In late 2018, Gartner ran a survey of more than 3000 CIOs on the trends in digital business. Among the many findings, they reported that, when asked which technology CIOs expected to be most disruptive, the most mentioned by a large margin was Artificial Intelligence (taking the place of data and analytics, which moved down into second place.)
In fact, 37% percent of the leaders surveyed confirmed that they had either already deployed AI into their business or that deployment was in short-term planning.
With AI set and ready to become a crucial part of human daily life in the upcoming years, it’s important to understand what AI can actually do.
In fact, AI is a broad term covering several subsets or types of artificial intelligence. These subsets can be divided by the type of technology required – some require machine learning, big data or natural language processing (NLP), for instance. These subsets can also be differentiated by the level of intelligence imbedded into an AI machine – more commonly known as a robot.
1. Reactive Machines
Reactive machines are the simplest level of robot. They cannot create memories or use information learnt to influence future decisions – they are only able to react to presently existing situations.
IBM’s Deep Blue, a machine designed to play chess against a human, is an example of this. Deep Blue evaluates pieces on a chess board and reacts to them, based on pre-coded chess strategies. It does not learn or improve as it plays – hence, it is simply ‘reactive’.
2. Limited Memory
A limited memory machine, as the name might suggest, is able to retain some information learned from observing previous events or data. It can build knowledge using that memory in conjunction with pre-programmed data. Self-driving cars for instance store pre-programmed data – i.e. lane markings and maps, alongside observing surrounding information such as the speed and direction of nearby cars, or the movement of nearby pedestrians.
These vehicles can evaluate the environment around them and adjust their driving as necessary. As technology evolves, machine reaction times to make judgements have also become enhanced – an invaluable asset in technology as potentially dangerous as self-driving cars. Improvements in machine learning also helps autonomous vehicles to continue to learn how to drive in a similar way to humans – through experience over time.
3. Theory of Mind
Human beings have thoughts and feelings, memories or other brain patterns that drive and influence their behaviour. It is based from this psychology that theory of mind researchers work, hoping to develop computers that are able to imitate human mental models. That is – machines that are able to understand that people and animals have thoughts and feelings that can affect their own behaviour.
It is this theory of mind that allows humans to have social interactions and form societies. Theory of mind machines would be required to use the information derived from people and learn from it, which would then inform how the machine communicates in or reacts to a different situation.
A famous but still very primitive example of this technology is Sophia, the world-famous robot developed by Hanson Robotics, who often goes on press tours as an ever-evolving example to the public of what robots are capable of doing. Whilst Sophia is not natively able to determine or understand human emotion, she can hold basic conversation and has image recognition and an ability to respond to interactions with humans with the appropriate facial expression, as well as an incredibly human-like appearance.
Researchers have yet to truly develop theory of mind technology however, with criticisms of Sophia for instance being that she is simply “a chatbot with a face”.
Self-awareness AI machines are the most complex that we might ever be able to envision and are described by some as the ultimate goal of AI.
These are machines that have human-level consciousness and understand their existence in the world. They don’t just ask for something they need, they understand that they need something; ‘I want a glass of water’ is a very different statement to ‘I know I want a glass of water’.
As a conscious being, this machine would not just know of its own internal state but be able to predict the feelings of others around it. For instance, as humans, if someone yells at us we assume that that person is angry, because we understand that is how we feel when we yell. Without a theory of mind, we would not be able to make these inferences from other humans.
Obviously, self-aware machines are, at present, a work of science fiction and not something that exist – and in fact, may never exist. As it is, we’re probably best focusing on the development of machine learning in our AI. A machine that has a memory, that can learn from events in its memory and then can take that learning and apply it to future decisions is the baseline of evolution in Artificial Intelligence. Developing this will lead to AI innovation that could turn society on its head, enhance how we live in the day to day exponentially and even save lives.