 I'm 23 and I live in Kyoto. But I think the development of full artificial intelligence could spell the end of the human race. Once humans develop artificial intelligence, it will take off on its own. Should we be fearful of artificial intelligence and the pace at which it's progressing? Or should we fear fear itself and the risk of its stifling innovation? It will not be man or machine. Our purpose is to augment and really be in service. According to a recent report published by researchers from the universities of Cambridge and Oxford, Open AI and the Electronic Frontier Foundation, there's an assortment of ways AI could go wrong and in the not so distant future. Focusing on what could happen in the next, say, five to ten years, the paper draws attention to what may seem like more prosaic concerns than, for instance, a global takeover by humanoid robots with superintelligence. But whether advances in AI lead to a rise in malicious bots or the weaponizing of consumer drones, just two of the risks listed in the report, we can't help but fear what we don't understand. So, to begin, the experts weigh in. Is artificial intelligence actually intelligent? I'm Kai Fu Li. I run Sinovation Ventures. It's a high-tech venture capital firm in China. My name is Fei Fei Li. I'm a professor and director of the Stanford Artificial Intelligence Lab. I'm also currently at Google on my sabbatical from Stanford. I am Jennifer Juscott, co-founder and principal of Radium Partner and Radium Blockchain Venture. A lot of people ask about AI and weather intelligence. I think today's AI, based on the current state of the arts, artificial intelligence is really more of an optimizer. It's not intelligent in the same sense that we think of humans as intelligent. We acquire knowledge, we abstract, we create. There's emotion involved and there is judgment involved and there is creativity involved. Prosto may knowledge common sense and ability to argue, the ability to understand humor and also self-awareness and compassion. Today's machine intelligence is far from that level of complexity. It's probably not most useful if we generalize AI. AI is an umbrella term from machine learning to deep learning to NLP, which is natural language processing. To reinforcement learning, there are many different technologies coming into this umbrella. And right now we're at the stage where AI is very good at doing one single task, well, concentrating the task and repeating the task really, really well. The consensus among these experts is that AI is far from reaching human general intelligence. Currently, AI is a collection of techniques which cleverly uses maths to create the appearance of intelligence. That, along with the appearance of emotional intelligence, shows that we need to build an ethical compass into these systems or, at the least, developers have to use one in order to avoid unanticipated outcomes. But wherever this may be heading, the march of progress shows few signs of slowing down. And who's at the forefront of these advances? Kaifu Li names the internet giants holding the largest amounts of user-generated data, companies such as Google, Facebook, Amazon, and Microsoft, as well as Baidu, Tencent, and Alibaba. It's important to note that of these seven companies, four are American and three Chinese, a near-even split revealing just how fiercely competitive the field of AI has become. Over the summer, the Chinese government pledged to become the world's primary AI innovation centre by 2030. Though the US still leads in terms of private investment, experts suggest that several key factors may enable China to push ahead. I think it's quite interesting to see the duopoly between US and China. China is very interesting. It's my home country and I see you with lots of energy where things are in China. If we look at the top 100 people in AI, US probably has 80, UK and Canada probably have 15. China might have two. On the other hand, China has four things going for it. Number one is that China has a huge, very strong engineering education and a mass of people with fantastic engineering training who are already able and willing to jump into AI. Its economy is very entrepreneurial right now, so there's an embracement of new technology and new ideas and that is a very fertile playground for new technology. The third reason is the vast amount of data. Data is the fuel for AI and China has more data than anybody else. Three times more data because of three times more mobile users compared to the United States. It's actually 10 times more data in terms of ordering takeout, 50 times more data in terms of mobile payments. The fourth would be that there's a lot of support both top-down and bottom-up in the society for AI technology and education. The government will put billions of dollars, not at the country level but at the first city level to help subsidize AI companies. A land takeoff and also bring back overseas talent. I think the two countries no doubt have become AI superpower but they are two very different beasts. In US we have companies like Boston Dynamics, DeepMind now is effectively an American company and you have companies like Handsome Robotics, etc. What they are producing is very inspiring, it's very futuristic. Some might argue it's slightly scary, especially with Boston Dynamics, when you see that robotics can do all this. At the same time, most of the people, including myself, we haven't really seen their products being used by everyone. And the difference in China is the engineers and developers in China are very practical. If what they are developing cannot be commercialized in the next year or two, they will not develop it. And in the US you see very advanced research being very well funded, that's probably going to be playing a dominant role in the next 10, 15, 30 years. One such American endeavor showing promise is being led by Rita Singh and her team of researchers at Carnegie Mellon. They're building AI-powered technology that analyzes the human voice. Why? Their voices carry a lot of valuable information. Indeed, so much information that Singh can construct a three-dimensional image of a person's face by studying the voice alone. Your voice carries a lot of information about you, an enormous amount of information. Everything about your physical stature, your facial structure, your ethnicity, what's going on in your body, your state of health, state of mental health, physical health, it captures everything that influences your body and mind and is evident in your surroundings. This information is embedded at very, very fine levels, micro levels in the speech signal. This technology, says Singh, could have many applications in medicine and is already being employed by law enforcement to help identify criminals, not only would it likely reshape the future, but also perhaps the past. I was curious to know what made Hitler so powerful. Was it because he said the right things? He was at the right place at the right time. I think the speculation is that it might have been something in his voice. And so I picked up Hitler's voice and the first thing I discovered that there were signs of Parkinson's. He would hold his fists clenched like this. That was to hide a small tremor that he had developed. And from that kind of evidence, people have speculated that he might have had Parkinson's. So what's the likelihood that he had Parkinson's? I think 100% from his voice. I've seen it in his voice. It's clear. How does the AI behind voice profiling fit into the big picture? With its incredible power and potential for misuse, this technology reveals the urgency in establishing best practices across the field. As the use of AI in criminal justice is just one of the ways in which injustice has been recognized from the use of biased data sets. But returning to the original question of whether or not AI is on track towards achieving superintelligence, that remains to be seen. AI is such a young field. It's only about 60 years old from ground zero to where we are today. From that point of view, we have come a long way. I think it's better to think about artificial intelligence as a single domain optimizer. And that in itself has huge implications, but it is not human or encompassing intelligence, which we call artificial general intelligence. There's various disagreements and debates on whether artificial general intelligence is ever possible and how long it will take. I'm personally a pessimist on achieving that level of intelligence, and I am an optimist on even today's single domain narrow intelligence creating huge societal impact. Artificial intelligence is about learning and making the right decisions using what you've learned. We are at a point where we have the understanding now to do this, and therefore we are in the new AI era.