Listen to this story
India currently does not have any legislation specific to AI. The closest to this was the draft Personal Data Protection Bill (2019). But that, too, was withdrawn by the government as the Joint Parliamentary Committee proposed a large set of changes and decided to call for a new Bill instead of modifying the existing one.
Analytics India Magazine interacted with Mariagrazia Squicciarini, chief of executive office and director of AI at UNESCO. She suggested a few aspects when drafting AI policies and reforms for India, talking about some of the challenges, solutions, government’s role, and more.
Highlighting one of the instances from a policy perspective for India, she said, “In theory, when you give your data, you will have this principle that in a year, you have the possibility of removing/deleting them (data). But, in practice, it becomes extremely challenging. So, the question for India, and many other countries is, how to then concretely operationalise some of the prescriptions.”
Sign up for your weekly dose of what's up in emerging technology.
She said that India should consider having a harmonised policy that has the possibility to go global in activities, instead of each and every country having different approaches, which actually impinge upon the possibility of compliance by companies, startups, and customers using technology.
“I hope they are considering this now in the development, which I am sure they are,” added Squicciarini, saying that perhaps they will mirror several other constraints in the final wording.
Download our Mobile App
At UNESCO, Squicciarini works around strategizing and advancing the organisation’s work related to social transformation and the reduction of inequalities – human rights, anti racism and the gender divide – the ethics of science and technology, and in particular, of AI and neurotechnology, among others.
Before this, she worked at the OECD Directorate for Science Technology and Innovation for over a decade as senior economist – head of unit. Her work revolved around digital transformation, artificial intelligence and others.
Squicciarini defines AI ethics as ‘putting technologists at the service of people, and not people being just used by technologists,’ and ensuring people are at the centre of it, based on human rights.
India vs the world
Last year, UNESCO announced that all of its 193 member states had adopted the global framework agreement on the ethics of artificial intelligence, emphasising AI technology benefiting humanity and raising fundamental ethical concerns. Recently, the US, not a member of UNESCO, also unveiled the AI Bill of Rights, which aims to prevent the harm caused by the rise of AI systems.
Slowly but surely, many countries are now waking up to the new reality developed by artificial intelligence and other emerging technologies before it’s too late to act. As per the AI Index report, the UK and the USA topped the list, following the pattern of the proportion of AI mentions in legislation that became laws. With 939 mentions, the UK comes first, followed by Spain, Japan, the US, and Australia.
On the contrary, the number of mentions of AI in Indian legislative proceedings, between 2016-2021 stood at 34. Of 24 countries mentioned in that list, India stands sixth from the last position.
(Source: Bloomberg Government)
Challenges of drafting an AI policy
Squicciarini said that today many companies that implement AI call themselves differently. Most companies today say they leverage AI and machine learning, but things like convolutional neural networks aren’t often mentioned. According to her, this remains a challenge on how we can translate this into policies.
She said once this is done, how do you take principles to practice? “Because the challenge AI has is that if you look at the recommendations, it touches so many parts of governments. We know that governments often work in silos, simply because it is very complicated to bring everybody on board,” said Squicciarini. She said, “Let’s face it, coordinating is not easy. There are a lot of coordination costs.”
Citing healthcare, she said the impact is seen from health to employment to privacy. “Of course, what we are talking about already by the example – I put in three ministries,” she added, saying that they need to regulate and legislate in the same direction. They will need to enforce this in a way that ensures coherency. “So, these are the challenges we see,” said Squicciarini.
What’s the solution?
She said that there is a lot of ferment around the world and many initiatives. This is both good and bad news. “We shouldn’t reinvent the wheel; we should check with what works and share the good practices,” she added.
Squicciarini said that countries should work together, those belonging to the same region because they share some commonalities. This can be leveraged so that one can test or try to experiment in one direction. Learning and sharing mutually help cut down cost, save time, and improve the efficiency of the whole process.
Citing Germany and Canada, alongside Namibia and other South African countries, Squicciarini said that these countries have seen several efforts and are sticking together to advance on implementation. “So there are a lot of good practices emerging,” she added.
Squicciarini’s version of AI policy aligns with finding a global solution to a global challenge. “In the sense that AI doesn’t know any boundaries unless you put them, you can’t stop it,” she added, saying that it doesn’t make any sense to have individual, completely different types of regulations or strategies.
Further, she said that they see the solution being generational as more and more young people are demanding the solutions, and the older generations who are sitting at the top of the establishment can not ignore the problem anymore. “So, we need to listen to more youth. We need to involve them more in a number of initiatives. Ethics of AI being one of them,” she suggested.
“Let’s face it, I am here, speaking for myself, only for a few more years. But, the people that today, say, are 15 and 18 years old are the ones that will have at least 50 years ahead of them. If we are after long-term solutions and long-term commitments, they will be the ones that will need to implement them, so they need to be involved in this discussion,” opined Squicciarini on a strong note.
Tackling data and algorithmic bias
While coming up with global AI policy sounds fair, is there a framework or a system to tackle biases in data and models and their consequences? “So your question is very important because actually, UNESCO can’t go and check each and every algorithm,” said Squicciarini.
However, they are working with researchers to develop an algorithm that checks on other algorithms for biases, gender representativeness, identification, discrimination in terms of the outcome, etc.
Squicciarini said that the biases come from the person coding them and the data being used to train the model. She said that there are a lot of ways in which codes can be adjusted and checked. “While UNESCO doesn’t have it, we do really need an army of people to be able to do that at the scale of the world,” she added. However, she suggested that at this point, it relies on the company self-assessing, and there needs to be accountability in eliminating data and algorithmic biases, which requires education and awareness work.