International | 10:30 AM
Nilesh Jasani, founder of the Geninnov Fund chatted with FNArena from Singapore about generative artificial intelligence, change and investing in innovation.
The discussion includes the size and scope of generative artificial intelligence; how technology and innovation are changing the world; what Geninnov invests in and why Jasani still prefers public markets to private markets.
Below is a curated transcript of the interview which is available via the FNArena Talks section of the website:
As well as via YouTube:
https://www.youtube.com/watch?v=0mVrFMA2CUY&t=1054s
This interview was conducted on October 9, 2024.
****
Danielle Ecuyer: How has your career to date brought you to the place of starting Geninnov, a generative innovation fund?
Nilesh Jasani: It appears like my entire career was always trending towards this point. I was an engineer, and I was fortunate enough to attend a college which had a semiconductor fabrication unit at that time. I was one of the few guys in India who manufactured something on a silicon wafer during my education. I was introduced to neural networks. So, two fields diametrically opposite, are the fields which are of critical importance today and were part of my education.
I started my career at IBM, and things that are important today, when you try to really compare x86 architecture with ARM, those kind of things were, again, part of my career. I also got to manage a team based in Seoul in Korea in the late 1990s. I got to manage a team based in Taipei in the middle of 2000 when companies like Samsung Electronics and TSMC were coming out.
I have been following these fields and areas all these decades, even when I was managing Jefferies and doing all sort of other things. These things were somehow always important to me, for no other reason than they were simply interesting.
Something happened in 2022. I said to myself I want one more inning of the career. The way I see it, 2022 is when humanity learned how to manufacture intelligence without two people having s-x. And that is ushering us into a completely new era. Almost everything that was happening in the name of AI and technology until 2022 meant we could not create machines that had ‘intentionality’.
Intelligence is something that is very easy to see but very difficult to define. It’s like you have an infant. You’re barely teaching her/him how to poop and how to cry, and one day, he/she knows how to manipulate your feelings by crying in a particular way to get picked up. This is something the baby is doing that our machines could not do.
Development of ‘intentionality’, a narrow neural network method in 2022, completely upstaged that, and that is the reason why the world is in a completely new stage for technology, and a completely new stage for humanity. We have machines that can address every problem out there at a different level of complexity compared to our human brains, and that is ushering us into this era of innovation.
I want to have a front row seat to possibly the biggest theatre of our lifetime. That’s why I am starting on the journey of this generative Innovation Fund, because I want to bring together various experiences and interests I’ve had in life.
I’ve been fortunate to start new companies and new divisions and worked in so many different markets. I’ve seen different aspects of technology, from IBM to software, to Palo Alto where my all my friends are, and so on and so forth, bringing all these together appears like something that will allow me to incorporate various things I’ve learned and that have interested me.
Question: Could you explain how you view the size of the generative AI market?
Jasani: Look. It’s going to be all-pervasive.
Transformers started as a narrow neural network method to roughly guess the next word in say, English. That’s how the guru math was initially designed in Google Labs about five years ago; and humanity stumbled upon something that is most amazing, one of the greatest discoveries of all time, and we are unable to fathom where its applications are going to end.
What started as a narrow neural network method to understand one language, somehow, the same formula figured out all human languages. The same formula, in no time, figured out all computer languages that we have. It not only did that at the next level of scaling, by 2023 it figured out human vision at the next level of scaling, which is ongoing. It is figuring out inter-relationships and things in our DNA data, or earthquake tremors, or crystal structures; that nothing in our mathematical or statistical toolkit could do until now.
Where this is headed and where it is going? All of us are shocked by the applicability of this method, and machines have started working on themselves, so the implications are not in chat boxes and Copilots. Those are narrow day one use cases, like the way email was for Internet. In the same way the Internet did not remain about computers to computers, communication through emails, but turned out to be much, much more.
This is far bigger.
While popular imagination is still fascinated by what chat boxes can do or cannot do, and the various issues over there, the implications are for driver-less cars. Also, the implications are for robotics, in molecular sciences, drug discovery, renewable energy and whatnot.
It’s impossible to even forecast the fields of applicability, let alone the amount of business one can do. I am more in the camp that this is going to be as pervasive as internet or mobile telephony, and with far more applicability over time.
Question: Is the evolution dependent on the build out of infrastructure, like data centres?
Jasani: The simple reality, the way I see it, is that nobody knows exactly where this technology is going and how it is going to evolve. In Chapter One of this era, we are in this concept of data centres and the global brain concept that if Danielle wants to know what is two plus two, that thing will possibly go to some data centre somewhere where 20,000 Nvidia chips are interconnected, and somebody will do a trillion-by-trillion matrix multiplication, and come back saying ‘four’.
Do you really need to do that? Are there other ways to do it? Should you not have an agent that says this is a solved problem that can easily get solved on your calculator, rather than go to a data centre?
The reality is that over time, the AI compute, which is 95% in cloud today, will possibly go to 70% in your pockets; 70% in the gadgets or instruments that you have. Your compute will have to be decentralised. Things will have to move to the edge.
Many other things will have to happen. We may stop thinking more in terms of our queries and chat boxes, and we may start thinking in terms of how we use the new technologies and the ones that come thereafter, to really change the driving, to really change the way we work at home, and in all sorts of devices and infrastructure.
So rather than waiting for something to emerge, I think all of us need to start participating. The sooner we get started, the sooner, we ourselves will come up with new innovations and new solutions.
Question: How do you look across the universe of stocks that you are afforded, because you’re looking across multiple countries, you’re looking across multiple industries. Is it a case of trying to isolate down those companies that have large R&D spends that, as you alluded to, are already first movers?
Jasani: The way we arrive at our conclusions and our starting point is always through material, scientific and technological innovation. We are in a new era. We strongly believe that internet era business models are shifting. The Internet era gave rise to a particular way of businesses, a particular kind of businesses succeeding. Most of the gains were captured in the application layer.
It gave rise to a different way of investing, venture capital investing. That’s when innovation moved to garages.
I think in the generative AI era all these business models are shifting. Ideation is getting commoditised. Computers are no longer speaking specialised languages. One can deal with machines in human languages. It is giving rise to completely different kinds of forces.
It is giving rise to this era where anyone can come up with products and features, but making money out of ideas alone or products and features alone is becoming more and more difficult. So, the business posts are shifting, more and more value is captured in hardware or near hardware layers, rather than in the application layer, and that’s the material shift for all of us.
Currently, there’s a lot of debate on this topic. We have our views, which are somewhat extreme, but there’s a lot to think about. The second thing we do is we try to understand not only innovations, but the companies and trends by being close to the sources. We try to spend far more time reading MIT journals and Nature magazine than say, reports that come out of investment banks.
We are trying to understand innovations at the root by talking to the right people, rather than talking to CFOs and CEOs. We are trying to really then focus far more on monetisation rather than simply products or features.
You see what’s happening with Microsoft, or what’s happening with Salesforce, or what’s happening at SAP; suddenly the numbers of products and features are going through the roof.
You can say that some of these companies have introduced more features in the last one-and-a-half-years than over the previous 15 years, but the ability to monetise is in question when the value capture is happening somewhere else.
We try to really focus on those things. We are also at the application end. We realise that lots of these things will spur innovation, like mobility. There are some amazing things happening in terms of level three and level four autonomous driving.
The field had somewhat stagnated over the last 10 years, but suddenly, as machines learned how to really train themselves through observation, level three and level four mobility have become feasible. We also invest in mobility.
We look for where robotics could be going from here. If people like Tesla or Apple are indirectly telling us they are new businesses, there’s something else going on. We try to look at demographic forces, why automation could be resisted in certain spaces, say in countries like the United States, while in similar spaces, the automation would be more than welcome in places like China or Korea, because demographic forces are very different.
Our entire thing is we don’t care about how many companies are listed out there, and we don’t care, obviously, about learning all of them for their quarterly results or for their valuations and so on and so forth.
Our processes are the other way around; we focus on innovations we believe in, and we look for the right expressions. At various points, we add angles like monetisation, macro factors, regulations, valuations et cetera. In a way, it’s a process that’s been successfully adopted by people like Warren Buffet that does not worry too much about what we don’t know, but try to really do what we know reasonably well.
Question: You hold Novo Nordisk, could you just possibly run through some of the details of the thesis for this so people can understand a real-life example?
Jasani: You possibly captured one of the three examples in our portfolio that has very little to do with generative AI. This exception is possibly the best way to explain the real nature of the portfolio, or rather our fund; we are an Innovation Fund.
We are not a hardware fund; we are not an AI fund.
Innovations can happen without any machine efforts, and that’s fine. The way we see it, is that we are in the era of hyper change. We were in the era of change between 1992 and 2007, when innovation was one of the biggest drivers of businesses and investments globally; when our consumption baskets were changing, when our languages were changing, when we added hundreds of new words to our dictionary, like web, internet, laptops, Palm Pilots, Blackberries and whatnot.
A lot of things had to go away from our basket, like video VCRs and fax machines, to make way for the new things.
Now we are in the era of hyper change. In between, there was a bit of a stagnation. But once again, our languages are changing. Agents are no longer chain spots, and our consumption baskets are changing.
We are replacing our potato chips with GLP-1s. I think GLP-1s is one of the most amazing innovations of all time. Hopefully there are no side effects that we come to know 20 years later.
Somebody I read last week even claimed it’s a bigger innovation than generative AI. Here is a molecule which is a gift that keeps on giving. It was supposed to be for your appetite, a suppressing molecule and not only after it became a solution for diabetes. What we are witnessing, almost on a quarterly basis now, is that it’s a molecule that could be a solution to certain kinds of heart diseases, kidney diseases, liver diseases, depression. GLP-1 is currently going to less than 2% of the people who need it.
It is such an amazing new molecule, and its potential is so huge. While prices will have to keep coming down over time for it to reach more and more parts of the population, there will be further innovations in the same space.
I think everyone has realised there’s a lot of potential in surrounding this space, the GLP-1 inhibitors. So absolutely, a lot more things will have to happen. But in companies like Novo, companies like Eli Lilly, you have completely different types of innovation that is sweeping the world. And they are attractive to us. They’re part of the kind of things that we look at; it’s fascinating and it’s interesting.
Question: Can you explain to people how autonomous driver-less vehicles are being created?
Jasani: Human vision was a problem. We could not make machines learn through observation in the way our kids do. Instruction-based programming is always something that doesn’t work in real life. It could not work in translation.
You may remember there was a time when Google Translate completely went away from instruction-based ways of translating from one language to another and went to learning based method. That happened a few years ago. Now, somehow, we could not do that with vision until transformers came in. With vision transformers coming in, roughly about a year, year-and-a-half-ago, it’s become possible for us to use vision data and let machines figure out whatever lessons they need to learn.
They’re spurring them suddenly in a new direction. Both robotics and autonomous driving are taking off because we can use visual data and let machines figure out what they need to do, rather than provide line by line instruction on 10s of 1000s of 10s of millions of things that happen in real time and how to deal with them.
This is effectively the reason why people like Elon Musk, companies like Tesla want to be in China versus, say, rely on the vision data only in the West; because the laws are very different.
The way you can possibly use camera data from cars moving on the street in China versus the US and the way you can use those data train your models. There are a lot more restrictions in privacy-conscious countries like the United States, while there are far less restrictions in China, where your models can improve very rapidly, because you can use a lot more data.
The entire driver-less mobility is going in different directions, even in terms of model developments. In many countries in the world people will be far happier overall if they can pay the cost for, say, getting a cab ride in a driver-less vehicle compared to a driver vehicle, and that may become the primary factor in an election.
In some other countries the attention could be more on the employ-ability of all the people who are involved in that industry, and producer utility versus consumer utility. All of us are producers and all of us are consumers. Which one dominates is different in societies at different times, and that is what will give rise to completely different innovation directions, not just in robotics or mobility, but across the board.
We’re also witnessing how currently chat boxes, consumer applications, and corporate efficiencies are far more important in countries like the United States compared to China. So different countries are moving in different directions with all these technologies, and that’s something that one should keep in mind by evaluating them
Question: How do you look at or do you look at rationalising valuations versus those companies that are at the forefront of innovation, with or without the use of generative AI? What interplay does macroeconomics have, if at all in your decision making?
Jasani: Investing is very difficult in public markets. As you know, you always end up having to look at too many factors that don’t have much to do with businesses, whether they are interest rates or some election or some geopolitical factors and so on and so forth. One must be somewhat mindful.
For the fund, the reason why we are in public markets and not in private markets, is because we think that a lot of material innovation out here will have to happen with large companies that offer innovation monetisation.
Innovation is more and more moving away from garages and small companies and into the domain of giants. That’s our belief and that’s the reason why we must be in public markets.
We also want to keep an exit window open. Too many things are changing. We don’t know where we are headed. I don’t think anyone knows where the world is going to be next year, same time, let alone 2030, or 2025, and investing with your exit window closed in some multi-year vehicle has become risky when the world is innovating.
The world is moving so rapidly.
I think the most important thing to do is understanding the business investment. Try and tune out as much to other macro news as possible, if you are doing genuine business investment. Tuning out is difficult, but that’s something we have to do.
Valuations are important, but equally, if we are thinking about investing in companies that are going to be growing at, let’s say, 15% or more over a long period, and are accumulating IPs, accumulating leaderships in space and riding ahead of competitors like, say, Nvidia, creating the competitive distance.
Then you must think about valuations very differently. It is not value investing. It is growth investing, and as a result, yes, absolutely, you must look at valuation, but not in the traditional sense. You can’t look at a five-year price to earnings chart; you have to compare to growth prospects.
****
More information on Nilesh Jasani and the Geninnov Fund can be found at https://www.geninnov.ai/
Find out why FNArena subscribers like the service so much: “Your Feedback (Thank You)” – Warning this story contains unashamedly positive feedback on the service provided.
FNArena is proud about its track record and past achievements: Ten Years On
