ExclusiveHow AI Models Are Impacting the Environment and Society | Vishnu Pendyala

So there's a lot of call for regulating AI. Some of the senior folks in the industry said that AI could cause a lot of trouble later on. These models hallucinate and produce misinformation. They produce misinformation, that's one thing, and they are used to produce misinformation as well, like deepfakes and things like that. There's so much fraud happening. Our Bay Area innovator today is Dr. Vishnu Pendiala, professor in the data science department at San Jose State University and AI expert. Vishnu is also part of the San Jose State Academic Senate body. He was a distinguished speaker at numerous conferences and wrote several books on big data and data mining. These models are not explainable, because there are like 175 billion parameters, floating point numbers, being crunched. We don't know how a decision has been produced, how an outcome has been produced. So when we don't understand the system, how can we control it? This is Bay Area Innovators, and I'm your host, Steve Vespas. Professor Pendiala, welcome to our program. Sure, it's a pleasure to be here. Glad to have you here. So recently you wrote an op-ed, actually in our publication in the Epoch Times, about SB 1047, and this had to do with putting some regulations into AI. Can you please comment on that? You're very involved with AI, you're an expert. Tell us about your thoughts on the op-ed of what you were actually talking about. So there's a lot of call for regulation, regulating AI. Probably because some of the senior folks in the industry said that AI could go out of hands, and it could cause a lot of trouble later on. There has been a lot of call for regulation. So this bill was about regulating artificial intelligence. But the bill was mostly focusing on the long term, and preconceived notions of the dangers of AI, rather than the imminent dangers of AI, in my opinion. So it was talking about a kill switch, but there are some imminent problems that need to be addressed. One problem is bias, fairness of the AI models. AI models, by their training, they are biased, because they are trained on the real world, which is not perfect. So they carry forward the same biases. So that is one problem with AI models. And the other imminent problem is hallucinations. These models hallucinate and produce misinformation. They produce misinformation, that's one thing, and they are used to produce misinformation as well, like deep fakes and things like that. There's so much fraud happening. And then the language itself. There were a couple of, or maybe more, instances of the word reasonable. What is reasonable? I mean, what is reasonable to me may not be reasonable to you, right? That will lead to a lot of ambiguity and lawsuits later on. And the other problem I have with the bills, these kind of bills, is that most of these companies are operating internationally. So if every state and every county, for instance, starts coming up with their own laws, compliance will become really difficult for the companies. So governments of countries talk about tariffs, taxes, but I think they should be talking about this uniformity of the law, which governs the companies, which operate in multiple countries. So there should be collaboration in that respect. Now, I think you were the victim of some fraud that resulted from deep fake or AI-related things, right? That was part of the reason to write the op-ed, right? Yeah, so yes, I was impacted by misinformation multiple times, multiple ways. One of the focuses of my op-ed was misinformation. So what I said was, if the law mandates data provenance, that means traceability of data, like how the data transformed, the origin, capturing the origin of the data, and how it got transformed, if at all there are any changes to the data. If all that metadata is maintained mandatorily, then my belief is that misinformation will come down. Just like tracing the flow of money, right? Once everything is on record, there are very few malpractices that can result from money flow. Same thing with data flow. When we trace everything, then there is very little chance for malpractices with data. So there are a lot of websites, for instance. There is a website called RateMyProfessor. There is no user ID, people can submit anonymously, and the person may not even be a student. They can write some nonsense there. And the law protects the website. It is really unfortunate that the law is actually protecting the wrongdoers, in this case at least. That was my pet peeve. So I thought data provenance will probably help in that respect. Let's talk more about the challenges that we're facing with AI in the future. So I think we talked about the kill switch, in the case that AI takes over in a way that we don't want it to take over. There's also bad actors using AI in a way that is damaging to us, to national security. But there's also probably more tangible things like energy and water. And I know you made that point. Can you emphasize more on that, on the energy aspect of it? The AI models consume a lot of energy. It's a well-known fact. In fact, there are a lot of analogies, but the analogy that immediately comes to my mind is that training a large language model like ChartGPT, for instance, produces probably more carbon emissions than for the entire life of a car. And answering one query can take the same amount of carbon emissions that result from watching like 5 to 120 videos on YouTube, movies on YouTube. So that's a lot of energy. That's because all the searching of ChartGPT is doing to figure out the information, right? Yeah, so the model itself, if we take the original version of ChartGPT, for instance, it had 175 billion floating point numbers in the model. Deploying that model, utilizing that many number of floating point numbers requires a lot of computation. We can argue this way or that way that the carbon emissions are causing climate change and all that. But more than climate change, what... My intuition tells me that consumerism is not really good. I believe that consuming will have its effects eventually. So, these large language models consume a lot of energy, not just energy but also water because the computation generates a lot of heat as well. For cooling, there's a lot of need for water. So, there's a lot of ecosystem, natural ecosystem, that is being consumed to get these artificial intelligence models to run. Anything else that you think we need to talk about as far as the potential dangers of AI? So, the biggest danger that people have been talking about is that it may go out of control. I don't think we are there yet, but there is a reason for that fear. The reason is that these models are not explainable because there are like 175 billion parameters, floating point numbers being crunched. We don't know how a decision has been produced, how an outcome has been produced. So, when we don't understand the system, how can we control it? One other problem that we should probably think about is the wealth inequality. So, AI is taking over a lot of routine jobs, which will be very disruptive. The routine tasks are getting taken over. So, what remains for people to do is highly intellectual. That leads to fewer and fewer people getting those kind of jobs, who will be making a lot of money. And the others will probably be struggling. I was reading some statistics. I don't exactly remember the numbers, but I think 25% or 10%, something like that, of us in the Bay Area are earning more than 300K, 325K, whereas the others are really struggling to make their ends meet. And that is going to further widen. That wealth gap is going to further widen because of artificial intelligence. That's a very good point. And now we're talking about these robots that are able to function pretty much to the work that a human would do, as far as taking out the garbage or cleaning or something. They will probably further impact even the lower-level jobs that people would be doing, I would assume. Yes, that's right. With drones and all these new technologies coming up and artificial intelligence powering them, I think we are in for a really big change. You come from India, and you were ranked 69th in your undergraduate class out of tens of thousands, maybe hundreds of thousands of students. And also you won a gold medal in the Math Olympiad in India. Please tell us a little bit, how did you accomplish this? What is the base of your education and your philosophy that you're able to accomplish what most people would consider pretty amazing accomplishments? Yes, math has been a favorite subject for me since my childhood. And the other thing is, I believe that to a good student, everything in this universe can be a teacher. And then we need just one teacher to make a whole lot of difference. And I had that kind of a teacher when I was in the 10th grade. So that's how the journey started, and I never looked back. The good thing about math is it is very analytical, which aligns with my kind of thinking. So it's kind of natural, I believe, that my liking for math has led me into this journey. It's interesting you mention, I think there's several times I talked to people who did very well in their career because of their teacher. I talked to a musician, and she said she got involved in music, in love music, because of her teacher or her coach. That's a very common theme. Now you came to the U.S., you worked in tech at Synopsys and Cisco for almost 20 years, and then you moved into teaching. Please tell us, why did you make the transition? Yes, teaching was always my passion. Even when I was young, I used to teach my neighbors, I used to teach my sister, younger sister. So teaching has always been a passion. In fact, I taught an engineer. He was kind of appearing for an exam, and he wanted to brush up his knowledge of some engineering subjects. So he was many years older to me. So I trained him, and he was very impressed. He said, I have that knack of putting things in the right way that students can understand. So that confirmed my belief that I'm a good teacher. So I used to take up part-time teaching every now and then whenever I got a chance, even in India. And then when I came here, I came to know that Ph.D. plays a big role in becoming a teacher, a professor. So I pursued my Ph.D. Then as soon as I got my Ph.D., I started teaching part-time at San Jose State University as a lecturer in computer engineering department. I really liked the experience, and the students liked my teaching as well. So it was a win-win situation, and then I decided, why not try this full-time? Because I'll have more time for research that way. Teaching-wise, whether it's part-time or full-time, I started teaching two courses initially. But the full-time opportunity helped me move into research. That's a big thing. And then service. So in the industry, there's not much of credit given for service to the community, service to the subject, to the profession as such. Whereas in academia, that's a given. I mean, we are supposed to contribute to the profession. So that's what lured me into full-time teaching. There was this opening that came up in Applied Data Science department. So I applied to that, and after the interviews and all that stuff, I got selected, and so far the going has been good. What would you say is the most satisfying thing about teaching? So there are many elements, but if you want the most satisfying element out of all those, I think training the next generation of workforce is probably the best reward that I have, I derive from this profession. And the other thing is staying with the young people makes me young as well. And the intellectual environment is really stimulating to the mind and body as well. Let me change gears a little bit and talk about music. I know you have an interest in music. Actually, I've seen you play some songs and sing some songs on video. How did you get interested in music? And tell us some of your favorite groups that you like. So music and math kind of go together. My dad used to say that music is permutations and combinations of the notes. You come up with the new permutations and combinations of the notes, permutation of the notes. You get a new genre of music. So it's very closely related to math, in my opinion. And that is one of the reasons why even today I use deep learning and machine learning models to analyze music and do stuff with it. But yeah, my interest in music dates back to my childhood. So we used to really wait for the Grammy Awards to be announced. Within a week or two, I think there was a company called Billboard, I think, which used to release cassettes of Grammy Award-winning songs. So we used to go and buy them and then listen to them very ardently. So I had quite a few favorites. So he asked about the albums. Some of the albums that I really loved as a small kid was ABBA, of course, and Boney M. As I started growing older, I mean a little older, into adulthood, Michael Jackson's albums were a big hit with me. It's interesting you mention ABBA and Boney M. because I grew up in Romania in the 80s and late 70s. Those were the most popular groups. Of course, ABBA, everyone knows, they're from Sweden. But Boney M. is a little bit less known. They're from Germany. I always thought they were American, but they're actually from Germany. And they had a lot of good songs, and I think you sang some of them. Daddy Cool, I think it's one of them. Yeah, Daddy Cool. That was my childhood favorite. Rasputin, I think, is another one. Yeah, Raul Rasputin. Let me change gears again. I love to talk to scientists such as yourself about the connection between religion and science. Many would say that they're at odds with each other. I really want to ask you, is religion and science at odds? Do they have to be at odds with each other? Or is there a common path that can be agreed upon? First thing is, I have my own opinion about religion. I don't think religion is really something... It's a set of faiths and beliefs. People gave it the name religion, but it is not supposed to be practiced in the way we are practicing religion today. It's a set of faiths and beliefs, right? I mean, I believe in science. Somebody else believes in social studies. Somebody else believes in languages. We don't fight with each other. We don't say that my belief in science is much better than your belief in English. So, I don't think religion really was invented to be the way it is today. So, for instance, a lot of people, I think, Epoch Times founders believe in Buddha or Buddhist teachings. People call Buddhism a religion, but many think that Buddha never started a religion. He never called it a religion. He just told some good things, some kind of things that we need to believe in and the faith that we need to develop. But I don't think he really used the word religion or he really meant religion to be in the sense that it is being practiced today, Buddhism being practiced today. It's not exclusive. It's all inclusive. So, I studied quite a bit of Sanatana Dharma, what we call as Sanatana Dharma. Sanatana means timeless. So, this timeless wisdom tells that it's again a set of principles, a set of guardrails, principles, theories, some of which are difficult to comprehend. That's where the faith comes into picture because we cannot perceive the whole thing. We may have to just blindly believe certain things that whatever is said there is correct, is true. So, that's where the faith comes in. But otherwise, most of what I studied there are quite related to the science and engineering and the math that we study in schools. Just to comment to you on the founders of Epoch Times, they were involved with the Falun Gong meditation, which is actually part of the Buddha school. So, there's definitely some similarities there. But the question would be basically how the universe was created. We talk about the Big Bang, or religions talk about God creating the universe. Maybe we can talk about some of the scientists who believed that God played a role, or a creator played a role in creating the universe. Yeah, that's a good point, actually. A lot of renowned scientists, like for instance Albert Einstein himself, he came to know about Bhagavad Gita, a holy book in the Sanatana Dharma, towards the end of his life. So, he regretted that he did not know about this book earlier in his life. And similarly, Oppenheimer, the inventor of the nuclear bomb. So, he was also quite impressed with a lot of Sanskrit books. So, I think most of these scientists were aligned with the thought in these scriptures. And we call them scriptures, but it's like timeless wisdom. We don't necessarily have to call them scriptures, but basically timeless wisdom. They were all aligned with this timeless wisdom. A lot of things I see are aligned with the science today. For instance, this sustainability issue that we talked, the energy and the sustainability. And you also asked about the creation of the universe and things like that. There is a famous mantra, we can call it mantra in the Veda, called Purusha Suktam. So, that describes how the universe got generated and all that stuff. Of course, it takes the name of God. There is God, but the concept of God described there aligns more with the science. So, they say that the God started creating the universe by doing something called a sacrifice. So, there is no gain without pain. We need to sacrifice to get something. So, that is part of science. And that is what this Purusha Suktam says. And coming to the concept of sustainability, Purusha Suktam says, So, it means that God exists in both the animate and the inanimate objects. We know that God is source of life. So, when the source of life exists in animate and inanimate objects, like rocks, there is life in rocks also. That is what it kind of means. So, everything in this universe has a value. So, that is where the sustainability comes into picture. Just because they are rocks, they are just gases, does not mean, or water for that matter, does not mean that we can use them indiscriminately. So, sustainability comes into, I mean, kind of relates to that concept, that even the rocks and everything in this universe has life, which is what this Vedic mantra says. I want to talk to you about a topic that, maybe it is crazy, you know, but I had this thought about the concept of infinity. And the way this came to me was that, I interviewed actually a scientist from Stanford, Leonard Susskind, many years ago. He is considered the father of string theory. And one of the theories he has is that, we live in a universe that is a multiverse. There is an infinite number of universes. And it just happens that the one we are in, is the one that is perfectly designed, programmed for life to exist the way it is. Because, as you know, probably, there is different constants in the universe. If any one of those would be just a little bit off, the universe would not exist. It would just be empty space or rocks or whatever. So, the way it exists is just amazing, how everything fits in, right? Now, he bases this on the theory that everything is infinite, and then it can, one of them is this way. Now, what if the concept of infinity is not there, in the sense that there is no such thing as infinity? You know, we can argue that we can get to the smallest particle, that whatever it is, so then the distance between two points would be a finite number of points. The universe is finite in some way, and also, in math, of course, you would have infinity, because if you have one divided by zero, or those kind of things, you would have infinity. But, as far as tangible things, I wonder your thoughts about this concept, that if infinity was not there, what would this mean to science and to math? Yeah, that is a very interesting question. So, infinity is the fundamental concept around so many theories, and even spirituality, for that matter. So, I was talking about this, this scripture, Sanatana Dharma, there is this word called sahasra. Literally, it means thousand, but in concept, it refers to the infinite manifestation of the power, the creator and the power, and things like that. So, infinity is kind of coming from those, has some spiritual roots as well, but infinity itself is key to many aspects in math and science. So, for instance, the whole idea of asymptotic behavior, like approaching, which is very commonly used in calculus, in algorithm theory. So, that is entirely based on the concept of infinity. So, the very concept, because we were talking about the spirituality, that is also one of my favorite topics, is the very concept of God is infinity. God is always seen to be elusive. We can never reach him, according to the scriptures or the ancient wisdom, we can never reach God. It is always, the power is always elusive. So, that is also a concept, aligned with the concept of infinity. So, that is why I say that, a lot of what we think of religion today is aligned with science and math. So, the whole notion of asymptotic behavior will vanish without infinity. And, without asymptotic nature, I think a lot of theories will fall apart. May be, we can try setting apart infinity and revising the concepts of calculus, concepts of algorithm analysis and things like that, and see how it, and of course, physics, quantum mechanics and that kind of stuff, and see how it plays without the rule of infinity. I think that is a good research question, actually. Professor Bentayala, thank you so much for your time. Anything else you would like to add in conclusion? I think it is a great time to be in, with all the happenings in artificial intelligence and a lot of developments. The developments in artificial intelligence are bound to cause change in all other aspects of our living. Regulation aspects, cultural aspects, behavioral aspects. So, I think we need to carefully watch for them and contribute to them also. I mean, we do not want to be just mere spectators, but we actually want to contribute to the movement that this artificial intelligence revolution is bringing about. So, that is the biggest message I have for your audience. Thank you again for your time, and thank you for the variety of topics. Many times we think of a scientist just talking or knowing about science, yet we talk about so many other topics that you are interested in, and they are all connected. So, thank you for your time. Thank you for watching another episode of the Bay Area Innovators. If you have someone in mind who should be on our program, please send us an email at sfbay.epochtimes.com. We would love to hear from you. Thank you for watching, and see you next week.