Perplexity AI Reaches $520M Valuation, Raises $70M

News
January 4, 2024

Perplexity AI, an AI-Driven Search Engine, Achieves $520M Valuation, Secures $70M in Funding

In a digital landscape where giants like Google are supercharging their platforms with cutting-edge AI technology, a wave of startups is emerging, determined to revolutionize AI-powered search from scratch. This might look like an uphill battle, especially when competing against firms that boast billions of users. Yet, these innovative search startups are confident they can secure their own unique space by offering a distinctly better user experience.

Among these trailblazers is Perplexity AI. Just this morning, the company made headlines with its announcement of securing $70 million in a funding round. This round was led by IVP and received backing from a host of notable investors including NEA, Databricks Ventures, Elad Gil (former Twitter VP), Tobi Lutke (Shopify CEO), Nat Friedman (ex-GitHub CEO), and Guillermo Rauch (founder of Vercel). The round also saw participation from industry heavyweights like Nvidia and, quite remarkably, Jeff Bezos.

Informed sources indicate that the recent funding round places Perplexity’s valuation at $520 million post-money. This might seem modest in the expansive world of general AI startups, but it’s notably impressive considering Perplexity has been operational only since August 2022.

The brains behind Perplexity include founders Aravind Srinivas, Denis Yarats, Johnny Ho, and Andy Konwinski — a team with deep expertise in AI, distributed systems, search technologies, and databases. Srinivas, serving as the CEO, brings valuable experience from his time at OpenAI, where he focused on research in language and general AI models, including projects like Stable Diffusion and DALL-E 3.

Perplexity stands out from traditional search engines with its chatbot-style interface. This allows users to pose queries in natural language, on a wide array of topics, from “Do we burn calories while sleeping?” to “What’s the least visited country?” In response, Perplexity’s AI provides a concise summary complete with source citations, mostly drawing from websites and articles. This feature also enables users to ask follow-up questions, offering a more engaging and in-depth exploration of subjects.

“Perplexity is designed to provide immediate and comprehensive answers to any query, complete with thorough sources and citations,” explains Srinivas. “It’s a tool for anyone who relies on technology for information search.”

At the core of the Perplexity platform are various gen AI models, some developed internally and others by external partners. The Pro subscription, priced at $20 monthly, offers users the flexibility to choose from a range of models including Google’s Gemini, Mistra 7Bl, Anthropic’s Claude 2.1, and OpenAI’s GPT-4.

This plan also unlocks advanced features such as image generation and unlimited access to Perplexity’s Copilot, which tailors search results to individual user preferences. Additionally, Pro subscribers can upload files, including images and documents. The platform’s AI models can then analyze these uploads to provide detailed insights, such as summarizing specific pages of a document.

The features of Perplexity might remind you of Google’s Bard, Microsoft’s Copilot, and ChatGPT, and for good reason. Its chat-centric user interface echoes the design of today’s leading gen AI tools. Yet, Perplexity is not alone in this arena. The startup search engine You.com also provides AI-powered summarization and source-citing capabilities, with an option to use GPT-4.

However, Srinivas argues that Perplexity stands out through its enhanced search filtering and discovery features. For instance, it allows users to refine searches specifically to academic papers or explore trending search topics shared by other users on the platform. While I remain skeptical about the uniqueness of these features – wondering if they are truly distinctive enough to prevent replication or if they haven’t been replicated already – Perplexity has broader aspirations than just search functionality.

The company is starting to deploy its own gen AI models, which utilize both Perplexity’s search index and the broader public web to potentially enhance performance. These models are accessible to Pro subscribers through an API, suggesting a move towards offering more specialized, high-performance AI tools.

As a journalist, I harbor some doubts about the long-term viability of general AI search tools, primarily due to the significant operational costs associated with AI models. For instance, at one stage, OpenAI was reportedly spending around $700,000 daily to meet the demand for ChatGPT. Similarly, Microsoft is said to be incurring an average loss of $20 per user each month for its AI code generator.

Insiders report that Perplexity’s annual recurring revenue currently lies between $5 million and $10 million. On the surface, this appears quite robust. However, when considering the substantial investment required to train general AI models like those used by Perplexity, the financial health of such endeavors becomes less clear. The costs involved in developing and maintaining these advanced AI models can run into millions of dollars, posing a significant challenge to their long-term sustainability.

The advent of general AI search tools like Perplexity brings with it valid concerns about misuse and the spread of misinformation. These concerns are justified, as AI technology is not always perfect in summarizing information. It can occasionally overlook crucial details, misinterpret or exaggerate language, or even fabricate facts with seeming authority. Moreover, these models, including those used by Perplexity, have shown a tendency to propagate biases and toxic content.

Another challenge that Perplexity and similar platforms face is related to copyright issues. General AI models rely on “learning” from a vast array of examples – including essays, codes, emails, and articles – to develop their capabilities. Many providers, presumably including Perplexity, gather millions or even billions of these examples from the web to enhance their training datasets. While these companies often cite the doctrine of fair use to justify their web-scraping practices, this stance is not without controversy. Copyright holders, including artists and authors, have raised objections to this approach. The disagreement has escalated to the point where lawsuits have been filed, seeking compensation for the use of copyrighted material in AI training processes.

In an interesting aside, it’s worth noting that while many general AI vendors are introducing policies to shield their customers from intellectual property (IP) claims, Perplexity has chosen a different route. As per their terms of service, customers are required to indemnify, or “hold harmless,” Perplexity against any claims, damages, or liabilities that may arise from the use of its services. This effectively absolves Perplexity of any responsibility for legal costs related to such matters.

Concerns have been raised by entities like The New York Times, which argue that general AI search experiences could unfairly divert content, readers, and advertising revenue from publishers. They label these practices as “anticompetitive.” Whether or not this is the case, the technology is undeniably influencing web traffic patterns. Research by The Atlantic suggests that if a search engine like Google integrated AI into its search functionality, it could directly answer a user’s query 75% of the time, potentially eliminating the need for click-throughs to the website hosting the original content. While some companies, such as OpenAI, have established agreements with certain news publishers, most others, including Perplexity, have not pursued such partnerships.

Srinivas, however, views this as a positive aspect of the technology rather than a drawback. He suggests that this feature of AI search tools represents an evolution in how information is accessed and disseminated, aligning with the changing landscape of digital information consumption.

“With Perplexity, the need to click through multiple links, cross-reference answers, or endlessly hunt for information is eliminated,” he explained. “We are moving away from the era of navigating through SEO-driven content, sponsored links, and disparate sources. Perplexity is ushering in a more streamlined approach to acquiring and sharing knowledge, which promises to accelerate learning and research across society.”

Despite the various uncertainties surrounding the business model of Perplexity, as well as the broader landscape of general AI and consumer search, the startup’s investors remain undeterred. Perplexity, boasting ten million active monthly users, has successfully raised over $100 million to date. This funding is primarily being allocated to expand its 39-person team and enhance product functionality, according to Srinivas.

Cack Wilhelm, a general partner at IVP, shared via email her optimism about the company’s direction. “Perplexity is vigorously developing a product that will harness the potential of AI for billions of users,” she said. “Aravind has a rare talent for maintaining a grand, long-term vision while consistently delivering product advancements, a combination that is crucial for addressing a challenge as significant and foundational as search.”

Conclusion

Perplexity AI is making waves in the world of AI-powered search engines. With its innovative chatbot-like interface and a range of gen AI models, Perplexity offers users a superior search experience. The recent funding round, led by prominent investors, is a testament to the potential and value of Perplexity. As the company continues to grow and refine its technology, it is poised to carve out a niche in the competitive search engine market. Whether you’re a tech-savvy professional or simply someone who relies on technology for information, Perplexity AI is a promising tool that provides instant answers with full sources and citations. Stay tuned for more exciting developments from Perplexity AI.

Have you ever used an AI-powered search engine before? If so, what was your experience like? Would you be willing to try out Perplexity AI as an alternative to other popular search engines? Why or why not? Share your insights below.