Artificial Intelligence (AI) has been just around the corner for a long time – but looks like we have now arrived.
Computers can beat us at Chess and Go and respond to voice commands.
Navigation systems are so good many of us now have never really learned to use a map.
There are so many ways of looking at and classifying the field of AI and machine learning that it’s almost impossible to get a sense of the field.
But we can start by looking at some broad domains – what do humans do a lot of the time?
We sense things – taking in vast quantities of visual, auditory and tactile information and responding to our environment.
We can detect the edges of things, work out which way is up or down and work out what is near us and if we are going to bang into them.
A particularly human thing to do is reason. Our brains are essentially prediction machines – we can think about what has happened and use reasoning to work out what we should do next.
But we don’t exist in isolation – as social creatures we interact with others – listening, speaking and responding.
We make plans – choose between alternatives or options – that range from what to eat to how to get somewhere.
We are also teleological – the conscious part of our brain helps us do things with purpose.
Our brains have evolved to be the way they are – but how would we go about creating an artificial one?
We could start by writing down all the rules we follow.
For example, doctors get to a diagnosis by considering and eliminating possibilities based on the symptoms they see and the measurements they take.
Rule based or expert systems take all this knowledge and use it to create if-then rules – if the temperature is above X, check Y next.
These systems are now pretty effective – and help us select the best flight, the cheapest online store for an item and schedule calendar entries from text in emails.
If there is too much data and variation to come up with rules, then we might use probabilistic approaches.
For example, we can run weather simulations that are probably accurate over hours or days but less so over weeks and months.
We can look at the distribution of a time series and use that to predict the range of probable future values – which then lets us pick out values and events that fall outside expected levels.
The rule based and probabilistic approaches are pretty easy to build and many systems in use now will be based on them.
A more complex approach is pattern matching, where a learning algorithm adjusts itself and learns from the data that goes into it.
For example, every time we type a search term into Google, we are training its AI engines. If we type in the word “eagle” and then click on a picture of an eagle, Google can learn what eagles look like and eventually predict that a picture contains an eagle.
With pattern matching, the more information we have the better our algorithm gets – and so it’s a winner take all situation where the systems we interact with most will learn the most and pull away from the rest.
But where can we use this technology now?
Three areas that are of interest in the energy sector are forecasting, scheduling and trading.
The energy system is all about balancing supply and demand, whether at the grid level or the domestic level.
If we know when the wind is going to blow, then we can make a call on the number of fossil fuelled power stations we need.
If we can see when demand or prices are high, we can schedule when we do work to avoid costs or take advantage of high prices.
We could even trade between ourselves – selling or buying electricity from the grid or a peer-to-peer network for a profit.
An interesting thing that happens with AI is that as it gets cleverer we tend to dismiss the things it does as simply something a machine can do.
As a result it is quietly augmenting how we do things without us really noticing. For example, how many of us now choose a different route based on Google’s recommendations first thing in the morning?
Many AI applications will be almost unnoticed, simply transforming the essential building blocks of our economic system.
Eventually, one hopes, AI will free humans up to do more creative fulfilling work and leave the mundane to the machines.