Human”. John Bohannon. (01 Oct. 2010). Science. Vol 330. no. 6000. pp. 30-31. Science Article.
  •  “Unreal Tournament 2010: Narrowing the Gap between Human and Bot”. Surfdaddy Orca. (10 Sept. 2010). h+ magazine Article.

Consciousness Evolves Besides Genetics

Raúl Arrabales Moreno, Machine Consciousness researcher at Carlos III University of Madrid

“CONSCIOUSNESS EVOLVES BESIDES GENETICS”

By Ana María Jaramillo V. (Translation of interview published by Blog Sistemas Inteligentes)

Inspired by the Strong Artificial Intelligence school, the same that captivated audiences in movies like ‘The Matrix’ or ‘2001 Space Odyssey’, this engineer by profession, multidisciplinary scientist by passion, believes the ultimate goal of Machine Consciousness research is to understand human nature. He pursues, as only a few do, the dream of creating self-conscious robots, as he asserts the best way to prove that something is understood is by recreating it.

Arrabales believes the real advancement of this field will come thanks to the synergy between mind research and technological disciplines. He knows he will live to see important qualitative changes and advocates the application of cognitive models from psychology or neurology to computational architectures.

This young scientist works in a controversial but fascinating field, where everyday research can be turned into fantasy, raising questions about free will and determinism in both humans and their creations.

AMJ: From what I understood reading you blog, you believe in the creation of artificial consciousness, don’t you?

RA: Yes, I believe so. However, it is not clear to me when and to what degree we will achieve this goal. Actually, one of the most important research lines I am currently working on is focused on the measure of the degree of artificial consciousness. There is no consensus about how to address this challenge. In fact, we don’t have a clear answer about the degree of consciousness of a coma patient. The definition of the term consciousness is a problem itself.

Continue reading “Consciousness Evolves Besides Genetics”

ConsScale. A Scale for Measuring Machine Consciousness

Measuring Machine Consciousness

ConsScale is a tool for assessing the functional level of consciousness of a creature. It has been specifically designed for the evaluation of Machine Consciousness implementations.

Now a ConsScale microsite is available where you can explore the conceptual levels of consciousness defined in the scale, learn how agents can also be rated using a quantitative score, and use the online calculator to rate your own implementations:

http://conscious-robots.com/consscale/

ConsScale is a framework for characterizing the cognitive power of a creature. ConsScale includes the definition of an ordered list of cognitive levels arranged across a developmental path. The arrangement of the levels is inspired on the ontogeny and phylogeny of consciousness in biological organisms.

The basic assumption is that there exist different kinds of minds, and they can be characterized in terms of ConsScale criteria. Using ConsScale, characterization and assessment of consciousness can be performed using three related tools:

– the ConsScale conceptual levels of consciousness (levels),
– the CQS (ConsScale Quantitative Score) (CQS), and
– the ConsScale radar graph representation (Calculator).

In order to assess the level of artificial consciousness of an agent using ConsScale, its architectural components have to be identified and its cognitive skills tested. Using this information as input, the scale can be used to obtain both a qualitative and a quantitative measure of consciousness:

ConsScale Process
ConsScale Process

CERA

CERA (Conscious and Emotional Reasoning Architecture) is a software architecture that allows the integration of different cognitive components into a single autonomous system. It is designed to be a flexible research framework in which different consciousness and emotion models can be integrated and tested. The CERA native components have been already implemented following the object oriented design methodology. Original design requirements are to fulfil nine modules of reasoning consciousness and their associated functionality [1]. These foundation classes can be extended and modified in a way that the desired models are represented and interrelated. This software engineering process is called CERA instantiation, as it produces a domain specific instance of CERA.

An instantiation called K-CERA (Khepera CERA) is described in [1], where we have adapted the foundation classes for the specific domain of unknown environment exploration using the Khepera robot. CERA foundation classes are designed to integrate reasoning consciousness with the rest of possible cognitive components of a model of the mind. CERA is structured in a three-layer architecture.

CERA layered design. The core layer is where the reasoning consciousness model foundation classes are located. Then, the instantiation layer adds the domain-specific cognitive systems. Finally, the top layer encloses the agent-specific perception and motor systems.

The inner layer, called CERA Core, encloses the reasoning consciousness model. Next layer is the instantiation layer, which contains the domain-specific cognitive components as discussed above. On top of the instantiation layer, an additional so called physical layer is required to adapt the cognitive components to the actual sensorimotor machinery of the autonomous robot.

The CERA core, which comprises the reasoning consciousness modules, defines a framework for implementing versatile cognitive processes. However, the knowledge representation is not concretely defined in this layer. An abstract knowledge class is used in CERA core in order to make the high level RCM processes definition representation-independent. This means that CERA core per se cannot be instantiated. A domain-specific instantiation layer is always required in order to build a complete cognitive model. Analogously, the physical layer is required in order to implement the actual autonomous agent control system.

[1] Arrabales Moreno, R. and Sanchis de Miguel, A. “A Machine Consciousness Approach to Autonomous Mobile Robotics”. In: 5th International Cognitive Robotics Workshop. AAAI-06. Boston, MA. July 2006.