1Mby1M Virtual Accelerator AI Investor Forum: With Venktesh Shukla, Founder and Managing Partner at Monta Vista Capital (Part 2)

Sramana Mitra: Understood, but now let’s switch that to a different case study. Let’s look at the medical domain, which is very high potential for Generative AI. There is so much medical information on which you can train physicians, telemedicine providers, etc. However, if that starts to hallucinate, is the human in the loop completely capable of catching that?

Venktesh Shukla: I think that’s actually a very good case of using Generative AI, because ultimately it’s the doctor who’s going to make a judgment, write a prescription, and do the diagnosis. He will use this as a way to help him or her. So that’s an excellent case study of a human in the loop, because it is doing all the grunt work for him – looking at all the test reports, latest research, and studies. Based on that, it says, given these symptoms, this lab test, and research in this field, the chances are that this is diagnosis or one of these two. Then the physician can exercise his or her judgment. That’s very good case study of how generative AI could be useful in the healthcare industry.

Sramana Mitra: I completely agree with you. The question that I’m still skeptical about is, if there is that much data that it’s processing and bringing something to the doctor; in that moment, the doctor has to make a judgment call, and we are assuming that the doctor has more knowledge than the generative AI. And that is not true. He’ll never have more knowledge than generative AI.

Venktesh Shukla: No, he’ll never have more knowledge than generative AI, because a human cannot deal with the amount of data that’s being processed and the speed with which it’s being processed. But based on the data, if it comes up with some diagnosis that the doctor says that it makes no sense to me.

Sramana Mitra: If it’s obvious, then it’s fine. You know, maybe first degree, second degree, it’s fine. But the moment it gets to complexity, I think medicine actually is a dangerous place where I don’t think this problem is managed yet.

Venktesh Shukla: But I think that’s actually a very good place to use generative AI.

Sramana Mitra: It is the place where it should be used. At the same time, the hallucination problem has to be contained there. You can’t let AI hallucinate in medicine.

Venktesh Shukla: Yes. But then you have a human in the room. You have an expert in the room.

Sramana Mitra: You have a human in the loop, but the human is not equipped to catch hallucination. That’s my issue.

Venktesh Shukla: Yes, but he’s applying some judgment based on his years and years of study and expertise. So I think that’s a much better use case than an enterprise one.

Sramana Mitra: Sure. Where are you investing, Venk? Have you invested in Generative AI companies? What has passed the smell test of where you can get around the problems and actually can go to market with something interesting and maybe with the human in the loop? A lot of investors we’ve talked to are only investing in human in the loop companies.

Venktesh Shukla: Yes, we have. Interestingly, we invested in a generative AI company before that word or expression entered popular culture. So we invested in the company in October of 2022, a month before ChatGPT came out.

This team, originally from Nvidia, is using generative AI to create 3D digital humans that can interact with people, perform functions, answer questions, and provide guidance.

Initially, we envisioned its application in gaming. In fact, Jensen Huang highlighted this startup’s demo in his keynote, which was viewed by seven million people. The demo showcased a gaming scenario where, unlike traditional characters whose reactions are all pre-programmed, these digital humans can respond dynamically. For example, if a character has a backstory like Rocky (Sylvester Stallone), they wouldn’t give a long-winded response to “How are you feeling?” Instead, they’d reply tersely, like, “Yeah, get out of here.”

This technology reduces the need for pre-programming every response, making interactions more natural. While we initially targeted gaming, we’ve been surprised by its potential in enterprise applications. For instance, Unilever uses it across 20-30 of its shampoo brands; a digital human could help a customer choose the right product by interacting as a human would. NASA is also exploring its use for training astronauts for life on Mars.

These digital humans possess human-like language abilities, intonations, and movement. They can discuss any subject as if they were real human experts.

Sramana Mitra: But it’s not built on generative AI?

Venktesh Shukla: No, it is. It’s a classic case of generative AI.

Sramana Mitra: So then how are you dealing with hallucination in that context?

Venktesh Shukla: In this context, hallucination doesn’t really matter. If you’re discussing shampoos, for example, and there are 35 options, the system might ask, “Do you have dry hair?” If the answer is “yes,” it simply follows a specific path. Hallucination isn’t an issue here because it’s not relying on extracting precise data or making a recommendation that could go horribly wrong.

Sramana Mitra: OK, but in the Mars case study, it could go horribly wrong, no?

Venktesh Shukla: Well, I think there it is. I’ve just seen a quick demo of it, but I think the mass thing is about how do you deal with the environment? So if you see this, how do you react to it? If you have this kind of weather, how do you react to it? That’s the kind of use that it is being put to.

Sramana Mitra: OK, so you’re saying it’s good for non mission critical use cases. This is not a human in the loop example. You have taken the human out of the loop and that is the benefit of it.

Venktesh Shukla: Exactly. So in this case, where it’s not catastrophic, if it makes a decision, it’s not a catastrophic disaster. But it is in the medical example we discussed. That’s why you need to have that physician in the loop.

This segment is part 2 in the series : 1Mby1M Virtual Accelerator AI Investor Forum: With Venktesh Shukla, Founder and Managing Partner at Monta Vista Capital
1 2 3 4