Predicting the future of AI is fraught with difficulty. Once upon a time, it was believed that handwriting recognition would be proof that a machine had achieved artificial intelligence (AI). However, now that the technology is widely available, it’s rarely considered to be evidence of ‘intelligence’. Users are of course still impressed by a machine’s ability to match patterns in handwriting, yet somehow this doesn’t convince them that the machine is really aware of, or understands what it is reading. Instead, it just shows how clever its designers were.
Our definitions of what AI really is are constantly changing. Dig into the literature around the topic and you’ll come across plenty of arguments about language—whether a technology is really an artificial intelligence, is in fact deep learning, machine learning, augmented intelligence or, for that matter, assisted intelligence. These differing definitions do actually matter. When people misuse terms, calling something AI when in fact it is something entirely different, our understanding of what machines can do evolve.
However, do we really need a definition of AI?
What even is intelligence?
Over time, our notion of what artificial intelligence means has shifted. The term was first coined in the 1950s. Its original proponents were mainly focused on a machine’s ability to simulate intelligent behavior. Modern definitions take into account a far more holistic view of what AI might mean, encompassing a computer’s ability to respond to information in its surroundings and autonomously choose to impact those surroundings.
Today, more than ever, we hear about the ‘future of AI’. With the advent of autonomous cars, Amazon’s Alexa or Google’s search engine results, humans are increasingly coming into contact with technologies which appear to be intelligent and we read about various games where machines have managed to beat humans, from chess to Go.
Perhaps part of the problem is that we don’t even have a widely agreed-upon definition of what human intelligence is. Is intelligence something people are born with? Is it learned? Does it exist in and of itself, or does it emerge in our social interactions? Given how hard the task of defining intelligence is, defining the artificial kind is just as (if not more) complex.
The future of AI may be as a support
Perhaps the biggest problem with discussion of the future of AI is that we’re really talking about a huge range of more or less related technologies. When pundits discussing AI talk about personal assistants, search predictions and self-driving vehicles in the same breath, it can give the impression that all these technologies are somehow working in concert, underpinned by the same technologies. In fact, they often represent very different forms of technology, built by different people and which only have a few similarities.
Attempting to sum up all these diverse technologies and putting them under one umbrella is pretty hard, if not impossible. So, rather than focusing too much attention on whether these technologies are intelligent or not, it might be better to focus on how they augment or support human intelligence. For instance, search engines and autonomous vehicles function in completely different ways. What they do have in common, however, is their ability to support human capacities (in these examples: finding information and moving from place to place).
Augmenting workers with topic computing
At harmon.ie, we’re really focused on this more practical definition of artificial intelligence: looking at how we can use advanced technologies to support humans and enhance their work. We developed Collage, a technology that uses machine learning to find patterns in what an employee is working on, to surface content in context, in a way that is useful for them. For instance, Collage is able to scan the contents of an email, identify key topics in the email, and then surface related application notifications and content for that user—what we call topic computing.
Perhaps the key point about our definitions of AI is this: while it can be tempting to get involved in convoluted arguments about what does or doesn’t count as ‘artificial intelligence’, more important is to switch the focus and look at how machines can be used to help enhance humans and help them become more capable, efficient and productive.
As the expectation and reality of our relationship with technology continues to charge forward, we may never come to a point where a universally-agreed upon definition of AI is reached. But what we are confident of is that technology will continue to be designed in a way that supports and enhances humans. And for us, that’s more exciting.