AI: progress or peril?
Artificial Intelligence (AI) is the tech buzzword of the moment, with a vast array of products and services claiming to be powered by AI. But what does AI really mean, and is it all that it’s hyped up to be?
In the popular consciousness, AI usually means machines and robots that can think, like Skynet in the Terminator or Alicia Vikander as the beautiful robot, Ava, in Ex Machina. This type of AI, called general AI, can learn, reason and adapt, but doesn’t yet exist.
Today, the only kind of AI that does exist is narrow AI. Narrow AI can only do what it has specifically been taught or learned to do. Despite this limitation, today’s AI can process data faster than ever before, help solve previously unsolvable problems and power a vast array of new capabilities, from Netflix recommendations and language translation to facial recognition and self-driving cars.
Machine learning powers most of these innovations, and is a subset of AI that edges closer to the long-imagined version of AI. Machine learning can learn and adjust itself when exposed to more data. Netflix reports that its machine learning-based suggestions are so accurate that more than 80 percent of the TV shows people watch are chosen through its recommendation system. Image and speech recognition both use machine learning, and machine learning has meant enormous improvements in both fields. Through machine learning, AI can now detect cancer in tissue slides better than human epidemiologists.
Deep learning is the next step further and a sub-category of machine learning. Deep learning is modelled on how the human brain draws conclusions, and like a brain’s neural networks, uses algorithms that are layered “deep” to learn from large amounts of data, with greater accuracy and more intensive compute. Autonomous vehicles use deep learning to recognise street signs and even pedestrians, utilising millions of individual deep-learning models to do so.
AI is accomplishing a great many things, but it still has a long way to go before anything approaching the synths (robot servants) portrayed in the TV series, Humans, become a reality. Experts have differing opinions about when general AI will happen. Author of “Architects of Intelligence” and futurist Martin Ford interviewed 23 of the preeminent people in the field of AI today, and surveyed them as to when they estimate general AI will exist. Opinions ranged from 10 years to 181 years, with the average being 81 years from now.
Although willing to venture a guess, most experts would agree that it’s nearly impossible to accurately predict what will happen, and when, with AI because it’s unknowable now what developments and breakthroughs are on the horizon. It’d be like asking experts from 1804, when trains travelled 5 mph, to predict when humans would be able to fly.
Stephen Hawking famously observed that the rise of artificial intelligence will either be the best thing that’s ever happened to us, or it will be the worst. But it’s important to remember that humans have always been fearful of significant technological inventions. Even the telephone – and the idea of copper wires carrying human voices across the sea – was a cause for concern at the time. While we can’t ignore the risks of AI, we also can’t afford to ignore its rewards. Only time will tell if our existential fears are real, or just science fiction.