My education (BEng from UCL in electronic and electrical engineering, MSc from University of London in solid state physics, and PhD from University of Sheffield in solid state electronics) led to research work in novel semiconductor device simulation for ministry of defense contracts at University of Sheffield, and then Rutherford Appleton Laboratory on UK and EU research contracts.
In 1989 I joined a startup in the University of Nottingham Science Park, which included working with neural networks, funded by an award from the UK government department of trade and industry. On Black Wednesday when the UK left the ERM our startup’s funder backed out and I started working as a consultant through my company Netnumerics. With Perot Systems (now acquired by Dell) I developed electricity generation price forecasts for East Midlands Electricity (now acquired by Powergen) using neural networks. I also built a Microsoft Excel add-in, Prognostica, for time series forecasting. During this time, I engaged with partners, including a boutique investment house, to apply neural networks to time series forecasting of financial markets. My models generated 55% accuracy in predicting the direction of major financial indices like the FTSE 100. We considered this not good enough (kind of mistakenly – interestingly I recently read that Jim Simons at Renaissance had similar statistics and with deep enough pockets could monetize it). I subsequently wrote up my knowledge of neural networks in a book (Azoff, 1994), out of print but soon to re-appear as a Kindle.
In the past two decades I’ve worked as a high-tech industry analyst, mostly at Informa business Ovum, and after an independent break currently back with Informa business Omdia as chief analyst in the cloud and data center practice. This book arose in my time as an independent consultant.
My interest in AI followed the invention of backpropagation, the breakthrough and excitement it stirred touched me around 1988. I joined a newly formed UK group with membership drawn from industry and academia, the Neural Computing Applications Forum, which led to the formation of a research journal published by Springer: Neural Computing and Applications. After Netnumerics in subsequent decades I kept my interest in AI on hold but began to cover it as an industry analyst with the emergence of deep learning in 2010-12, the rise of AI accelerator processors, and more recently by the breakthroughs with generative AI and LLM.