The Generative AI Revolution: Leveraging while Mitigating Concerns is Key

By Arvind Tiwary, Chairperson, Cybersecurity Working Group, IET Future Tech Panel

Generative AI (genAI ) such as ChatGPT and Midjourney have the power to revolutionize industries and transform the way companies operate. There is a technology stack of foundation LLM( Large language model) , training data ( English and vernacular languages as well as voice and video in multimodal engines , large compute engines for conversations that are 10-100X more expensive then Google search.
The data centers to host LLM like ChatGPT are constrained by availability of expensive GPU farms (10-20,000!!) and will take 100MW electricity and 100 tonnes of water to service 250 million conversations a day. Even after more then 6 months of general availability there are quotas on queries and response times are well over a second

Efficiency and Cost Reduction
One of the most significant advantages of using generative AI is efficiency, which can lead to cost reduction. For instance, companies can use generative AI to automate customer support services, reducing the time and effort required to handle customer queries.
Another example of using generative AI for efficiency is in the healthcare industry. Hospitals can use this technology to automate patient consultations, which can reduce the time required to see patients. This can lead to significant cost savings for hospitals, as they can service more patients with same doctors

Translation between languages are vastly superior with these tools and may enable greater use. Dr Vivek Raghvan at AI4Bharat have demonstrated voice query in Tamil being answered back in spoken Tamil but using English OpenAI backend.
GenAI tools can be utilized as a virtual assistant to assist in editing and proofreading blog articles. It can help identify grammatical errors, suggest sentence restructuring, and provide general writing suggestions. Many experts use them for summarizing long articles and helping triage research effort

Effectiveness and Quality of Results
Education can be vastly enhanced by tailored coaching to young students, Tools can vary different explanations and convert quizzes to storytelling and relate to fables and other pedagogy.

Generative AI can also be used to improve the effectiveness and quality of results. In computer systems support they can be leveraged to identify and fix problems before they become dangerous and costly. It does so by analysing system logs and user feedback using technologies that are capable of identifying potential issues and suggesting solutions in real-time

Another example of using generative AI for effectiveness is in the legal industry. Law firms can use this technology to automate legal research and generate legal documents, such as contracts and agreements. This can lead to significant time savings for lawyers, allowing them to focus on more complex legal issues. Consumers can use them as assistants to research product comparisons, travel itinerary

BlueSky new usage
GenAI tools allow artists to create very different new art in different styles of painting, re-mixing songs and SciFi themes. Creative storytelling and infinite games to entertain and provide joy and fun are possible

Model-as-a-Service (MaaS) is one of the most popular business models for generative AI developers. Providers like OpenAI, Microsoft, Google , Databricks Anthropic and META may offer free access at some usage level and paid access for more features. In Indian and enterprise context fine tuning these base models is necessary and may be done using open source alternatives like Hugging Face.

Users will play and create far more new applications and usage then we can imagine. We need to encourage and let that rip

Disadvantages and Ethical Issues
There are also some disadvantages and ethical issues that need to be considered. Deepfakes and false storytelling in political campaigns, advertising etc is a major concern. GenAI are not factual query but engaging story tellers and happily invent fake data and narratives. A Lawyer in the USA got sanctioned by a court for submitting false references created by ChatGPT to non-existing cases.  Most GenAI tools are being socialized in the USA and may have in-appropriate values and norms. Specific legal requirements would need to be placed as EU has done to tame the tools. A basic requirement should be that outputs should be clearly labeled as having used GenAI and provide complaint and redressal mechanisms. Restraints on data acquisition and data provenance and copyright safe is a big requirement

Policy to speed adoption
We need to vastly increase access and develop skills in various industries to use these tools. We should also provide subsidized compute infrastructure for using open source lower cost engines and re training for Indian context like agriculture , education etc

Conclusion
In conclusion, GenAI has the potential to revolutionize various industries in India.. The change will reduce costs, scale productivity and quality outputs and help create new business cases and industries. Government needs to help create more access and alternatives using opensource and India appropriate products