Berkeley Graduate Aims to Enhance Global Intelligence
Since the introduction of ChatGPT two years ago, individuals have been finding innovative ways to utilize generative AI chatbots for automating tasks and addressing queries. While these AI chatbots can be functional, their reliability as accurate information sources is sometimes questionable.
One notable endeavor in the realm of AI chatbots is led by Aravind Srinivas, who completed his Ph.D. in Computer Science at UC Berkeley in 2021. Together with Andrew Konwinski, he co-founded Perplexity in 2022, aiming to create a chatbot that functions like a virtual Ph.D. student. This chatbot not only generates text but also cites sources, maintaining a strong foundation based on peer-reviewed data.
Srinivas explains the essence of their project: “Ph.D. students write papers referencing previous work for every sentence to ensure their conclusions are scientifically supported. We applied that principle to develop Perplexity.” The goal is to minimize errors, often referred to as ‘hallucinations,’ where AI might produce incorrect or fabricated information. Perplexity is designed to provide answers that can always be backed by references or sources.
He describes Perplexity as a blend of Wikipedia and ChatGPT, suggesting that users can still enjoy a conversational interface while receiving structured responses akin to Wikipedia articles, complete with citations. The mission of Perplexity extends beyond just providing answers; it strives to enrich human curiosity and offer a learning experience. According to Srinivas, within thirty seconds of using Perplexity, users should come away with new information, aiming to cater to the collective curiosity of humanity.
Srinivas’s fascination with artificial intelligence began during his studies at IIT Madras in India, where he engaged in research about machine learning. His initial experiences included teaching AI to play vintage Atari games such as Pong and Breakout. He also explored concepts like transfer learning and hierarchical reinforcement learning. Following several published works, he interned at the Montreal Institute of Learning Algorithms, where he was introduced to deep learning by Turing Award winner Yoshua Bengio. He later pursued his studies at UC Berkeley because of its prominence in the field of AI, inspired by groundbreaking research from esteemed professors and fellow students.
His time at Berkeley was marked by intense challenges and significant learning. He appreciated the independence expected from Ph.D. candidates, who were responsible for all elements of research and development. Srinivas noted that seeing the dedication of his peers was instrumental in understanding the importance of software engineering alongside theoretical concepts. This practical experience formed the foundation of his subsequent journey in AI.
After his first year at Berkeley, he landed a summer internship at OpenAI, which he described as a pivotal experience. At OpenAI, he witnessed the early development of GPT technology, which reshaped his views and ambitions regarding AI.
Srinivas had the desire to establish a startup from the start of his time at Berkeley. While involved in research, he also wanted to enter the entrepreneurial realm but needed clarity on his direction. Many successful startups emerged from academic backgrounds, like Google, and he aimed to replicate this model by transforming academic insights into user-friendly products, leveraging user data to enhance the service further.
His inspiration was further fueled by popular media. The comedic TV series “Silicon Valley,” which centers around tech startups, resonated with him, especially its focus on generative models, leading him towards the field of generative AI without him initially realizing it.
Launching a startup proved to be a complex endeavor, requiring a convergence of the right idea, team, and funding. After his Ph.D. and a year of work at OpenAI, he focused on crystallizing his vision for Perplexity. He then reached out to potential investors and assembled a talented team.
His Ph.D. advisor, Pieter Abbeel, recognized Srinivas’s potential early on. He remarked, “Aravind not only executed well but also had a clear vision. This same talent is evident in his leadership of Perplexity, where he melds deep AI knowledge with innovative ideas.”
For aspiring engineers at Berkeley aiming to transition to entrepreneurship, Srinivas offers straightforward advice: begin with a clear vision. He emphasizes the significance of passion, stating that a genuine obsession is attractive to others and essential for attracting talent and funding. He advises future entrepreneurs to remain resilient, especially during uncertain times, urging them to persevere and keep pushing forward.
