“Think of foundational AI models as much more than that.”

New Delhi As Vice President of IBM Research AI, Sriram Raghavan heads the US company’s Artificial Intelligence (AI) Research Laboratories. Until recently, he was the director of the IBM Research Lab in India and the Research Center in Singapore. In an interview, he shared IBM’s AI strategy, his thoughts on how to look for return on investment (ROI) from AI, the impact of developments in quantum computing on AI, and how AI needs an ethical framework. Edited excerpts:

Why are the CXOs of so many companies around the world and in India as well, who have embraced AI still suffering from ROI?

The use cases are well understood. So, if you can run and build an AI model with the right investment, the business impact is clear. But will it take six months? Will I still ask 300 people to keep the form? These are the ROI questions they (CXOs) struggle with. This is why we are so excited about base models (large models like 3 or GPT pre-trained generative transformers) but think about them far from just big language models. The following idea is at the core of foundation models: Can I train a model to create representation without human supervision – under self-supervision? If yes, then I am only limited by the computing power and infrastructure to process all that data.

Imagine I have to do 20 NLP (Natural Language Programming) tasks including answering questions, sentiment analysis, and extraction. The traditional approach to addressing this has been to move from collecting and processing all of your data into a form.

With foundation models, you are not restricted to label data (such as “cat” or “dog”) because your model can be trained without it. I also don’t need to start with the raw data every time, so 20 AI models can be created using the same dataset. Thus, I pay the cost of data organization engineering once instead of 20 times (hence a better ROI). However, the challenge is that you must have the skills and computational ability to train these large AI models.

IBM has been talking about NLP, AI automation, advanced AI, AI scaling, and trusting AI as part of its holistic approach to AI. What do these terms mean for companies?

The focus on NLP and trust at its core is the recognition that there is a science around creating trustworthy AI. Then there is the activation of trust. In an enterprise context, this could include NLP to build conversational systems. NLP also allows us to extract ideas that help us automate information technology. There is also AI automation – applying artificial intelligence to doing business and automating IT.

Lots of people have to build AI models. How can we enable them to make building the fitting of them easier and faster?

This is sometimes referred to as Scaling AI. All of this is based on the fact that we still think of our AI and hybrid cloud strategy as tightly interconnected because we’re always building AI to run where the data is.

Given the huge strides that AI has made over the past few years, do you think we’ve reached the point where a breakthrough in AI could happen at any time?

Unequivocally, AI is not conscious. We’ve continued to enhance our ability to do pattern recognition, representation, and intelligent representation at scale with these recent advances – we don’t just predict and classify but we produce, but we’re still data-driven. Today’s data representations are becoming more powerful because they learn from the data. But we’re a long way from anything in AI that would be called conscious.

Give us some examples of how to automate artificial intelligence?

A use case that crosses industries and geographies, and that people find easiest to start with, is the conversation in artificial intelligence for customer interaction. The second is the application of artificial intelligence to IT automation. The third is process automation, workflow automation, or business automation.

We are also seeing a shift from automating tasks to coordinating tasks. Can AI Transcend Task Automation into Tasks – How to Perform a Credit Check; How is citizenship verified? Can he compile the flow knowing that is what you want to achieve? This is the vision behind Orchestrator, and it will expand the scope of automation.

The work around network automation is (gaining momentum) as carriers increasingly adopt new networks, 5G, etc., due to their need for more and more AI technologies. The research lab in India, for example, has been helpful in some of the work that we’ve done globally with this network and there are 5G operations where they wanted to use AI to help figure out the automated resource allocation for 5G segmentation (splitting the network into several segments). Virtual connections that can be customized to traffic requirements for different uses), etc. I also see a huge opportunity for more and more AI to emerge in sustainability, which is why IBM Research is putting so much into our work with our business units to launch the Environmental Intelligence Group.

Find all company news and updates on Live Mint. Download the Mint News app for daily market updates and live business news.

more less

participation in Mint newsletters

* Enter an available email

* Thank you for subscribing to our newsletter.

Add your comment

Leave a Comment