PRESS RELEASE: Advantages of the “SP theory of intelligence” over AI alternatives
Gerry Wolff
A new paper argues that the SP theory of intelligence has advantages over several AI-related alternatives, including the popular idea of “deep learning in artificial neural networks.”
The SP theory has been under development for several years and is realised in the form of a computer model. From that computer model the researchers Dr Gerry Wolff and Dr Vasile Palade aim to develop a powerful SP machine, driven by an existing supercomputer.
The SP system is different from most other AI systems because, right from the beginning, the aim has been to simplify and integrate ideas across a broad canvass including artificial intelligence, mainstream computing, mathematics, and human thinking and perceiving.
What has emerged from this attempt at 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.”
One of the main parts of the paper describes several problems with deep learning and how, in the SP system, they may be overcome.
One of the problems with a deep learning system is that, when it is used to recognise things, it can make mistakes. For example, it may correctly recognise a picture of a car but may fail to recognise another slightly different picture of a car which, to a person, looks almost identical. Because of the way it works, the SP system is unlikely to make this kind of mistake.
Another problem with deep learning systems is that it is often difficult or impossible to know how they reach their conclusions. This is likely to have legal implications if, as in driverless cars, AI systems make decisions on their own. The SP system has the advantage that all the workings of the system are transparent and open to inspection.
“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”.
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 www.cognitionresearch.org/sp.htm .
2 Dr Gerry Wolff, the author of the new paper, may be contacted via j...@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.