Muddu Sudhakar and I share the perspective that the real opportunity for Generative AI startups is not in building platforms but in piggybacking on other platforms.
This conversation deep dives into the subject with real world examples and explores all the nuances entrepreneurs need to consider.
Sramana Mitra: Good to have you back! Our AI conversation has suddenly entered mainstream. We have a lot to catch up on and reflect on what’s happening around us. We know what Aisera does and how you’ve been one of the pioneering thinkers in applying AI to the customer support problem.
We had an extensive discussion on this about a year ago. Since then, ChatGPT and all these things have entered the conversation. Let me ask you a bit of an open question. How does this influence your thinking about what you are doing in your domain?
Muddu Sudhakar: ChatGPT has done great service to humankind. I put this one on the same level as what happened with cloud in 2006 and 2007. It’ll take years to come to see what this will generate. My son is in the middle school and he’s using ChatGPT. The bar for AI in the past was really high. You don’t set the same bar on humans. Now people are realizing AI can be humans too. AI will make mistakes. That whole thing has changed a lot.
At Aisera, we focus only on enterprise. In the enterprise, people have started trying it three or four years back. The payback now is that it allows you to cut down the cost. I can now use AI to eliminate my costs. AI will flourish in multiple ways. It can also improve our operations. I see a lot of promise, excitement, and interest.
Sramana Mitra: Let’s narrow down the conversation down to Aisera. How does what ChatGPT offer impact how you are thinking about your product roadmap?
Muddu Sudhakar: In the past, we started with NLPs. We started language models. Now that Microsoft has opened up Azure Open AI, we can build our NLMs better and faster. We’ve been compiling and using generative AI since 18 months. That was GPT3. Now it’s GPT4. Let’s say you wrote an article and you want it to become a dialog, we can do that. You want to generate code. We can do that. We’ve been using it, but this is coming out to the party now.
One area where we’re using generative AI is in making it domain-specific. You need a domain-specific NLM. We create enterprise domain packs on top of the base generative algorithms. Then I use your data to customize to your needs. As it evolves, you’re going to get more business, more articles, more content. Maybe user interactions will make mistakes.
The question is, “Should it be semi-supervised or unsupervised?” Maintaining those NLMs and how often you update, those are the things people are now thinking about. That’s where the unsupervised AI comes in. How do we maintain this NLM? Who owns it?
This segment is part 1 in the series : Capital Efficient Strategy for Generative AI Startups: Aisera CEO Muddu Sudhakar
1 2 3 4