Computers With Common Sense – A Close Possibility

May 22, 2015 - Written By Soumabho Dutta

Will we ever witness a computer with human-like intelligence? Yes, says Professor Geoff Hinton, a leading AI scientist. According to him, computers can develop ‘common sense’ in the next decade or so and from then on, no dream will be too big.

Around two years back, Professor Hinton was hired by Google to develop intelligent operating systems. During a recent interview with The Guardian, the talented researcher said that Google is on the verge of developing an algorithm with the capacity for logic, striking conversations and even flirtation! He pointed out that Google’s new algorithm encodes thoughts in a sequence of numbers, which they refer to as ‘thought vectors’. However, he also stressed on the fact that the work is still in its nascent stage and that Google is still a long way from having a sophisticated version of the software that will eventually allow the computer to think and sound like human beings, with the capacity of logic and reasoning.

On being quizzed about it even further, Professor Hinton pointed out that thoughts can be captured and distilled to a mere sequence of numbers. While he agreed that many in the world will think of all this as ridiculous, yet, he stood by his vision and explained, that there is no reason why thought can’t be captured in vector form. He further harped on the fact that the ‘thought vector’ approach will help solve two critical challenges in the world of Artificial Intelligence – to help master natural, conversational language and give the AI the power of logic. According to him, in the near future, people will not only use their computer to extract information, but will also engage in conversation – reminiscent in the film Her, wherein Joaquin Phoenix falls in love with his self-made AI. According to him, none of this is far-fetched and there is every chance that human beings will eventually befriend their computers.

Richard Socher, an AI scientist at Stanford University recently came up with a program which he named NaSent. It has been trained with 12,000 sentences from film review site Rotten Tomatoes, to better understand and recognise human sentiment. Hinton too added that in the last two years, scientists around the globe have made significant progress in the field of AI. He played down concerns about the dangers of AI by saying that he’s not afraid of an autonomous killer robot and is instead scared to think that the technology might allow NSA to misuse computers even further.