Position Statement on Generative AI
From the R&D Committee
We are glad to announce that the Indian Society of Artificial Intelligence and Law is hosting the 3rd virtual edition of the Indian Conference on Artificial Intelligence and Law, 2023 on November 22-25, 2023.
You may find the itinerary of the programme at isail.in/indocon.
To attend the event, you are required to become a member of ISAIL or a paid subscriber at indian.substack.com.
You can also express your interest to attend at executive@isail.co.in.
We are also glad to release a Position Statement on the state of Generative AI authored by the members of the AIGA R&D Committee of ISAIL, led by Pavithra Manivannan (Chairperson), Tushar Sharma, Kailash Chauhan, Indrasen Poola, Rishabh Bezbaruah, Samridhi Jain and Hari Shivan.
The Position Statement on Generative AI
The development and use of GenAI models are rapidly increasing. They are changing the way we communicate, work, think and conduct business. However, even developed nations with better infrastructure are grappling to understand its full potential and the challenges that come along with it. In India, the GenAI industry is growing at 48% CAGR. Given this, it is essential to understand and appreciate its current state in India, to be able fully harness its use and take preemptive actions to govern its functioning. This position statement is offered with this intent.
Statement 1: Democratisation of AI
The benefits and efficiency impact that GenAI can bring in is indisputable. However, it is essential that this benefit is reaped by more sections of the population, both in terms of its development and use.
- Infrastructure challenges
To innovate high-end Gen AI tools and train them with vast data pools businesses require large hardware infrastructure, which are expensive. This can cause emerging startups and MSMEs to take a foot back. Lack of access to high performance computing and general disinterest by investors to fund for hardware infrastructure hampers the development of such technology by young and small businesses.
- Labour displacement
Use of GenAI has the ability to disrupt labour both skilled and creative. It is difficult for human workforce to keep up with the growing pace of technology. However, at this stage, most GenAI tools are likely to be human-aided. Therefore, it is essential that government, businesses as well as educational institutions invest in upskilling labour with valuable digital skills so as to enable them to boost their productivity.
Statement 2: Regulation
Many nations such as China, Brazil, EU, Japan, are in the process of adopting regulations for governing AI. The difference in each jurisdiction varies from pro-innovation soft touch models to outright bans.
- Problems of invisible infrastructure
It is tempting to borrow and adopt one or more principles from other nations. But policies that work in one country may not be successful in others owing to invisible infrastructure, that is, the super set of laws, institutions and state capacity in a country that is crucial for ensuring the success of a policy intervention. A governance framework for India should be developed and implemented keeping in mind the existing laws, institutional capabilities, level of freedom to businesses and socio-economic conditions of the demography, of the Indian State.
- Adaptation of existing laws
Copyright and data privacy are top concerns surrounding the use of GenAI in the global ecosystem. To the extent possible, these could be solved for by amending existing laws as opposed to creating new laws. There is precedent for doing this, for instance, in the adoption of digital evidence. On the government’s front, this will reduce the cost of setting up new regulators. On the other hand, for businesses, it will reduce the cost of compliance and enhance ease of doing business in India.
Statement 3: Misinformation and ethical standards
Organisations developing and using GenAI tools and models must have a strong emphasis on curbing misinformation and polarisation. Content generated through GenAI must clearly be marked and disseminated as such. This not only enables autonomy for users but also ensures that the model is being trained on appropriate content. Secondly, since the model learns from its users, it is susceptible to furthering the disparities in the society. Educating and increasing awareness of the population about the mode of functioning of GenAI models is crucial. Both businesses and the state has a role to play in this.
Statement 4: One size does not fit all
When it comes to regulation or public discourse about the subject, it is essential to understand the nuances of the models and tools. Not all AI tools use personal data. Some are trained based on public source of information which essentially perform more sophisticated functions of search engines. Further, the industry or sector that a tool caters to is also a key determinant factor. In some, the benefits would far out-weigh the potential harm or hype or harm.