2023 will likely mark an inflection point in humankind’s quest for Artificial Intelligence (AI). The year saw public attention focussed on ChatGPT and the unveiling of other major AI products, such as Google’s Bard. These generative AI initiatives, underpinned by large language models (LLMs), are significant because they may pave the way for a more general AI indistinguishable from human intelligence. These developments demonstrate the enduring merits of promoting cross-border and cross-sectoral flows of science and technology, making a case for why India should embrace techno-globalism.
Generative AI learns from datasets that form neural networks, which help machines learn tasks by analysing data. (Thinkstock)
First, globally accessible science and research are the bedrock for AI innovation. Generative AI learns from datasets that form neural networks, which help machines learn tasks by analysing data. Computational gains made via hardware advancements allow products such as ChatGPT to learn from large datasets quickly. But hardware’s brute force is impotent without cutting-edge research and development in underlying neural network architecture. These architectures are now globally available, thanks to open-source science and research.
The role of open science is best exemplified by the developments in 2017, when Google Brain and Google Research scholars introduced a new network architecture called “Transformer”, which forms the bedrock of ChatGPT-like models. The Transformer AI model is a neural network that learns context and can track meaningful relationships in sequential data, reducing the time required to generate responses to questions.
The researchers at Google open-sourced their work, recognising the benefits of knowledge spillovers — a hallmark of techno-globalism without which generative AI would not be where it is today. Similarly, Meta released a model called LLaMA or Large Language Model Meta AI in February — designed to help researchers without access to large amounts of computing infrastructure study large language models.
Today, data from the public internet is used to train generative AI models. Thus, researchers and developers in different parts of the world are on a level playing field when it comes to data. They don’t need access to proprietary datasets to run powerful AI models until they need to train them for very specialised applications. For instance, “Common Crawl” is a globally accessible repository of raw web page data, extracted metadata and text extractions mined from the public internet since 2008. This was a crucial input in the development of ChatGPT.
Finally, governing the mass adoption and use of technologies such as ChatGPT will require global cooperation — national responses will not suffice. For instance, OpenAI, the company behind ChatGPT, cautioned us on the various harms that could stem from the misuse of such technology, including the automated generation and propagation of disinformation on a global scale. In 2021, the United Nations secretary-general António Guterres called for a “global code of conduct that promotes integrity in public information” to “better regulate and manage our digital commons as a global public good”. Analogous international, multi- stakeholder efforts to guide responsible AI development are on the cards. For instance, the Global Partnership of AI is an international initiative to support the holistic development of AI. It is a coalition of 25 countries, which includes India as a founding member and current chair.
India benefited immensely from the free movement of people and ideas in the first wave of information technology (IT) growth, and it should espouse openness to navigate the age of AI. Our IT industry is among the largest in the world because our globalist entrepreneurs made early outward forays. Consequently, India’s IT exports account for nearly half of total foreign exchange earnings.
Similarly, early AI adoption may help the domestic IT industry reduce reliance on labour arbitrage as its core business proposition. Conversely, it may increase the creation and use of scalable and flexible software products. The proliferation of 1,500 global capability centres across the country — captive operations hubs that handle IT support to enhance workplace productivity and drive digital transformation — indicates that India is well-placed to help the world automate with tools such as AI. An open policy stance encouraging cross-border digital trade and innovation is the surest path to achieving this.
E-Paper
