Artificial Intelligence — Lessons From Fire
Technology is something we have always had. The advancement that most shaped humanity was learning how to control fire, it enabled our ancestors to keep warm, cook, make better tools, and migrate to areas with a cooler climate.
Fire, while extremely useful, comes with many risks. There are countless examples of damage and loss of life caused through fire related accidents and also through intended misuse.
We are on the edge of another major advancement, that will enable changes to our lives in ways difficult to imagine. Instead of learning to harness one of the four elements that existed long before humanity, it will be entirely our creation. Something that has been talked about in books and movies and anticipated for nearly 100 years. I’m talking about Artificial Intelligence.
Through books and film, there are many examples of AI being helpful servants, evil overlords, and everything in between. The most common fear is that AI-powered robots will take our jobs. There have also been recent examples of algorithms and machine learning (basic AI) solutions that have delivered a racially biased outcome. It is worth noting that AI isn’t racist, it simply doesn’t have the computing capacity to apply wider societal context to the function performed. The bias, however, is often created through providing a limited data set, the fault lay with the creators of the AI instance.
Before we go further, it is worth pointing out that the basic premise of computing and communications is based on a binary outcome, on/off, yes/no, one/zero. Everything we do with a computer is supported by these small binary states, thousands per millisecond. It has been this way since computers were first invented. For complex activities to occur, we have simply increased the speed and power of the computers. Future generations will liken this method of computing to using an abacus.
The next stage of computing that will power AI is quantum. The basic premise of quantum computing is difficult to explain, in fact, current supercomputers are being used to assist in the design of quantum computing. For me, the easiest analogy to assist in understanding the difference between binary and quantum computing is to think of a legacy computer as using a coin. The coin has a binary value, placed on a table it will show one value at a time, increasing power simply adds more coins. Quantum computing, however, would use a sphere. The value of the sphere could be more of a one, more of a zero, and anywhere in between. The value we observe on the sphere could also vary depending on where we are sitting at the table.
Once quantum moves beyond the current infant stage, things will start to get interesting. Quantum computers could be used to create the next phase of computing, a capability we would struggle to comprehend. Sound farfetched? In 2017 it was reported that Facebook shut down an experiment when it became evident that two instances of AI had developed their own language to communicate. It was also discovered that the neural network used by Google to translate between human languages had developed its own language to handle the tasks it was assigned.
With all this uncertainty and a whole range of possibilities, we have risk and understandably fear. We’ve had human made cyber-attacks that have crippled health services and infrastructure, the societal impact could be likened to an arsonist starting a forest fire. In the future, the impact of misuse from a human or an errant AI algorithm could be worse.
Here is where I suggest that we take lessons from managing fire and apply it to AI. Given how transformational and risky the applications can be, really embed the lessons for development, deployment, and use. These include risk assessment, containment, control, detection, and response. With fire, we’ve all been educated on risk assessment, how a safe fire is contained, control of the fire using fuel, using devices to observe and alert when something is not right, and lastly how to respond when things go wrong.
While fire and AI aren’t the same, our relationship could very well be similar.
