Friday, April 10, 2009

LinkedIn and an STL backstory

I'm on LinkedIn here.

For a long time I treated this pretty passively, but as LinkedIn is, apparently, the 'thinking person's FaceBook', I decided to treat it a little more seriously and so gave my profile a makeover.

One of the consequences was that I decided to try to add to my network some of the colleagues I used to work with back at STL in the early eighties - Will and Paul, Mike and Jill, Bernie and Mel. This is not the place for a memoir, but let me outline the background.

I had moved to STL from a systems programming job in 1982, and suddenly found myself in a first-rate research environment, tasked to develop part of a software support environment for 'formal methods' such as CCS, Z and VDM. My first assignment was to develop a term-rewrite system using Lisp (on a VAX 11/780).

My real interests however lay elsewhere, and having done my software engineering apprenticeship, I was able to set up a funded team to work on formalisations of Artificial Intelligence. This was the subject of the parallel Ph. D. programme I did at Surrey University between 1984 and 1988, supervised by Professor Bernie Cohen.

Sadly, following the takeover of STL by Northern Telecom/Bell-Northern Research, all this blue-sky stuff was terminated, and I had to re-launch my career in telecoms as a carrier network designer. I recall Professor Cohen being distinctly irritated, to put it mildly, that what looked like a promising field of further research had been so abruptly terminated. Perhaps a small explanation is owed to him here.

Bernie, at the end of my thesis I concluded that if we were going to understand human mental architecture (which includes concepts such as emotion, personality differences, intelligence and consciousness itself) we had to understand what specific environmental conditions held back there in the Pleistocene, such that human cognition was the appropriate consequential evolved response. I did some preliminary exploration, but there was almost no academic research to draw upon.

It must be a feeling familiar to physicists today. There were many, many directions which could be mathematically explored, but in the absence of experimental real-world feedback there was no ability to prune the cul-de-sacs and focus on the most promising directions.

The feeling that my research programme had stalled has been vindicated over the years. Subsequently there was the explosion of work in sociobiology (E. O. Wilson et al.), and the further development of evolutionary psychology with all its manifest strengths and weaknesses. AI research meanwhile stultified, while sociobiology continues to be the area I would work in, if I could restart my research programme today.

But we still don't understand enough about the ecological drivers of human cognition over the last 100-200,000 years. In fact this whole area, at a conceptual level, is subject to enormous contemporary debate.

It was Richard Feynman who said that if you can't construct it, you don't understand it. A very profound thought of course, and a pointer to the fact that the falsification of our best, speculative evolutionary psychological theories requires we can build agents effectively mimicking humans in social interactions. Sadly, that's a generation away though, showing what we have here is a genuine research programme, not simply a problem to be directly addressed and solved.