There are already many forms of artificial intelligence (AI) in the world today. Weak or narrow AI is a kind of machine-based expertise that is focused on a single or very narrow range of tasks. This is already very widely available. Siri and Alexa might be thought of as weak AI, although they are already sophisticated compared to achievements only a decade ago1: they combine speech recognition and natural language processing to ‘understand’ what you are asking them. They then have a set of capabilities that they can act upon (such as making a call, setting an alarm or ordering flowers for Mother’s Day on the internet, but not something in the real world such as making a cup of tea).
The current level of even the most sophisticated AI in the cement industry would be regarded as weak or narrow, since it can do a small range of jobs very well. These include the following:
- Fine-tuning pyroprocessing systems to optimise fuel mix, flame attributes, air-flow, feed rates, damper settings etc, in order to achieve a better outcome than even the best operator;
- Listening-to and understanding the spectrum of vibrations from a mill or fan and diagnosing any problems, long before a human could do so;
- Optimising delivery truck logistics and planning in real-time using GPS and neural networks, beyond the capabilities of any human handler.
These narrow AIs use a combination of approaches to solve their tasks, including algorithms, neural networks, fuzzy logic and machine learning (where optimised outcomes are iteratively achieved without explicit training, using statistics2).
Computers have also mastered some other skills, such as chess, Go and Jeopardy (in the last case, the computer, IBM’s Watson, first has to work out what the question means, then works out some possible answers based on the information it has to hand - 200 million pages of text - then work out which answers are most likely to be correct and then give the best answer - all faster than the very best players of all time 3). It turns out that the game-playing ability of Watson can be used to provide services for almost any information-heavy industry, including healthcare, banking, insurance, and telecoms. The original machine that won Jeopardy was as big as a master bedroom, but is now the size of a microwave oven. It’s also 240% faster and capable of answering simultaneous questions from multiple users. Adding image processing and the ability to understand other ‘dark data’ such as graphs and photos will further enhance Watson’s abilities. However, there are plenty of other Watson-like AI alternatives out there on the market already, available as a service on a rental plan. AI is already out there - who knew?
However, all this weak or narrow so-called AI is just a warm-up for the real thing. Strong or general AI is understood to be expertise over a wide range of tasks, at least as good as a typical human. This does not currently exist, since it would be prohibitively expensive to endow a machine with all the myriad abilities that you have yourself (driving, drawing, chatting, thinking, feeling - and so much more). However, I can see a case for the construction of such an AI, in a robotic body - for piloting a craft to and then exploring Mars. Without doubt this AI robot should be named ‘Robert.’
Incidentally, a third form of AI is postulated - that of superintelligence - when AI far exceeds the capabilities of even the brightest of human minds - possibly coming after a putative moment of ‘singularity.’ Let’s see.
So, how will the various evolving forms of AI affect the construction industry? Peter Debney4 makes the following suggestions:
- AI will be used to create 3D maps, blueprints and construction plans, incorporating extensive automated Building Information Modelling (BIM);
- AI will be used to take over the administrative and project management roles;
- AI will be used to advise on how specific construction projects should be planned (for example the design and construction methods to build a bridge);
- AI will be used to make buildings more efficient in use, for example by automatically reducing a building’s energy consumption to the lowest possible level.
I can certainly see a time when cement plants and their logistics operations are almost entirely AI-controlled, possibly from some remote location. The cement trucks may be driverless, delivered to an AI-augmented construction site, to build AI-designed houses, buildings and bridges. Truly, the future is AI.
1 https://magoosh.com/data-science/siri-work-science-behind-siri/
