Everyone wants better web search – is Perplexity's AI the answer?
'Conversational engine' still hallucinates, cites its sources at least
AI search engine startup Perplexity has raised $73.6 million in a series-B funding round led by Nvidia, Jeff Bezos, and other investors.
California-based Perplexity announced the cash infusion on Thursday, and is now said to be valued at $520 million. The year-old startup has raised around $100 million so far.
Perplexity bills itself as "the world's first fully functional conversational answer engine." It extracts information from the web using its own search, indexing, and crawling techniques, and uses a variety of large language models to generate concise, conversational responses quickly to users' questions. Perplexity cites sources and provides links to the sites it scraped information from in its answers – like you can do with, say, Google's Bard search bot.
The search tool is powered by a range of large language models built in-house and by other developers. Subscribers can pick from OpenAI's GPT-4, Anthropic's Claude 2.1, Google's Gemini, or the startup's own LLM Perplexity in Perplexity Pro, at a cost of $20 per month.
Perplexity has added other tools to improve its AI search – like Copilot, to add follow-up questions to fine-tune search inquiries. Users can click on the suggestions to minimize having to type in fresh prompts repeatedly. Subscribers can also keep and manage their search threads by topics, and can share them with others to collaborate with its Collections feature.
- Microsoft Bing Chat pushes malware via bad ads
- After failing at privacy, again, Google is working to keep Bard chats out of Search
- OpenAI reinstates ChatGPT's internet browsing privileges
- Google Bard can now tap into your Gmail, Docs, more
Dominated by Google, internet search is a tough industry to break into – several businesses have failed trying to change netizens' search habits. But at a time when people are clamoring for a return to better search results, large language models could provide an opportunity in the form of a novel way for people to recall information from the web.
Google was slow to roll out its AI web chatbot, ceding first-mover advantage to OpenAI and giving rival Microsoft's Bing a rare opening.
Founded in August 2022, Perplexity seems particularly promising and has reportedly raised the largest amount of money for any web search business in recent years. The startup has made between $5 million to $10 million in revenues so far from subscriptions, and from selling its AI software to other vendors. It may follow Google into the advertising business too in the future, CEO Aravind Srinivas told the Wall Street Journal.
Like all LLMs, however, Perplexity's product isn't foolproof and can still generate false information – known as "hallucinating" in AI parlance. Users should double check facts and inspect sources, especially if it gives contradictory responses, but hopefully the source citing will help with this.
"In an era where misinformation and AI hallucinations are causing increasing concern, we're built on the idea that accuracy and transparency are prerequisites to making AI-powered search ubiquitous," Srinivas explained on Thursday.
"The times of sifting through SEO spam, sponsored links, and multiple web pages will be replaced by a much more efficient way to consume and share information, propelling our society into a new era of accelerated learning and research."
The latest funding round was led by VC firm IVP and includes Nvidia; Jeff Bezos; Bessemer Ventures; co-founder and CEO of Bloom Institute of Technology, Austen Allred; ad the former CTO at Coinbase and ex-general partner at Andreessen Horowitz Balaji Srinivasan.
Interestingly enough, we see that Srinivas says he got a PhD in computer science at Berkeley in 2021, and within the past few years was an intern at Google and DeepMind, and a research scientist at OpenAI. ®