By S V Krishna Chaitanya | Express News Service | Published: 09th January 2018 02:27 AM |
CHENNAI: It reads like a scene from a classic sci-fi flick. But it’s not. It is really a human talking to a machine - a machine being built by 14-year-old Indian-origin whiz kid Tanmay Bakshi, who is currently working with IBM on artificial intelligence (AI). The AI machine that he envisages will have brain of its own and play the role of a human therapist diagnosing and counselling people with mental health illness, especially the kids and teenagers.
And there may be good reason it sounds like a movie. Part of the trick is that this machine is learning to converse by analysing an enormous collection of old movie dialogues. Bakshi is using Cornell Movie-Dialogs Corpus to train the neural networks of the machine. This corpus contains a large metadata-rich collection of fictional conversations extracted from raw movie scripts numbering 2,20,579 conversational exchanges between 10,292 pairs of movie characters from 617 movies.
Sharing details about his path-breaking venture ‘E-Therapy’, Bakshi, who on Monday delivered a keynote address before jam-packed audience on artificial intelligence and machine learning organised by E-Cell of IIT Madras, said depression-driven suicides were on the rise and in almost 80 per cent of the cases, teens who commit suicide give out clear patterns. “In Australia alone, 40 per cent of calls to the helplines go unanswered because of shortage of manpower. This is where artificial intelligence comes in handy. Eight months down the lane, we will be ready with a prototype (proof of concept) that will engage in a conversation with a human and diagnose at least a few types of depression,” he told Express on the sidelines of the event.
Bakshi said the reason why movie transcripts are being used to train neutral networks was if the AI machine can learn and make sense from such huge volume of data, it would be easy to make the machine think and analyse a limited scope like mental illness.
A mobile app is being designed for public to volunteer and share their data. “To design a prototype, at least 2,000 hours of conversation are needed. AI will allow a free flow of conversation between AI therapist and the patient. We are not here trying to replace professional therapists. We are only trying to mimic them to bridge the gap so that a teenager who needs help get it on time just over his/her phone.’’
