Brad Smith, president of Microsoft, recently visited India. He expressed his views on India’s data protection law and AI. He appreciated that the data protection law has recognised data processor and data fiduciaries . Even data’s cross border movement has been allowed.

AI has the potential to do a lot of good. It is for us to do so responsibly and wisely. Machines may make decisions which so far were made by humans. We have to get it right.

AI may act on its own. That is an issue. AI should be a tool to serve people. AI products are coming too fast. There should be adequate usage of the product over a period of time.

AI laws should be built on existing laws. Non-IT persons should also develop AI skills. There should be collaboration with other countries. There should be focus on safety when a public agency uses AI. There has to be arrangement to slow down AI or turn it off when AI is used for infrastructure. It will ensure public safety. The control could be at the level of a model, or an application or at the level of a data centre, or at multiple layers.

Policy makers may be overwhelmingly optimistic or overwhelmingly concerned. Mostly they are balanced. They are testing the waters before acting.

2023 is the year of AI, and generative AI. It will stand out historically as an inflection point for the tech sector. Gen AI will be far more transformative for people’s lives than perhaps any other invention in people’s lifetime.

Brad Smith compares ChatGPT to printing press. Gutenberg perfected the printing press in 1462 using movable-type. Printing press made knowledge accessible to more people. Generative AI is similar. It is a tool to learn and to do research.

AI’s costing causes some concern. However, these are early days. These days everybody deals with LLMs or large language models, say GPT-4. In future, we may shift to narrower models, say some open source models which are smaller. These may not be as efficient as the larger models in one or two tasks. However, at the same time, they may do as well as the large model in one or two tasks. There will be constant innovation. The large models consume more energy. Companies should optimise the GPUs so as to make models more efficient computationally. There should be more investment in green energy.