Generative AI, the creative frontier: how generative AI is transforming content and innovation

Hi. I'm Ravinder Singh, I'm Head of Digital Systems, working with Cabinet Office and particularly with Government Commercial Function. I have been in this role for the last two years and with the civil services for the last eight years, and before that, I was in the private sector. Today's talk is regarding generative AI, which everyone has used. ChatGPT is the most popular one that people have used. We will touch upon ChatGPT, and all the other large language models, and how this new generative AI technology is transforming content and innovation. This is very crucial because everyone has to use this technology.

Emerging technologies, as I said, are not just AI and machine learning, or generative AI, or ChatGPT, as people have thought of. It is much beyond that, and this slide presents the various technologies that come under this. It could be Automation, Hyper-automation, Internet of Things, Virtual Reality, Blockchain, Quantum Computing, Low Code/ No Code, Edge Computing.

This slide presents what's the hierarchy of these technologies. Overall, I can see as a Venn diagram that artificial intelligence, what is machine learning, and how the generative AI sits. What are the parts of it? You can see a lot of overlaps will happen among various technologies. It's not that just one technology is overwhelming. Basically, what happens is the machine is learning something, only then, once the machine has learned, then the deep learning comes into play. Then, from deep learning, you can have generative AI, large language model, functional models, symbolic AI, rule-based AI, rule-based systems, neural networks, fuzzy logic, which people have used for many years, Prolog, LISP, and other languages people have used for rule-based languages. They have been in use since '60s, '70s, '80s; It started with neural networks, fuzzy logic, and these all technologies they form the basis for artificial intelligence to work. What is needed here is the data, which has to be collected, based on that data, the machine can learn itself and develop its own intelligence. It's like training a kid from small to the adult that they go on learning the language. It's the same thing the data goes on increasing in the machine, and it can take better decisions for you rather than a human taking that decision. But those decisions have to be fed that if this is the use case, this is how the decision has to be made. If this is the use case, the decision has to be made. So the number of cases and the data, which is for those use cases, as it increases, machine becomes more efficient, more effective in giving you the right decisions, so that human intervention can become less and less. But otherwise, if you don't do that, the machine can hallucinate. The machine will thus think, "Oh, I can do this." That's where there is a challenge. That machine has to be trained with supervised learning or unsupervised learning, whichever way, active learning or passive learning machine has to learn. Only then machine can start behaving as you like it.