New frontiers: I think, therefore I am – a machine
Dr. Sadeep Jayasumana
Amachine that could think for itself, something like Arnie’s Terminator, and able to take decisions on its own without any human input – this is what the future might look like according to Dr. Sadeep Jayasumana, an Oxford -based researcher who dabbles in the field of artificial intelligence and deep learning.
Intrigued by reports of a recent talk given at WSO2 in Colombo on the somewhat esoteric subject of ‘Deep Learning in Computer Vision and Algorithmic Trading’, we buttonholed Dr. Jayasumana in a bid to uncover what on earth it exactly means .
After an hour of exactly conversation we emerge worried for according to the Kalutara-born whiz-kid – in 2010 he graduated from Moratuwa University top of a class of 750 engineering students – it is entirely a plausible scenario where in the future the world will be filled with robots and machines, all turbo-charged with artificial intelligence and the ability to do their ‘own’ thing.
“Maybe there could be a ‘Terminator’ doomsday scenario one day if we are not careful. Researchers are always trying to push the boundaries (of invention) and it is a challenge to build something with AI (artificial intelligence) which works like Bots with the ability to think for itself,” is Dr. Jayasumana’s frank verdict.
“If there are no restrictions from governments, perhaps in 50 years it can get to an uncontrollable stage, but I think before that people will start intervening like they did in the gene industry where you can’t just clone a person because of regulatory and ethical reasons. So at some point these same obstacles will start to kick in and safeguards will be built in. Hopefully,” adds Dr. Jayasumana.
Ever since watching Stanley Kubrick’s epic science-fiction film (2001) based on the Arthur C Clarke classic ‘A Space Odyssey’, where Hal, the space ship’s computer, takes control, there has been an atavistic fear for machines superseding mankind. The Terminator did nothing to quell those fears.
Dangers of AI
But never mind the ideas of science fiction writers and of Hollywood for more recently, none other than pre-eminent British scientist Stephen Hawking warned that Artificial Intelligence could herald the end of mankind.
In a 2014 interview with the BBC, Hawking said efforts to create thinking machines pose a threat to our existence. “The development of full artificial intelligence could spell the end of the human race,” cautioned Hawking. “It (machine) could take off on its own, and re-design itself at an ever increasing rate”.
So what is Deep Learning in Computer Vision all about we ask Dr. Jaysumana. “It is the science of making a computer see the world. When we see an image we understand it but for a computer when you show an image it is just a bunch of numbers. To make the computer understand we use machine learning. In simplistic terms if we show an image of a car and tell the computer it is a car and do it over and over again showing lots of images of cars, this is what is called machine learning.
“Through repetition you teach a computer, machine learning. We have to input. It is like a new-born child who doesn’t know anything at first but you teach a child showing images. It is pretty much the same thing, you build an artificial brain inside a computer but when you first build it the computer doesn’t know anything and then you start showing examples,” explains the 29-year-old Sri Lankan who was on a short break home last week.
This technology is nothing new. It has been happening in the US and the West for decades and has reached a point where today we get automated cars – Tesla and its self-drive cars for example – and much more.
“There are many applications being used today which use deep learning. For instance in the medical field it is being used in MRI scans, Siri for voice recognition on smart phones, Amazon for its retailing purposes or Google for its various tasks,” reveals Dr. Jayasumana.
Having finished his Bachelor’s degree at Moratuwa University, Dr. Jayasumana won a scholarship to do his PhD at the Australian National University. He chose computer vision and machine learning. He had always been fascinated by machines, even as a schoolboy at Kalutara Maha Vidyalaya, when he used to toy with programming and coming up with text-based games.
Deep learning
After completing his PhD, in 2013, Dr Jayasumana continued working on deep learning allied to computer vision at Oxford University as a post-doctoral researcher. He is still based there although he is now putting all his hard work to practical use working for a hedge fund based in London which manages over US$19 billion in funds.
“I’m a Quant – a quantitative analyst in machine learning. Although I’m still based at Oxford, I work for a finance company.”
He is involved in algorithmic trading – whether to buy a stock or not, with the computer analyzing and deciding. The algorithm can predict with some accuracy, but the stock market can go up as well as down. In the long-term, and on average, it pays off.
“We analyse the stock market using techniques like maths, machine learning, physics, computer science and then try to build an algorithm which can automatically trade in the stock market. We are trying to mimic what goes on inside a human brain but in a computer. It acts like a human brain. If you ask if a computer can make a decision the answer is yes in cases like self-driving today.
“The grand goal in AI is to build a machine which can think like a human. But to build that you first need to build a sensory vision and that’s where computer vision comes into play. You can’t build an artificial human without building vision first.”
Machines learn fast. Dr Jayasumana related a story of how a few years ago a machine was taught to play a game. It lost at the start but after more training, did better although still losing. More training led to the machine discovering on its own steam how to get an edge and win the game.
“The neural network learned this trick on its own. This is frightening, actually. There are some movements in the West where people are saying we need to be more careful. We can’t build safeguards into the system at the moment. Right now neural networks are not sophisticated but I think it will be in 10 or 20 years,” Dr. Jayasumana predicted.
So what does this mean for mankind? In the 17th Century, French philosopher Rene Descartes wrote a Latin philosophical proposition ‘Cogito ergo sum’ which translated means ‘I Think, Therefore I Am’. In the future this might not only hold good for humans.