A stimulating wide overview of the enormous potential of AI for interdisciplinary technological applications.Dr Vincenzo Tamma
The workshop “Introduction to AI” was similar to one that I had delivered at the 39th BCS SGAI International Conference on Artificial Intelligence in Cambridge. The premise was that AI is a broad field. So, even if you already know about part of it, you are unlikely to have encountered the full range of tools and techniques.
I started off with a general discussion of what AI means. My preferred definition is “mimicking human mental faculties with a computer”. That definition has its limitations, not least that current AI outperforms humans in some tasks. Nevertheless, the definition is nicely broad enough to cover a range of techniques.
I break those techniques down into knowledge-based and data-based (or computational) intelligence. The former set of tools explicitly capture specialist human expertise, such as narrow aspects of law, medicine, or science. Such “expert systems” became popular in the 1980s and remain an important technique today. However, most of the current excitement is around the data-based techniques, especially the latest generation of deep-learning neural networks. These models are great for spotting patterns in large quantities of data. That makes them useful for classification and prediction, but it is important to remember that they have no contextual understanding.
The workshop explained, with some risky live demonstrations, how several AI techniques work in practice. On the knowledge-based side, rules and case-based reasoning were introduced, as well as fuzzy logic and Bayesian updating, which sit at the intersection with the data-based tools. On the data-based side, different designs of neural networks were presented, along with genetic algorithms for finding optimal combinations of variables.
Practical problems require a mixture of the right tools for the right tasks. So, the workshop finished by highlighting a range of real-world examples that bring the techniques together like a team. Those examples ranged from rail-freight logistics to oral cancer diagnosis, from communications network management to automated umpiring, and from image interpretation to the control of manufacturing processes. It is the practical examples that excite me most and they generated a lively discussion.