The SP theory of intelligence has advantages over several AI-related alternatives according to a new paper by Dr Gerry Wolff of CognitionResearch.org.
The SP theory, which has been under development for several years, is realized in the form of a computer model. From that computer model the researchers Dr Wolff and Dr Vasile Palade (of Coventry University) aim to develop a powerful SP machine, driven by an existing supercomputer.
The SP system is different from other AI systems because it simplifies and integrates ideas across an exceptionally broad canvass including artificial intelligence, mainstream computing, mathematics, and human thinking and perceiving. The key to this simplification and integration is the powerful concept of multiple alignment, borrowed and adapted from research in biochemistry. This provides a framework for several different aspects of intelligence, including learning, pattern recognition, several kinds of reasoning, problem solving, and more.
Multiple alignment could be the “double helix’ of intelligence” says Dr Wolff. “It could prove to be as significant for an understanding of intelligence as is DNA in biological sciences.”
There are several problems with “deep learning”, an idea which is popular at present. Problems include the fact that deep learning systems can make bad mistakes in recognizing things, it is often difficult or impossible to know how they reach their conclusions, they fail to account for how people can learn from a single experience, deep learning systems require unnaturally large amounts of data and processing power, and there are problems with generalization of their learning. The SP system provides solutions to these and other problems with deep learning.
Although deep learning has been making the headlines” says Dr Wolff, “it is likely that the SP system will provide a firmer foundation for the long-term development of AI. At the same time, it has the potential to deliver useful applications on relatively short timescales”.
The SP system has advantages over several other AI systems, as described in the paper.
NOTES
1 The new paper may be downloaded via “The SP theory of intelligence: distinctive features and advantages” (PDF, J G Wolff, IEEE Access, 4, 216-246, 2016, bit.ly/21gv2jT). Related papers are detailed, many with download links, on http://www.cognitionresearch.org/sp.htm.
2 Dr Gerry Wolff, the author of the new paper, may be contacted via jgw AT cognitionresearch.org or +44 1248 712962 (landline) or +44 7746 290775 (mobile).
3 Since people often ask what the name “SP” stands for, it is short for “Simplicity” and “Power” in accordance with “Occam’s Razor”. That principle says that a scientific theory should be as simple as possible without sacrificing its descriptive or explanatory power. Simplicity and power is also equivalent to compression of information, which is at the heart of how the SP system works.
Sign up for the free insideBIGDATA newsletter.
Speak Your Mind