‘Regulations in artificial intelligence have not caught up with speed of development … make sure we don’t lose core humanity’ by Sethu Vijayakumar

‘Regulations in artificial intelligence have not caught up with speed of development … make sure we don’t lose core humanity’ by Sethu Vijayakumar

Topic: ‘Regulations in artificial intelligence have not caught up with speed of development … make sure we don’t lose core humanity’

Writter:  Sethu Vijayakumar

Publish Date: Wednesday, 20 December 2017

Published on : The Times of India

‘Regulations in artificial intelligence have not caught up with speed of development … make sure we don’t lose core humanity’

What is the role being played by robots today?

Robots have been playing a role in traditional manufacturing sector. What’s changing now is that robots are becoming more ubiquitous in everyday life. They are playing a role in things that are much closer to home like surgery on the human body.

What’s your role in NASA’s ‘Valkyrie’ project?

The project is a small part of NASA’s bigger vision of making us a two-planetary species. In other words, potentially inhabiting the Mars surface in a long-term perspective. That is where we come in: robots going in and doing things human workers would do. Our role is to build algorithms and control systems that can really take these complex humanoid robots and give them the dexterity, mobility, biped locomotion capabilities, sensing and reaction capabilities that humans take for granted.

Is artificial intelligence (AI) the final frontier in robotics?

If you think of AI, it is a bigger concept which embodies everything from taking sensory data you’re getting, processing it in a clever way and creating additional value in order to do things in an interesting cognitive fashion. That’s a broad way of thinking about AI. Within that there are subsets that deal with pure data streams and that is what social network media is doing. But the point is it’s not changing the state of the world, it’s just data. Robotics is considered the arms and legs of internet of things. In other words, if you want to take the processed data and change the state of the world then you need robotics and activation systems. That’s where robotics fits within the AI hierarchy.

This requires human programming. Can a machine ever mimic the human ability to learn?

A significant part of the research I’m doing looks at learning systems. The systems learn and adapt. The bigger aim of that approach is to equip a robotic system or an activation system with some core capabilities and then use the sensor data it receives to constantly evolve its behaviour. The core initial behaviours are built in by the human programmers but the systems are able to evolve and adapt based on the feedback it gets and changing environment.

That’s not yet reality.

There are significant aspects of it that are already reality. In terms of autonomous driving cars, we have trials in the UK about systems collecting data, and then driving. There are special events which federal agencies have regulated. These are intervention events when a person has to take control when something is going to go wrong. In those scenarios, it is not just that machine which learns. It feeds that into a database which networked machines around the world can learn from. Eventually, if you have seen one kind of accident in autonomous driving cars, it’s unlikely that the same sequence of events will be repeated in any of the cars.

Is this human capability without human weakness?

That’s true. Random stochastic nature of human failures will be taken away. At the same time adaptation will evolve it towards a better system. But there is a downside to it. Typically, a programmer has the ability to define a goodness criterion. The question is who will control the goodness function in there and does it capture the subtleties of ethics that we take into account in our everyday life.

Stephen Hawking has a sense of foreboding about AI. How would you view it?

My belief is that as with any advanced technologies there is a risk for it being used in a way that is not originally planned for. You have to have a certain level of ethics and morals underlying the developments you are building into it. It’s the responsibility of scientists to potentially understand the applications of the work and at the same time cooperate fully with agencies creating ethical guidelines. Regulations in robotics and AI have not caught up with speed of development. We have to get people who understand the law and folks who deliver technology together because they cannot work in isolation.

Do we have to become cyborgs to deal with this world?

As machines become more capable, there is this notion we would have to augment ourselves with artificial devices to catch up. It’s an interesting thought. Just take a calculator. It’s immensely more powerful at computing vast numbers than your mind. If you start extrapolating that capability to various fields of life, you may get to a stage where a machine is a lot more capable in many, many areas than an average human. Does that mean we are threatened? Who knows? It may mean we keep some safeguard in place to make sure we don’t lose what we think is core humanity.