Future of Artificial intelligence - discussion on opportunities and threats. From predictions to controlling life

Dr Patrick Dixon, YouTube Futurist Keynote Speaker - Recent Futurist YouTube Videos - Keynote Speaker

Discussion / presentation by Patrick Dixon to a charitable Foundation. Infosys has invested hugely in predictive analytics, for example in the oil and gas sector. Engineers planted around 30 sensors on several oil rigs in the Gulf of Mexico, and connected their data outputs to AI. The sensors were recording things like temperature, air pressure, low frequency vibrations, high pitched sounds and so on, and matching the data to other events such as mechanical breakdowns. After a few weeks, the team had an alert from the system, warning that there was going to be a problem soon on one of the rigs, which was not taken very seriously because the system had only just been set up. A few days later there was a major incident, and the engineers realised that AI discovered how to prevent disasters. But how? AI had detected unusual and obscure data patterns beyond any human understanding. AI is better than doctors at diagnosis from images The same has happened with doctors reading X rays of bones or 3D scans of the human brain. AI was fed with tens of thousands of images, matched with diagnosis made at the time by experts. It was also fed older images of the same person. So for example, a brain tumour may have been obvious in the later images but invisible to the human eye in earlier ones. And as a result, in many hospitals around the world, AI is now detecting cancers far earlier than any human being can manage. Indeed, AI is better at image analysis than the human beings that trained AI. Not just for medical images, but any images: face recognition, recognition of people by the way they walk, and so on. AI predicts human behaviour - tennis champions How about predicting human behavior? Of course, autonomous cars can only drive safely by anticipating what other drivers may decide to do. And AI in marketing is used extensively to map out what kind of ads to present us with online, based on lifestyle choices, age and so on. But we can human predictions much further. Infosys has also invested a lot of effort in predicting the play of professional tennis players. To do this, they fed video of every major tennis tournament into an AI system, going back 30 years, plus all the associated data such as serve speed, scores and so on. The Infosys team were astonished at what happened next. They started testing the AI tennis system with matches the system had never seen before, with players whose previous matches were in the system. The questions was this: can AI predict precisely what shot a particular world-class player is going to play next, at any point in any game? The answer was yes, in many cases. So much so that Infosys started taking groups of VIP clients to Wimbledon Tennis Championships to demonstrate the technology. Infosys also started talking to coaches of the best players in the world, offering fresh insights. Turns out that even the most successful tennis players are far more predictable than they imagine, in the shots they play or don’t play, in the impact on their playing of their own mood, physical tiredness and so on. And all of this was being picked up and inwardly digested by AI. AI in detecting fraud and commit crimes AI is being used in similar ways to detect fraud. AI collects many tens of thousands of data points about each credit card user, to predict what kind of purchases they are most likely to make next. So the moment an “unusual” purchase is detected, the card is blocked. Data points can go far beyond normal size of payment or location of the person. It can include the IP of the device being used, fingerprinting, browser data, keyboard pressure, the angle the smartphone is being held when a PIN number is being entered, the interval between each of the numbers being entered, intonation of voice and so on. On the other hand, AI can also be used to trick such systems for criminal intent. For example, faking someone’s voice, faking a video of someone talking (both image and their voice), faking IP address. AI will have power to lock people out of their own lives The trouble is that now we already live in a world where AI can and does automatically block us from part of our day to day lives, in an instant, in ways that can be hard to challenge. In future you could find that your entire day to day life has been effectively “cancelled”. It happened to me the other day. I was travelling in Asia and one of my credit cards was blocked – because I had been through so many different nations in the previous week, none of them places I usually visit. AI from one bank alerted AI in other banks so I soon found that all my other cards were also blocked. In that moment I had effectively lost the ability to travel, buy food, pay for a taxi, pay for a hotel room. AI robots are also monitoring social media. It may not matter to you if a particular social media platform is blocked for you for a few days but for many it could be hugely damaging.

https://www.youtube.com/watch?v=E6len92r7SU

< Prev		Next >