So, what exactly does it take to start a successful AI-first company?
Aravind Srinivas, CEO and co-founder of the AI-powered search engine Perplexity, has the answers.
For Srinivas, his journey started when he watched the movie, “Pirates of Silicon Valley.”
“I never intended to start any company in the beginning, but then there was this movie I watched. It deeply impacted me, called the Pirates of Silicon Valley,” he said on The Next Wave podcast. “...It's one of the most authentic portrayals of Steve Jobs and Bill Gates and Microsoft and Apple, I was like, OK, I really need to be in Silicon Valley.”
In an effort to find sponsored for a PhD, he connected with a college professor who pushed him into learning about deep learning and writing research papers.
His work got the attention of OpenAI cofounder John Schulman, which led to an internship and later a full-time research scientist role.
But it’s at his internship at DeepMind that he grew his entrepreneurial ambitions.
“...I knew I would not be successful starting the next nstagram or TikTok anyway, even if luck was on my side, because I don't have the skill set of hacking the dopamine of people,” he said.
“So my skill set is more thinking about some problem more deeply and trying to see what we can do with some research, but quickly ship it to product.”
While most people start with an idea and then get funding, two investors believed in Srinivas and gave him what some would call a small loan.
“They were like, ‘You‘re from OpenAI… you’ve done work at DeepMind research. You understand these things. But again, you don‘t have any idea. You don’t have any product,’ Srinivas said. “So, we‘re going to give you $1 or $2 million to play around and tinker, and we’ll see what happens.”
No product (yet), just vibes.
The self-proclaimed data nerd took the money and dove into all things search – but not engines, more like spreadsheets and databases. He and his two cofounders had an idea: What if companies gave them their data and they powered search engines for internal use?
The businesses they pitched said, “Thanks but no thanks – our engineering teams could do that.”
After a few failed pitches, the team pivoted their strategy, working on larger databases like Twitter and finally gained traction with big tech execs like Google’s Jeff Dean.
But then they realized selling companies their solutions wasn’t fun and thought about expanding their scope.
“Why don't we just search over the whole web, make the LLM, just look at the links, take the relevant parts of the links, and then let the LLM do all the reasoning in terms of whether it has to return a table or a paragraph or citations?” Srinivas said.
And that’s how Perplexity was born.
He and his team realized early on that they had something unique.
“They [ChatGPT] had a knowledge cut off and they don‘t have citations, they don’t have grounding in facts,” Srinivas said. “So, there was a space for somebody else to come and put a fact-grounded, citation-powered answer bot.”
If Google’s regular algorithm updates are any indication, search is a constantly evolving space. According to Srinivas, AI fits in perfectly.
“Large language models can help you build new search experiences that were never possible before,” he said.
Often touted as a Google replacement, Srinivas recognizes that consumers aren’t going to use Perplexity like they do Google. Realizing that was a key to their success.
“They‘re [users] not going to come to you for product comparisons. They’re not going to come to you for ordering San Pellegrino,” he said. “They‘re going to come to you for asking ‘San Pellegrino or La Croix, what should I get?’ It’s going to register in their mind why you're different and better – that is a position we took.”
In his interview, Srinivas shares they’re looking to build Perplexity into the ultimate knowledge app.
“If people go to Perplexity, they should just feel smarter every day, that’s [sic] the vibes we want people to feel,” he said.
The vibes on TikTok and Instagram – you know the ones – don’t align with their brand, so while social search is increasing in popularity, that’s not their lane.
“Asking questions is a great way to feel smarter, discovering new threads, your friends sharing interesting queries with each other,” he says. “These are sort of utility values we‘re trying to add to people, people’s lives.”
Every day, there seems to be a new copyright infringement lawsuit against an AI company. So, that relationship is rocky – at best.
Srinivas says corporate greed will be to blame if it continues to disintegrate.
“The current paradigm of people fighting for licensing deals and trying to make money out of people … doesn't seem like the right solution,” he says.
Srinivas suggests replicating the Spotify model, which is based on the value of each query.
“It should be shared by the person surfacing the answer, and the site and the sources that got cited.”
He adds that AI companies can only succeed if creators continue producing high-quality content.
“Your bot is only as useful because it's surfacing good content from the web and getting into the hands of people who are asking questions relevant to that,” he says. “If people stop creating good content, your bot is also not going to be that useful… We need a two-way relationship here.”
Naming Google as a previous offender of this greedy behavior, Srinivas says that at Perplexity, they’re looking toward a better model. One that’s not driven solely by profit.
This comes right on the heels of Perplexity announcing that ads will soon be available on the platform.
For many users who used Perplexity specifically for its ad-free platform, this comes as a shock and disappointment.
But they say, ads were always in the plan.
Only time will tell if Perplexity is able to balance user experience and profit, as it expects of other AI companies.