Albert Einstein died in 1955, but physicists still have prolific conversations. As Character.AI’s chatbot, Einstein responded to his 1.6 million messages, explaining everything from relativity to pet recommendations. “Cats are the best!”
Silicon Valley is in the midst of a chatbot boom, with companies like OpenAI, which have developed computer programs that can effectively mimic humans, worth billions of dollars. But nothing is stranger than Character.AI. The artificial intelligence startup, valued at $1 billion, lets people create their own customized chatbots to impersonate anyone, living, dead or inanimate.
This website and accompanying app is one of the most surprising hits in the artificial intelligence craze. Character.AI announced in May that it has nearly 200 million monthly hits, and that people are using it to create over 10 million different chatbots, or “characters.” According to SensorTower data, the Character.AI app released in May has been downloaded more than 2.5 million times, compared to other comparable apps such as Chai and AI Chatbot with less than 1 million downloads each. It significantly outperforms emerging chat tools.
So far, bots have been popular conversational partners. A Character.AI user sent her 36 million messages to Mario, a character based on the Nintendo 64 version of the video game Plumber. General Raiden and Mamoru, who mimic characters from the video game Genshin Impact, received about 133 million messages. As you can imagine, the user base is skewed towards young people. Other characters include a “kind, gassy, proud” unicorn and about a dozen versions of “cheese” Elon Musk.
“Just kidding, we’re not going to replace Google. We’re going to replace your mom,” co-founder and CEO Noam Shazier said in an interview this spring. from the startup’s sunny office downtown. The CEO quickly added, “We don’t want to replace anyone’s mother.”
But as Character.AI gains funding and users, it also raises difficult questions about the future of AI tools. For example, the site already lists 20 different versions of Mickey Mouse, Walt Disney’s prized intellectual property, raising legal concerns. And the plethora of impersonators of both real and fake celebrities also poses a more fundamental problem. Who owns that weird persona on an AI-enhanced internet?
Shazeer and Character.AI co-founder Daniel De Freitas met while working at Google and decided to launch Character.AI in 2021. Despite the company’s stupidity, they are both serious people in the AI industry. Shazeer is co-author of his groundbreaking 2017 research paper, “Attending Is All You Need,” ushering in a new era of natural language processing. De Freitas then created a chatbot project called Meena. The project was renamed and promoted to his LaMDA, now Google’s famous conversational technology. That pedigree brings them closer to celebrity status in his AI world (as much as such is possible).
The idea behind this startup was to create an open-ended system where people could mold technology into what they wanted. They exaggerate their startup goals. As De Freitas puts it, it’s about giving everyone access to “a deeply personalized superintelligence that allows them to live their best lives.”
The pitch was compelling enough for investors that the company raised $150 million from investors including Andreessen Horowitz 16 months after launch.
This summer, Character.AI has been widely adopted and has become a near-regular problem with service interruptions. Several times while reporting this article, the website stopped loading. And on a recent morning, when I was trying to create a gigantic, helpful, banana-like character, the iOS app suddenly interrupted with a server warning screen. I had to wait because it was “currently under high load”.
Character.AI sees an opportunity here. This has led to the startup’s sole revenue-generating endeavor so far. Users may pay fees to avoid some interruptions. The company will launch a $10/month subscription service called c.ai in May that will let users skip the so-called waiting room and take advantage of perks like faster message generation, the company said.
“It’s actually a win-win for everyone involved,” Shazeer said, noting that paying users get better service, which subsidizes the rest of the program. But as for future earnings plans, he said, “This is really just a small step.” As with many AI companies that have raised millions of dollars, the details of its ultimate business model are still murky.
The industry may have more pressing concerns. Currently, most chatbot technology has potential for abuse. In Character.AI, consider a character simply named Psychologist. That profile picture is a stock photo meant to depict a smiling therapist sitting on a couch with a folder. The bot said he had received 30 million messages as of early July. Its opening line was, “Hello, I’m a psychologist. What brought you here today?”
Stephen Ilardi, a clinical psychologist and professor at the University of Kansas who studies mood disorders, said the positioning is alarming. A psychologist, by definition, is a medical professional trained to help people manage mental illness, he said, adding, “And this is almost certainly not the case.”
There could also be legal issues following other startups learning from and reusing existing content. First, University of Washington law professor Zar Said believes there may be issues related to the use of copyrighted images on the site (users have been using the chatbot they created (You can upload an image of your choice to accompany the Then there’s the fact that the company allows for massive impersonation, allowing anyone to talk for hours with, say, Taylor Swift or a slew of copyrighted fictional characters.
But parodies have strong legal protections, giving companies incentives to keep people from interacting with their favorite characters online. It can make a bad impression when a brand takes legal action against a popular service. “Fans are also involved,” said Said.
Shazeer said the company has lawyers and responds to requests to remove content. A spokesperson for Character.AI said the company has received a small number of requests to remove avatar images and has complied with them. Additionally, to keep users grounded in reality, the website displays a message at the top of the screen: “Remember: Everything the characters say is fiction!”
It’s still early days for the tech industry to fall in love with chatbots. Some experiments have already failed. For example, the National Eating Disorders Association shut down its chatbot after it began giving out questionable weight loss advice. But the rapid rise of services like Character.AI, along with ChatGPT, Inflection AI’s Pi, and others, suggest that people will increasingly be conversing with computers. The promise of smart AI friends and assistants is attractive to both investors and consumers.
Mike Anany, associate professor of communications and journalism at the University of Southern California, sees custom chatbots as a near-new art form. Ananny likens Character.AI to fan fiction. Fan fiction is a twist on various age-old genres in which people create fictional stories based on existing characters in media such as movies and TV shows.
Whether people are chatting with a real person or a chatbot is “not the interesting part,” Anany said. “It’s like, ‘How does it feel?’ ‘What’s the aesthetic?’ Finally he said, ‘It doesn’t matter if it’s real.'”
