Tamil Nadu at work on safe & ethical AI policy
Tamil Nadu will unveil a policy for safe and ethical use of artificial intelligence (AI), in what is arguably a first-of-its-kind move in the country.
“While there is a lot of talk around AI, to condense it in the form of policy has not been done before, to be best of my knowledge," Santhosh Mishra, Chief Executive Officer of Tamil Nadu e-Governance agency, said.
The agency executes cost-effective governance projects using frontier technologies such machine learning, drones, augmented reality, among others.
The government will also bring out a rule book for state agencies and vendors who use AI for governance services, to help handle private individual data in a sensitive manner.
The policy will be released along with the blockchain and electronics hardware policy in about three weeks, according to another top government official.
“Government departments will begin to use AI progressively. We have created a matrix which will indicate if the AI is used in a safe and ethical way. Government departments, when they do a tender, need to ask themselves these questions. I should not be doing AI on your data; things like that and more,” the official said.
Ethical implementation of AI goes beyond data privacy. Machine learning algorithms, an instance of AI, give different results for different data, meaning the data fed into such software need to be free of all kinds of identity biases.
The question of ethics in AI has two aspects — explicability and constitutional validity, said Anindya Chaudhuri, senior economist with the Global Development Network, a Delhi-headquartered network of policy institutes.
“One of the central problems with AI algorithms is that they suffer from what's known as ‘Explainability Problem.’ Essentially, there's a black box between the input and the output," Chaudhuri said, referring to the decision-making processes in AI being too difficult for many practitioners to understand.
"The other is that the use of AI requires or ends up in situations that violate other democratic constitutional norms — privacy, equality before the law, presumption of innocence until proven guilty."
Chaudhuri had given a lecture at the e-governance agency in September on the challenges faced by agencies that implement AI for people.
The e-governance agency has held consultations with private software firms, government departments, and citizen groups to hammer out the policy.
The agency had earlier partnered IIT Madras’ Robert Bosch Centre for Data Science and Artificial Intelligence with impact areas such as education, health and agriculture.