How AI will answer questions we haven't thought to ask | Aravind Srinivas

0:06You're listening to TED Talks Daily, where we bring you new ideas to spark your curiosity every day. I'm your host, Elise Hu. Aravind Sreenivas is the CEO of Perplexity, a search engine that answers your questions by using the power of AI. In his 2024 talk, he explains why today's AI revolution may actually be a revolution in enhancing the very human trait of curiosity and questioning.

0:36There are a couple of ways I'm not a traditional tech founder. I never dropped out of college. In fact, I kept going. I'm an academic, you could say, and it's okay to be proud that I have a PhD in AI from Berkeley right here in the Bay Area.

1:02But there's something interesting in AI that I've noticed compared to other tech founders. Or they're stereotypes, at least. A lot of us hold PhDs. I mean, quite a lot. 11 out of 24 speakers just at this conference have PhDs, and over a third are assistant associate or full professors with major universities. Only time will tell if this is a new trend of seeing academics in technology startups. But I got pretty curious to find out if this is common or new. And it turns out this is somewhat new. Only over a year ago, the researchers at the University of Maryland found a 38 % decline at the rate of startup formation or share of employment by U.S.

1:55PhDs over the past 20 years. Yet, our attendance here today and the trend in AI technology broadly does not seem to correlate with this finding. As I said, only time and more data will tell. In the meantime, my curiosity led me to another question. What was the last major technology company founded by academics? Google. At Perplexity, we get accused of trying to kill Google a lot. But trust me, we're not really trying to kill things. We're motivated about building things. The co-founders of Google would probably say the same. Let's hear from Larry Page, an interview of his from the year 2000. Artificial intelligence would be the ultimate version of Google.

2:49So if we had the ultimate search engine, it would understand everything on the web. It would understand exactly what you wanted. And it would give you the right thing. and that's obviously artificial intelligence. You'd be able to answer any question basically because almost everything is on the web, right? Think about that. Artificial intelligence in the year 2000. I was only six back then. There are a few things interesting about this interview. One, Larry did accurately predict the future of search almost 25 years ago. The future of search is artificial intelligence. That's why I'm here, and we're going to talk more about it.

3:34Second, it's very interesting how a common theme in interviews like those or events like these is us thinking about the future. What is the future of search? What is the future of technology? What is the future of AI? I'm sure a lot of you have lots of thoughts about these questions. In some sense, that is the purpose of technology, to keep us thinking and to keep us evolving. But people like Larry or people like you or people like me, we are not building technology in a vacuum. We are building technology for us, the people. We are the people. So when we come here to think about what is the future of technology or what is the future of AI, let's ask ourselves this question.

4:17What is the future of us, the people? I believe that AI will make us even more human. Socrates, the Greek philosopher, is famous for saying that wisdom comes from realizing how little we know. Or that progress can only be made by asking better questions. The Socratic method is essentially the practice of relentless questioning. Relentless questioning is something academics do all the time. It has been core to the progress of human intellect over the past thousand years. Relentless questioning is also a practice that can be done orders of magnitude better with the power of AI. And by the way, relentless questioning is something South Indian parents do when you tell them you're leaving a good university or a stable job to go join a startup.

5:18So, jokes aside, relentless questioning is something fundamentally human. The physicist David Deutsch has proposed that we humans are the only species who have curiosity for what is already familiar. We can know so much about the stars above us or the machines in front of us and yet continue to have more questions about them. It seems like for humans, every answer leads to a new set of questions, questions that we haven't even asked before. That, to me, is what the future of technology should be about. And it's also how perplexity was born. I was raised as an academic in the comforting arms of universities.

6:03So when I actually entered the real world and tried to do my own company, I had an endless set of questions. SPV, safe notes, health insurance. I needed to figure all these things out. And all these required to do a lot of research. I needed actual answers. And traditional search engines left me lost. There was a ton of information and very little time to evaluate any of it. And neither did I have access to all of the experts on all these topics. So I was actually truly in a state of perplexity. so that's when I thought maybe I could have an AI do this for me maybe I could go ask an AI all these questions if it was able to pull information from the web and answer all my questions so my co-founders and I came together and we built a slack bot where we could just ask our own questions once we began using it is when we realized what we built was much bigger than ourselves For the first time, I had the ability to go ask whatever question I wanted about any topic, no matter my level of expertise in it, and get a well-researched answer from the web.

7:14And it's not just about an answer, it's an answer that I can actually trust. In this case, every answer in perplexity comes with sources from the web, in the form of citations, just like academics cite their sources. Now, this is pretty powerful because trust is not unique to animals or humans, but it empowers us pretty differently. In the case of humans, an answer you could trust allows you to ask better follow-up questions. More questions lead to more knowledge. That's the point of ensuring that you could always get an answer with well-sighted sources. And in perplexity, ever since the beginning, every answer has always come with sources that allows you to ask more questions.

8:00In my case, once I ask questions about safe notes or insurance, I ask more questions. What areas outside of insurance could I benefit from having access to better answers? Who else in the world benefits from having access to better answers? Now, the answer is basically all of us. Every single person benefits from having access to better answers. This is such a profound shift in human history. Until recently, if you wanted the best answers, you had to be someone who could afford it. You had to be someone who had access to the greatest minds in the world or the best materials, libraries, expertise.

8:42And now that's changing. If a major achievement of the Internet was to give everyone access to all of the world's information, a major achievement of AI would be to give everyone access to all of the world's answers. It doesn't matter if you're a Harvard professor or an underserved student in a developing nation. We all get access to the same answers. With AI that keeps getting better and better at answering all our questions, the marginal cost of research is rapidly approaching zero. In that new era of humanity that AI is powering, knowledge does not really care about who you are, where you're from, or who you have access to.

9:28Rather, what matters is the next question you're going to ask. When all of the world's answers are available to all of the world's people, one can only wonder, what will the best questions be? and how many such questions will get asked. This is again where David Deutsch argues that human potential is infinite. As long as we keep engaging in relentless questioning and keep asking interesting set of questions, sky's the limit in terms of what we can actually learn. For example, humans are always curious. You can see that in babies. Even before they learn to crawl, they're pretty curious about what's around them.

10:12That's a natural trait for all of us. Take an example for the technologies that we're building. In the case of BART, that became perplexity. Once I got answers to something like health insurance, I could ask an infinite set of new questions, ranging from very pointed ones, like what are concrete ways to improve the healthcare and insurance industry, to very broad ones, like who else would benefit from having access to such a technology. It seems to a curious species, every question and answer that you get is a lead to the next set of questions and spawns several paths of curiosity more than any one person can keep track of.

10:57So when we are here to wonder about what is the future of technology or what is the future of AI, we are merely talking about the outputs. the outputs of a much bigger question, what is the future of human curiosity? It is my strong belief that in an age where AI gets better and better at answering all our questions, this human quality that makes us so human will become even more essential, our innate curiosity and our relentless questioning. With all of the world's answers available to us, the tools we use to ask our questions, and the stuff that we build using those answers, those to me are the future of our technology.

11:42And more importantly, that is the future of us, the future of humans. We are all curious, and when we are curious, we want answers. We really do. But what we really want are those answers that lead us to the next set of questions. And I, for one, can't wait to see what you'll ask next. Thank you.

12:29team, Martha Estefanos, Oliver Friedman, Brian Green, Lucy Little, Alejandra Salazar, and Tonsika Sarmanivon. It was mixed by Christopher Faze-Bogan. Additional support from Emma Taubner and Daniela Ballarezzo. I'm Elise Hu. I'll be back tomorrow with a fresh idea for your feed. Thanks for listening.