One of the most captivating scenes from the recent Alan Turing biopic, The Imitation Game, sees Benedict Cumberbatch as Turing defend his perspective on the idea of machines being capable of thinking:
Turing was one of the pioneers of computer science; his research presaged the development of the field of artificial intelligence (AI). Since Turing’s time, we have seen great advances in AI. You may not realize it, but there are elements of AI incorporated into many of our favorite devices. In fact, the device you’re reading this on likely has AI in some form built in – perhaps as a virtual personal assistant. The virtual assistant is an exciting development in the field, allowing people to receive help with day-to-day tasks or find useful information in specialized domains like medicine.
The doomsday bunch.
Unfortunately, AI has drawn some rather extreme headlines recently, as the conversation has shifted from the progress that we are seeing to a feared end result – AI-based technology surpassing the levels of intelligence of humans and posing a threat to our existence. Apart from the popularity of such doomsday scenarios in science fiction, this outlook appears unfounded: there is currently no evidence to suggest that anything like this would necessarily happen. Perhaps even more importantly, we’re getting rather ahead of ourselves with these sorts of predictions. AI has indeed seen some encouraging and impressive progress over the last few years, but we still have a long way to go before we achieve anything capable of the scenarios that have been discussed.
AI as transformative technology.
In fact, why focus on such extreme scenarios when there are many alternatives that would see a peaceful co-existence and productive collaboration between humans and machines? These systems could, for example, become partners or teachers, or perhaps even feel indebted to us, their creators.
Consider instead some of the promising futures that AI could enable. AI has the potential to radically transform, in a positive way, the degree to which we can utilize and process data and information in ways that people simply cannot. In addition, simple everyday actions, such as interacting with the Internet of Things, that have become overly complex because of arcane interfaces (e.g., setting a thermostat or controlling a TV) can be radically simplified through natural language. As AI systems mature, they will drive important advancements for society, in areas like healthcare, education, the economy and many more. An AI system could help a doctor with a diagnosis, serve as a virtual teacher with the wealth of Internet knowledge at its fingertips, or be woven into the fabric of our daily lives, helping us with everything from basic decision-making to driving our cars.
Where we are.
Although we are in the early stages of achieving behaviors and intelligence that is in line with humans, recent advances have enabled the virtual assistants mentioned earlier to understand us and interact with us through spoken language. As AI technology improves, these assistants are also demonstrating proactive capabilities acquired through the identification of patterns in our behavior. And researchers – both at Nuance and at other AI-dedicated labs around the world – are in the process of driving not just new advancements, but new ways of assessing progress.
The Turing Test was long held as the benchmark for measuring AI. Since Turing’s time, however, researchers have uncovered a number of shortcomings with that test, stemming from the underlying requirement that the program try to trick a person into thinking that it is human. A good example of this was recently seen in a program that was claimed to have passed the test by mimicking a 13-year old boy. The validity of that claim has been debated by many researchers, but in the meantime, Nuance is exploring more suitable alternatives for assessing progress in AI through its sponsorship of the Winograd Schema Challenge – an exciting new proposal for gauging progress in the field through tests that involve answering multiple-choice questions that require commonsense reasoning.
Other methods are also being examined. The Association for the Advancement of Artificial Intelligence (AAAI) has called for a summit in January to discuss AI challenges and competitions aside from the Turing Test. Dubbed ‘Beyond the Turing Test,’ this event will see leading researchers in the field of AI present new ideas that could serve as more useful tests of progress in AI.
The recent focus on AI suggests an outcome that we will likely continue to debate. More importantly, these conversations speak to the potential of this technology – a potential that we remain committed to developing, understanding, and applying to our daily lives.