Solutions to problems with deep learning

[Gerry Wolff]

Dear Colleague,

Solutions to problems with deep learning

As you may know, this paper “Deep learning: a critical appraisal” (PDF, bit.ly/2CLJBY9), which has been posted recently on arXiv by Professor Gary Marcus, describes some of the shortcomings of deep learning.

I’ve now drafted a follow-up paper, posted on arXiv, “Solutions to problems with deep learning” (PDF, bit.ly/2AJzu4j). This describes how several of the problems with deep learning may be overcome via the SP theory of Intelligence, and the powerful concept of SP-multiple-alignment. The SP system has nine substantial advantages compared with deep learning.

On the strength of evidence in the paper, and other evidence, I believe the SP system provides a much firmer foundation for the development of artificial general intelligence (AGI) than does deep learning.

Papers from the SP programme of research may be downloaded via www.cognitionresearch.org/sp.htm .

Of course  it would be wrong to halt all research on deep learning, but there is certainly a case for opening up other avenues and I believe the SP framework is a good candidate for further development.

I would be happy to try to answer any questions you may have, and I would very much welcome your comments.

With best wishes,

Gerry Wolff

--

Dr Gerry Wolff PhD CEng MIEEE MACM MBCS

Cognition Research

j...@cognitionresearch.org, +44 (0) 1248 712962, +44 (0) 7746 290775, Skype: gerry.wolff, Web: www.cognitionresearch.org.