Sorry, fellas: your spicy AI anime girlfriend is actually destroying your mind — at least according to Perplexity CEO Aravind Srinivas.
During a fireside chat at the University of Chicago, covered by Business Insider, Srinivas charged that the huge rise in popularity of AI-powered companion chatbots is “dangerous.” The tech CEO fretted that the AI bots — which are designed to mimic doting lovers over text or voice chat — are becoming more sophisticated and human-like, with abilities like remembering intimate details about their users.
“That’s dangerous by itself,” Srinivas said. “Many people feel real life is more boring than these things and spend hours and hours of time. You live in a different reality, almost altogether, and your mind is manipulable very easily.”
Like any good entrepreneur, Srinivas isn’t just interested in analyzing the problem. He’s also selling a solution: his own software.
“We can fight that, through trustworthy sources, real-time content,” the CEO said of users losing themselves to AI companionship. “We don’t have any of those issues with Perplexity because our focus is purely on just answering questions that are accurate, and have sources.”
Though vulnerable people are definitely falling into dangerous rabbit holes with their AI lovers, the anxious comments are a little rich coming from a CEO whose own company is in the business of pedaling AI-powered search engines that work like a chatbot.
Sure, Perplexity isn’t selling a raunchy companion bot to satiate lonely gooners, or even a conversational encyclopedia like ChatGPT. But when you peel away the outer layers, Srinivas’ business model is no different than the others: providing a machine-learning solution, built on huge troves of unethically scraped data, for users trying to fill some kind of void. For companionship chatbots, that void is human connection; for Perplexity, it’s a functional web browser.
Claims of superior accuracy are a standard boast in the tech industry. Beyond Srinivas, Elon Musk has positioned xAI’s Grok as a “maximum truth-seeking AI,” and Anthropic’s Claude is explicitly marketed as a “constitutional AI,” designed to be fundamentally “helpful, honest, and harmless.”
In reality, these AI systems fall far short of their CEO’s claims. Over the summer, Grok suffered an infamous meltdown where it began spewing racist slurs, while Claude has a history of veering off onto tangents that are totally unrelated to its assigned tasks.
Perplexity also has a tendency to spew hallucinations, like when it was caught inaccurately summarizing reporting from reputable journalists to its users, which has landed it in hot water with publishers.
Ultimately, Srinivas isn’t diagnosing a societal ill as much as he is writing a script for his own product — and ignoring the root causes of the rise in AI companionship in the first place.
More on Perplexity: AI Browser Will Track Every Single Thing You Do, CEO Reveals
