The new talk of the internet and the search engines and LLM world is Perplixity, which is, in simple words, a combination of Google and ChatGPT. It is essentially a search engine that gives you an answer based on the latest Google search results, condensed into a single answer, along with links from sources of information. 

Founded by Aravind Srinivas in 2022, this startup has grown to 10M monthly active users after just a year of its launch in December 2022. Perplexity has raised $100M of capital from investors like Nvidea and Jeff Bezos and is valued at $520M. 

So let’s find out how an Indian entrepreneur and academician got inspired by Larry Page and Sergey Brin to build a $520M company that people are labeling as the “Google killer.”

Founding Story of Perplexity

Aravind Srinivas Academic Days 

Aravind Srinivas was a student at IIT Madras, which is amongst the most premium engineering universities. While pursuing a dual degree (B. Tech and M. Tech) in Electrical Engineering, his friend introduced him to machine learning and encouraged him to take part in a machine learning contest. 

Winning that contest sparked his interest in the field. Motivated by this success, he decided to pursue a Ph.D. in AI and deep learning at Berkeley. Aravind pursued his Doctor of Philosophy (Ph.D.) in Computer Science at the University of California, Berkeley, from 2021 onwards. 

During his research endeavors, he delved into various facets of computer science, focusing on Contrastive Learning for Computer Vision (CPCv2), Reinforcement Learning (CURL), Transformers for Image Generation (Flow++), Image Recognition (BoTNet), Video Generation (VideoGPT), and Reinforcement Learning (Decision Transformer). 

His academic journey also included a teaching role, where he imparted knowledge on Deep Unsupervised Learning during the Spring of 2020 and 2021.

Inspired by a TV Show, Larry Page and Sergey Brin To Pursue Entrepreneurship

Throughout his academic journey, Aravind was intrigued by entrepreneurship, an interest fueled by my exposure to Silicon Valley and the TV show "Silicon Valley." However, he struggled to find an academic-turned-entrepreneur as a role model. 

As the Silicon Valley bug took hold of Aravind, he was drawn to the energy of smaller teams and the audacity to pursue ambitious goals without the fear of failure. Inspired by the startup culture, he secured an internship at DeepMind, the top lab in 2019. 

Days were spent on launching training jobs, while evenings were dedicated to devouring books from DeepMind's impressive library. One particular book caught his attention – "How Google Works" by Eric Schmidt. 

Larry Page's forward, expressing a desire to either be a professor or start a company to execute ambitious goals, left a lasting impression on him.

Google DeepMind and Open AI Internships

Thereafter, Aravind completed the internship at Google and DeepMind for about a year and a half and then moved to OpenAI as a Research Scientist at OpenAI. 

However, his confidence took a hit during an internship at OpenAI in the summer of 2018. Despite excelling in India, he realized the vast skill gap when surrounded by exceptionally talented individuals. This experience became a crucial reality check, prompting him to focus on improving my programming skills and adopting a first-principles thinking approach.

During his internship at OpenAI, GPT-1 was unveiled, revealing a revolutionary form of learning through internet data. Recognizing its significance, he persuaded his advisor to shift their focus to this emerging field. They dedicated holidays and weekends to understanding and coding, embracing the challenge of mastering this new technology.

After two years of intense dedication, their efforts led to a breakthrough in combining generative AI and reinforcement learning. This laid the foundation for transformative technologies like ChatGPT, which goes beyond predicting the next word and emphasizes effective communication with humans.

Genesis of Perplexity

In envisioning artificial intelligence as the ultimate version of Google, he saw its potential to understand everything on the web, discern users' needs, and deliver precisely what they sought. 

This vision became his driving force, propelling him towards the intersection of academia and entrepreneurship ultimately contributing to the development of groundbreaking technologies in the field of artificial intelligence.

Transformers

Fascinated by the world of search and the potential for disruption, he reached out to the inventor of Transformers, Ashish Vaswani, a former Google researcher. Aravind’s vision was to integrate Transformers into search, allowing for a more nuanced understanding of language at different levels. 

Transformers, a deep learning architecture, had the power to process vast amounts of data and offer a comprehensive understanding of language.

Finding the company

The turning point came with the GitHub Copilot revolution. Witnessing its adoption and profitability, Srinivas felt the startup moment had arrived. Cold emails to investors like Nat Friedman and Elad Gill opened the door to initial funding. Yes, you read that right, COLD EMAILS!

Leaving OpenAI behind, Aravind, along with co-founder Johnny Ho (with whom Aravind connected by cold email), a former top-ranked competitive coder, embarked on the journey of building a company. Johny Ho (Chief Strategy Officer - Perplexity) and Denis Yaratas (now CTO of Perplexity) both worked at Quora, so they are already in the search engine background.

Their shared obsession with search fueled our efforts. Their initial team consisted of four or five individuals, inexperienced in running a company but determined to make it work. 

Initial Innovations

To streamline the early stages, the team created a bot using GPT-3.5 to answer our own questions, easing the burden of basic tasks.

All of that led them to conceptualize Twitter as a relational database, mining social graph data for meaningful insights. Building a search demo around Twitter, the founding team showcased it to potential investors, including prominent figures like Yan Liron, Andrej Karpathy, and Jeff Dean. The demo resonated with them, highlighting the potential of their approach.

Expansion & Launch

The initial team expanded to 30 individuals, with a mix of in-office and remote work culture. 

As we navigated the challenges of running a startup, Perplexity’s focus remained on search. 

The turning point came with the realization that their internal bot, powered by GPT-3.5, could be transformed into a public product. The launch of ChatGPT on November 30th, 2022, marked the beginning of their journey.

In September 2022, Perplexity received a $3.1M seed investment from an unknown source. 

They rolled out the company launch on December 7th, precisely a week after the initial announcement. The launch was a remarkable experience, unlike typical AI launches that involve flashy demo videos, sign-ups, and long waiting lists. The company decided to take a contrarian approach, presenting users with a straightforward search bar right from the start. No sign-up, no waiting list – just a landing page with the simplicity of perplexity encapsulated in a search bar.

The product garnered attention during the Christmas break, with people appreciating its simplicity and utility. 

Jack Dorsey's tweet endorsing their Twitter search feature brought unexpected virality. However, due to Twitter API rule changes, Perplexity had to shut down that particular feature. 

Raising Money

Despite the initial hype, their sustained usage indicated genuine interest. The company observed the product's trajectory, following Nat Friedman's belief that a good product should have initial wow moments, sustained usage, and continuous innovation. 

And yes, they had all of that. Thereafter, the users of Perplexity started going up and up, as the team gradually developed the product. Eventually, it raised $25.6M in Series A in 2023 and $73.6M in Series B rounds in January 2024 from renowned tech companies and entrepreneurs like Databricks, Jeff Bezos, Nvidea, and more.

What is Perplexity, and how does it work? 

Perplexity is a “conversational search engine” based on AI and NLP text prediction algorithms. It uses GPT3.5’s LLM and combines it with its NLP-based LLM.

Using Perplexity is very easy. Anyone with first-hand experience using any search engine will get the hang of it. You get the following search bar where you can type your query and press enter to let it search. 

Based on your search, it gives you results from all the sources it can gather on the web. Clicking on ‘Co-Pilot’ can increase the number of results. As you can see, it gives you about 25 sources/links, which is pretty decent. 

Is Perplexity the Chat GPT and Google Killer?

Some people are calling it the “Google killer,” and while it may take some of the limelight, it’s not exactly a direct competitor as Google serves a different purpose. Moreover, combining Google with a GPT extension called WebChatGPT can also give GPT’s LLM-based search results summary but is not conversational. 

However, with WebChatGPT or any other Web access extensions, Chat GPT can very closely resemble Perplexity. 

So, as you can see, ChatGPT with an extension can also do the same thing, but its processing is slower. 

So, to answer the question, Perplexity is a great tool, but it may not be the ChatGPT or Google killer. I believe people are going to use all three tools simultaneously. Moreover, it’s difficult to beat Google’s search engine algorithm, and Google can also come up with a similar conversation search engine. However, that will significantly bring down Google’s Ad Revenue, which is a tricky space for a big corporation like Google to be in.  

Perplexity v/s Chat GPT v/s Google: In-depth Comparison

Conclusion

Perplexity has a huge element of virality and is growing enormously, handling 3M daily searches. Without any marketing and PR, in just one and a half years, it has grown to 10M monthly active users and a $520M company. 

It’s not profitable yet but already has 10,000+ monthly users subscribed to its Pro model. Moreover, they have recently collaborated with AI smartphone company Rabbit on their latest gadget, called R1. 

While it still has to take longer leaps to take over Chat GPT and Google, it definitely will give them hard competition. Good for us consumers! We have some pretty cool tools in our hands. This AI race keeps getting more and more interesting!!!

‍