This article was first published in the September 2015 issue of WIRED magazine. Be the first to read WIRED's articles in print before they're posted online, and get your hands on loads of additional content by subscribing online
Co-founder and lead scientist, Emotient "We underestimate how hard vision is because our brains do it so well. The human brain has evolved for millions of years to be able to make sense of light patterns. Until ten years ago, getting a computer to recognise even simple objects in natural images has been elusive. Now, with advances in machine learning -- many of which are inspired by our understanding of the brain -- computers can recognise something as nuanced as facial expressions. This will transform the way we interact with machines. In the next 20 years it will be commonplace for devices to respond to non-verbal signals such as attention and mood."
Charlotte Golunski
Co-founder, Sense "Research is underway to allow computers to be able to recognise what is being said, understand what sounds mean and recognise objects in a scene. Once we achieve this, there is the challenge of 'data fusion'. We understand the world around us by fusing information from different sources simultaneously. This helps us make faster, more accurate decisions, but is still very difficult for computers. When this challenge is solved, AI will have more use cases than we can imagine today."
John Gordon
Vice-president, IBM Watson Solutions "All innovations start with a moment when a person sees something previously overlooked and recognises its potential to change the way we live. Today, new cognitive computing systems put science behind serendipity to help us see the world differently. From inspiring chefs to find new flavour combinations to helping pharmaceutical researchers spot new treatment pathways, cognitive systems will help us accelerate our work and inspire our thinking, ushering in the next era of innovation."
Matthias Scheutz
Department of computer science, Tufts University "Data-driven statistical machine learning will allow us to surpass human performance in many areas, from solving the hardest crossword in the blink of an eye to proposing architectural designs for cities. For autonomous systems such as service robots, it will be critical to attain sufficient moral competence in order for them to navigate the complexities and intricacies of human social and moral norms without causing us unnecessary harm."
Oren Etzioni
CEO, Allen Institute for Artificial Intelligence "What if a cure for an intractable cancer is hidden within the tedious reports on thousands of clinical studies? In 20 years' time, AI will be able to read -- and more importantly, understand -- scientific text. These AI readers will be able to connect the dots between disparate studies to identify novel hypotheses and suggest experiments which would otherwise be missed. AI-based discovery engines will help to find the answers to science's thorniest problems and ultimately revolutionise science."
Richard Socher
Founder and CTO, MetaMind "One challenge is understanding and retaining connections between facts and concepts. If a machine could achieve this, it could become a domain expert, answer tough questions and make difficult inferences autonomously. Another major practical challenge is efficiency. Today's AI approaches require tremendous amounts of computing power. Developing more efficient and effective algorithms will be key to making the technology ubiquitous and valuable."
This article was originally published by WIRED UK
