Research Thinking and Writing Toolbox DVA 403 Artifactual

Research Thinking and Writing Toolbox DVA 403 Artifactual and Natural Intelligence Symbolic, Sub-symbolic and Agent-based Gordana Dodig Crnkovic School of Innovation, Design and Engineering, Mälardalen University, Sweden 1

Thinking and Intelligence In our course, Research Thinking and Writing Toolbox, research thinking, along with writing, is a central topic. Research thinking as thinking in general is based on a set of abilities that we call intelligence, so let us start from learning some basics about how we today understand intelligence, the ways we think, acquire knowledge and produce knowledge. 2

What is Intelligence? 3

Intelligence This general ability is defined as a combination of a number of specific abilities, which include: : – Adaptability to a new environment or to changes in the current environment – Learning capacity for knowledge/skill acquisition – Capacity for reasoning and abstract thought – Ability to comprehend relationships/patterns/rules – Ability to evaluate and judge – Capacity for original and productive thought –…. 4

Intelligence Howard Gardner's theory of multiple intelligences identify at least eight different components: logical, linguistic, spatial, musical, kinesthetic, interpersonal, intrapersonal and naturalist intelligence. IQ tests address only linguistic and logical plus some aspects of spatial intelligence, while other forms of intelligence have been entirely ignored. 5

Artifactual/Artificial Intelligence – In an artifact, artifactual/artificial intelligence is such a behavior (function) which in humans would require (biological) intelligence. – The central functions include reasoning, knowledge, planning, learning, communication, perception and locomotion (movement, action). 6

Artifactual/Artificial Intelligence (AI) is the branch of computer science that aims to create the intelligence of artifacts/ machines. John Mc. Carthy coined the term AI in 1956. “Weak AI” refers to the use of software to specific problem solving, (e. g. expert systems). General intelligence (or "strong AI") is still a long-term goal of AI research (human-like intelligence). 7

Artifactual/Artificial Intelligence – In the beginning researchers started from human intelligence and tried to implement corresponding functions into machines (artifacts). – Today one starts to think about superintelligence – enhancements in natural level human intelligence. 8

Symbolic Intelligence: Deduction, Reasoning and Problem Solving Human ability to think was the first thing AI researchers tried to simulate. Early AI developed algorithms that mimicked the step-by-step reasoning that humans use to make logical deductions. However, soon it was evident that deduction is not enough. 9

Symbolic Intelligence: Deduction, Reasoning and Problem Solving A very central itelligent ability that human possess is our skill to handle uncertainty and incomplete (often even contradictory) information. Exact reasoning leads to the explosion of possible scenarios which must be analysed – known as ”combinatorial explosion”. 10

Symbolic Intelligence: Deduction, Reasoning and Problem Solving A big advantage of machines – their ability to perform exact and lengthy calculations is at the same time their problem – in real life we do not think perfectly exactly, but ”good enough”. Humans are taking into account relevant things, and neglecting irrelevant. How can machine know what is relevant? 11

Symbolic Intelligence: Deduction, Reasoning and Problem Solving Symbolic information processing: reasoning, on the level of language (natural or formal), that which we are aware of Sub-symbolic information processing: that which goes on in our brains and nervous system without our thinking of it – seeing, motion, feelings, etc. 12

Symbolic Intelligence: Deduction, Reasoning and Problem Solving Humans usually solve problems using fast, intuitive judgments (“feeling” on a level of sub-symbolic information processing) rather than step-by-step deduction from perfectly exact data. 13

Symbolic Intelligence: Deduction, Reasoning and Problem Solving Imitating sub-symbolic problem solving: embodied agent approaches emphasize the importance of sensorimotor skills to higher reasoning; neural networks (connectionist) research simulates the structures inside human and animal brains that give rise to this subsymbolic skill. 14

The Symbol Grounding Problem GOFAI Good Old-Fashioned Artificial Intelligence is an ironic description of the oldest original approach to AI, based on logic and problem solving in specific problem domains, for example chess playing. In the robotics research, the term is extended as GOFAIR ("Good Old Fashioned Artificial Intelligence and Robotics"). 15

The Symbol Grounding Problem The GOFAI approach is based on the assumption that the most important aspects of intelligence can be achieved by the manipulation of symbols, known as the "physical symbol systems hypothesis" /Alan Newell and Herbert Simon in the middle 1960 s/. The term "GOFAI" was coined by John Haugeland in his 1986 book Artificial Intelligence: The Very Idea, which explored the philosophical implications of artificial intelligence research. 16

The Symbol Grounding Problem GOFAI was the dominant paradigm of AI research from the middle 1950 s until the late 1980 s. The Symbol Grounding Problem is related to the problem of how words (symbols) get their meanings, and hence to the problem of what meaning itself really is. If symbols (words) always are explained with other symbols we get infinite regress. Somewhere symbols must be “grounded”! In what way does that grounding happen? 17

Sub-symbolic AI Opponents of the symbolic AI include roboticists such as Rodney Brooks, who aims to produce autonomous robots without symbolic representation (or with only minimal representation) and computational intelligence researchers, who apply techniques such as neural networks and optimization to solve problems in machine learning and control engineering. 18

Connectionist AI systems are large networks of extremely simple numerical processors, massively interconnected and running in parallel. The level of analysis at which uniformal principles of cognition can be found is the subsymbolic level, intermediate between the neural and symbolic levels. Symbolic level structures provide only approximate accounts of cognition. Paul Smolensky http: //web. jhu. edu/cogsci/people/faculty/Smolensky/ 19

Connectionist AI The Blue Brain Project simulation by reverse-engineering the mammalian brain. http: //bluebrain. epfl. ch/ 20

Connectionist AI A model of brain’s neocortical column, with a generic facility that could allow modeling, and simulation of any brain region for which the data are provided. http: //www. hiddengarments. cn/? tag=switzerland 21

Integrating the Approaches: Intelligent Agent Paradigm Nowadays, the term agent is used to indicate entities ranging all the way from simple pieces of software to "conscious" entities with learning capabilities. For example, there are "helper" agents for web retrieval, robotic agents to explore inhospitable environments, agents in an economy, and so forth. 22

Integrating the Approaches: Intelligent Agent Modelling An "agent" must be identifiable, that is, distinguishable from its environment by some kind of spatial, temporal, or functional attribute. Moreover, agents must have some autonomy of action and they must be able to engage in tasks in an environment without direct external control. 23

Agent Based Modelling Approach Agent-Based Modeling (ABM) is a relatively new computational modeling paradigm, is the modeling of phenomena as dynamical systems of interacting agents. Another name for ABM is individual-based modeling. This strongly resembles Marvin Minsky’s ideas of The Society of Mind and Douglas Hofstadter’s ideas about reductionism vs holism from his book Gödel, Escher, Bach: An Eternal Golden Braid. 24

References Basic material: – http: //en. wikipedia. org/wiki/Artificial_intelligence – http: //paul-baxter. blogspot. com/2007/01/lessons-for-symbolic-and-subsymbolic. html – http: //en. wikipedia. org/wiki/Society_of_Mind – http: //www. scholarpedia. org/article/Agent_based_modeling – http: //cogprints. org/3106/1/sgproblem 1. html Harnad, S. (1990) The Symbol Grounding Problem. Physica D 42: 335 -346. – http: //www. typos. de/pdf/2007_AI_without_representation_M&M. pdf Vincent C. Müller, Is there a future for AI without representation? Minds and Machines, 17 (1), 101 -15. 25