Nasscom issues guidelines for research, development, use of generative AI
Considering the recent powerful developments in the generative artificial intelligence space, tech industry body, and trade association and advocacy group Nasscom has released a set of guidelines for those engaged in research, developing, and using generative AI (GenAI) technologies.
In terms of objects and scope, Nasscom says the objective of these guidelines is to promote and facilitate responsible development and use of GenAI solutions by different stakeholders.
A stakeholder could fit into all three categories of research, development, and use of GenAI, say the guidelines.
The research on GenAI solutions also includes both fundamental and applied research on GenAI models, applications, tools, and techniques.
Development of GenAI solutions means building and developing GenAI models and sourcing training data for GenAI model design and development to design or power various applications.
The use of GenAI solutions includes employing AI models for products or services for commercial or non-commercial purposes.
In terms of potential harms, the Nasscom guidelines seek to help mitigate the proliferation of misinformation, disinformation, and hateful content; infringement of intellectual property; privacy harms via the violation of data protection norms and standards; propagation of harmful social, economic, and political biases; large-scale job displacement and loss of livelihood and economic strain for the industrial workforce; carbon and water footprints and environment degradation; and surge in malicious cyber attacks.
There are obligations for those conducting fundamental and applied research on GenAI models, applications, tools, and techniques, says the Nasscom report, adding that the industry needs to exercise caution and foresight; transparency and accountability; reliability and safety; inclusion; and work towards the progress of humanity as a whole.
For those engaged in the GenAI development, Nasscom says, they need to evaluate potential harms from the development and use of GenAI models; disclose data and algorithm sources; practice due diligence in the adoption of means and methods for solution development and deployment; devise tech means for furnishing explanations for outputs generated by GenAI solutions in high-stake contexts; show accountability and support the progress of humanity as a whole.
For those using these models for commercial and non-commercial use cases, the Nasscom report says they must do internal oversight and comprehensive risk assessment; disclose all tech, non-proprietary information about the development process, capabilities, and limitations of the downstream models and applications; use GenAI in compliance with terms of service; devise tech means for furnishing explanations for outputs generated by GenAI solutions in high-stake contexts; and use it in alignment with the positive goals of human progress.
The tech industry body says the joint obligation of researchers, developers, and users is to steer and support universal AI literacy and awareness programs, technical AI safety research, and regulatory reform projects focussed on designing guardrails for safe AI use.
As researchers, developers, and users of GenAI technologies, everyone must contribute to the co-development and adoption of actionable guidance tools and best practices to enable all stakeholders to comply with these guidelines, says Nasscom.