For most of my adult life, whenever I had a question about the world or needed to track something online, I turned to Google for answers.
But lately, I’ve been venturing into Google with a new AI-powered search engine. (No, not Bing. Bing tried to destroy my marriage last year and is dead to me.)
It’s called perplexity. This year-old search engine is quickly becoming one of the most talked about products in the tech world, as its founder previously worked on his AI research at OpenAI and Meta. It’s now one.Technical personnel rave about it On social media, investors like Jeff Bezos, who was also an early investor in Google, have poured cash into the company. The company recently announced that it has raised $74 million in a funding round led by Institutional Venture Partners, valuing the company at $520 million.
Many startups have tried and failed to challenge Google over the years. (One of his potential competitors, Neeva, failed to gain traction and shut down last year.) But Google doesn’t seem so invincible these days. Many users have complained that their Google search results are clogged with spammy, low-quality websites, and some have started looking for answers in places like Reddit and TikTok instead.
Intrigued by the hype, I recently spent a few weeks using Perplexity as my default search engine on both desktop and mobile. I tested both the free version and the paid product, his Perplexity Pro. Perplexity Pro costs $20 per month and gives users access to more powerful AI models and certain features such as the ability to upload your own files.
After hundreds of searches, I found that Perplexity is not perfect, but it is very good. And while I’m not completely breaking up with Google, I now believe that an AI-powered search engine like Perplexity could loosen Google’s grip on the search market, or at least force it to catch up. I’m sure of it.
I also fear that AI search engines could destroy my job, and that products like AI search engines could disrupt the entire digital media industry. But I’m getting ahead of myself.
A shining place
At first glance, Perplexity’s desktop interface looks a lot like Google’s, with a text box centered on a sparse landing page.
But as soon as you start typing, the differences become obvious. When asked, Perplexity does not return a list of links. Instead, it searches the web and uses AI to create a summary of what it finds. These answers are annotated with links to the sources used by the AI, which also appear in the panel above the answer.
I write about current events (“What happened to Nikki Haley in the New Hampshire primary?”), shopping recommendations (“What’s the best dog food for senior dogs with joint pain?”), and household topics. We tested Perplexity against hundreds of queries, including questions. Task (“How long can beef stew be stored in the refrigerator?”).
Each time, I received an AI-generated response. Usually one or two paragraphs long, they are interspersed with quotes from his websites, including NPR, the New York Times, and Reddit, as well as a list of possible follow-up questions, such as: Ta. “Can I freeze beef stew to make it last longer?”
One of Perplexity’s impressive features is “Copilot,” which helps users narrow down their queries by asking clarifying questions. For example, when we asked for ideas on where to hold a 2-year-old’s birthday party, Copilot asked if we wanted suggestions for outdoor spaces, indoor spaces, or both. If you selected “indoor,” you were asked to select an approximate budget for your party. Only then did we have a list of potential venues.
Perplexity also allows users to search within a specific set of sources, such as academic papers, YouTube videos, and Reddit posts. I found this helpful when I was researching how to change the settings on my home’s water heater. (Exciting, I know.) A Google search turned up a bunch of not-so-helpful links to DIY tutorials, including thinly veiled ads for plumbing companies. I tried the same query on Perplexity and narrowed my search to her YouTube videos. Perplexity found the video you needed for your exact model of water heater, extracted the relevant information from the video, and turned it into step-by-step instructions.
Under the hood, Perplexity runs on OpenAI’s GPT-3.5 model and its own AI model, which is a variant of Meta’s open source Llama 2 model. Users who upgrade to the Pro version can choose from several different models, including the GPT-4 and Anthropic’s Claude. (He used GPT-4 for most of his searches, but he didn’t see much of a difference in the quality of the answers when choosing other models.)
Puzzlement is also refreshing and adept at admitting it. I don’t I know something. In some cases, my questions were answered with partial answers, along with notes such as “No further details will be displayed in search results.” Most AI chat products I’ve used lack this kind of humility. Their responses sound confident, even if they’re spouting nonsense.
Where Google still reigns supreme
During our testing, we found Perplexity to be most useful for complex or open-ended searches, such as summarizing recent news articles about a particular company or recommending restaurants for date night. I also found it useful when the information I was looking for (passport renewal instructions, for example) was buried in a crowded, hard-to-navigate website.
But I sneaked back to Google for some types of searches. Usually it was when I was looking up a specific person or trying to access a website I already knew existed. For example: If you typed “Wayback Machine” into your browser’s search bar, you would be redirected to Perplexity, which would spit out his one-paragraph essay on the history of the Internet Archive, the organization that maintains the Wayback Machine. I had to look for a small quote link to get to his website on the Wayback Machine, but that’s what I wanted in the first place.
A similar thing happened when I asked Perplexity for driving directions to a work meeting. Thanks to its integration with Google Maps, Google should be able to give you turn-by-turn directions from your home. But Perplexity doesn’t know where I live. So all Perplexity could provide me was a link to MapQuest.
Location data is just one of the many benefits Google has over Perplexity. The scale is different. Perplexity has just 41 employees and is based in a shared workspace in San Francisco, but it has 10 million monthly active users, an impressive number for a young startup, but far behind Google’s billions. It’s insignificant compared to.
Perplexity also doesn’t have a profitable business model. The site currently has no ads and fewer than 100,000 people have paid for the premium version, according to CEO Aravind Srinivas. (Srinivas did not rule out switching to an ad-based model in the future.) And of course, Perplexity offers Gmail, Google Chrome, Google Docs, or any of the dozens of other products. Is not … Google’s ecosystem is inevitable.
Srinivas said in an interview that he believes Google is a formidable competitor, but that a smaller, more focused startup could give it a jolt.
“What gives me confidence is the fact that if they want to do better than us, they basically have to abolish their own business model,” he said.
What about hallucinations?
One of the problems with AI-based search engines is that they tend to hallucinate, make up answers, and deviate from the source material. This issue has plagued several AI search hybrids, including Google’s early release of Bard, and remains one of the biggest barriers to mass adoption.
In my testing, I found Perplexity’s answers to be mostly accurate. More precisely, they were only as accurate as the sources they utilized.
I found some mistakes. When I asked Perplexity when Novak Djokovic’s next tennis match would be, he gave me details of matches he had already completed. Another time, he uploaded his PDF file of his new AI research paper and asked Perplexity to summarize it, and received an abstract of a completely different paper published three years earlier.
Srinivas acknowledged that even AI-powered search engines still make mistakes. He said that because Perplexity was a small and relatively unknown product, users did not expect it to be as authoritative as Google, and that Google needed to maintain a reputation for accuracy, so it was not possible for search engines to generate He said it would be difficult to incorporate AI.
“If you were using our product and it worked for eight out of 10 queries, you’d be impressed,” Srinivas said. “So let’s say you’re using a Google product, and it only gets a 7 out of 10. You think, ‘How can Google get that product?’ 3 queries are wrong? ”
“That asymmetry is our opportunity,” he added.
Gain for users, loss for publishers
I enjoyed using Perplexity and will continue to use it with Google, but after seeing news articles, product reviews, and primitive, concise summaries of articles, I felt a sick feeling in my stomach. I admit that. How-to article.
Much of today’s digital media economy still relies on a steady stream of clicks on links from Google and ads appearing on publisher websites.
But with Perplexity, you usually don’t need to visit the website at all. The AI will browse for you and display all the necessary information on the answer page.
The possibility that an AI-powered search engine could replace Google’s traffic or encourage Google to include similar features in its own search engine, as it has begun with its “Search Generation Experience” experiment. It’s one of the reasons many digital publishers are scared right now. This is also part of the reason why some companies, including The Times, which sued OpenAI and Microsoft for copyright infringement last year, are fighting back.
After using Perplexity and hearing about other startups developing similar products, I was convinced that the worriers had a point. If an AI search engine can reliably summarize what’s happening in Gaza or tell users which toaster to buy, why would anyone ever visit the publisher’s website again? Is it necessary? If AI search engines can just devour information and spit it out, why do journalists, bloggers, and product critics continue to publish their work online?
I raised these concerns with Mr. Srinivas, but he responded with diplomatic avoidance. He acknowledged that Perplexity will probably send less traffic to the website than traditional search engines. But he said the traffic that remains is higher quality and the result of more targeted queries, making it easier for publishers to monetize.
I’m skeptical of that argument, and I still worry about what the future holds for writers, publishers, and people using online media.
So for now, you must weigh the convenience of using Perplexity against the fear that by using it you may be contributing to your own destruction.
