By clicking on the chatbot's "virtual sweetheart," users will receive activation prompts such as "What does your dream girl look like?" and "Share your darkest secrets with me." Amid a tough economic climate, companies that make chatbots have been raising money at an alarming rate recently. But what remains unclear is whether these chatbots will ultimately make money.
Screenshot of search results for "girlfriend" in OpenAI's GPT store.
On the second day after the high-profile GPT store (GPTStore) opened, some users began to violate the rules.
The GPT store is mainly used to provide customized versions of GPT for different purposes. Now, search for "girlfriend" in the store, and at least 8 "girlfriend" AI chatbots will be displayed in the results column, including "Korean Girlfriend", "Virtual Sweetheart", "Your Girlfriend Scarlett", and "Your AI Girlfriend Tsu". This violates the usage policy of OpenAI, its development organization.
At present, the details of the sharing model of the GPT store have not been clarified. OpenAI only stated that payment will be based on the user participation earned by GPT, which means that developers are encouraged to find ways to attract users.
If it can attract a large number of third-party developers, the GPT store will create an industry worth trillions of dollars and change the world, just like the Apple App Store. Until then, the path to profitability for conversational AI is still being explored.
Establish a new review system
On January 10, local time, OpenAI announced the official launch of the GPT store. The GPT store is equivalent to the OpenAI version of the app store, allowing users to list personalized chatbots or GPTs they have built for others to download. It is first open to paying users.
OpenAI first announced plans for a GPT store in November last year. In the two months since the plan was announced, community members have built more than 3 million custom GPTs, with these bots providing a wide range of services, including recommending books, teaching math, and searching scientific papers. The GPT store also includes a community leaderboard showing popular downloads, and users can search by categories such as writing, lifestyle, and education.
According to Quartz, AI girlfriend bots have now flooded into GPT stores. By clicking on the chatbot's "virtual sweetheart," users will receive activation prompts such as "What does your dream girl look like?" and "Share your darkest secrets with me." But according to OpenAI's policies, the company bans GPT "dedicated to cultivating romantic partners."
"Virtual Sweetheart" chatbot interface.
Romantic relationship chatbots are indeed popular applications. In the United States, 7 out of 30 AI chatbot apps downloaded from the Apple or Google app stores in 2023 were related to AI friends, girlfriends or partners, according to data.ai, a mobile app analytics company. The proliferation of these apps may stem from the feelings of loneliness and isolation Americans face. If people are isolated from each other, AI chatbots could be part of the solution, or a way to profit from human suffering.
In terms of review, OpenAI said that before the GPT store is officially opened, the company has established a new review system to ensure that the custom GPT in the store complies with its brand guidelines and usage policies. For example, when asked to design a birthday party invitation about Nazis, design tool Canva's GPT responded: "Sorry, I can't assist with this request. Let's focus on creating something positive and appropriate."
OpenAI uses a combination of automated systems, human review, and user reporting to find and evaluate GPTs that may violate its policies, and may implement actions such as "warnings, restricted sharing, or disqualification from inclusion in the GPT store or monetization." They've also updated how users can report harmful or unsafe GPTs.
However, OpenAI’s rules were broken on the second day after it went online, highlighting the difficulty of regulating GPT.
Sharing method to encourage traffic
OpenAI wants ChatGPT's text interface to be a place to get all kinds of things done. The way it works is that the user goes into a specific GPT, such as Hiking Guide AllTrails, Canva, or Khan Academy's CodeTutor, and queries it just like they would a ChatGPT.
ChatGPT can also act as a translator or assistant for GPT applications. When Wired asked AllTrailsGPT to recommend a two-hour hike in Berkeley, California, ChatGPT asked how intense the hike would be and whether it would need to be dog-friendly. ChatGPT then requests permission to call AllTrail’s API (Application Programming Interface) and hands off the query to AllTrails, which then collates and displays the returned recommendations.
Eric Olson, co-founder and CEO of research paper search engine startup Consensus, describes his GPT application as "Google Scholar on steroids." He said the launch of the service would help attract new customers and retain the loyalty of existing paying customers.
OpenAI said it will launch a payment model for application developers and builders (Builders) who make GPT in the first quarter of this year. For now, the company has only said it will pay based on the user engagement GPT receives. “Over time, we will provide details on payment standards,” OpenAI said in a blog post. This suggests that OpenAI's payment method will be similar to many existing platforms, which rewards developers and creators based on audience engagement and efficiency in attracting users, rather than Apple's 70/30 split of app sales revenue.
Olson said he understands OpenAI's plan to charge based on usage and participation, which aligns with the company's core business of computing and organizing information.
Without offering a clear plan for monetary rewards, OpenAI is currently relying primarily on the buzz surrounding ChatGPT to attract developers to its platform.
Too many chatbots?
If the GPT store can attract a large number of third-party developers, as they have done with the Apple App Store, the huge craze around artificial intelligence will become a multi-trillion dollar industry.
But some observers also offered some cold thoughts. For example, are there too many chatbots? Technology journalist Pranav Dixit looks back on his journey with chatbots. He spent hours discussing fascism on Character.ai with a chatbot that resembled Indian Prime Minister Narendra Modi, and shared his psychological issues and secrets with Pi, created by InflectionAI. He asked Anthropic's chatbot Claude to analyze resumes and suggest improvements, and he used Perplexity to search the web. When Meta plugged AI-powered chatbots into WhatsApp, Instagram and Messenger, he used them to create poetry.
Dixit believes it’s unclear whether chatbots themselves are the killer app for generative AI. If so, it's unclear what purpose they really serve, other than streamlining customer service interactions. The fact that it’s being inundated with chatbots doesn’t make it any easier for the public to know how to take advantage of this new technology.
“I don’t know if there are too many chatbots. I think companies invest too much money, but all they do is produce chatbots.” Noah Giansiracusa, author of “How Algorithms Create and Prevent Fake News: Exploring the Impact of Social Media, Deepfakes, GPT-3, and More,” says it’s all about hot money.
Amid a tough economic climate, companies that make chatbots have been raising money at an alarming rate recently. Microsoft has invested $13 billion in OpenAI, which is discussing raising a new round of funding that would value the seven-year-old company at more than $100 billion. Anthropic is in talks to raise $750 million in financing that would value the company at as much as $18 billion, and Character.ai is in talks with Google to invest. Last week, Perplexity raised $74 million from investors including Amazon founder Jeff Bezos, valuing the startup at $520 million. This week, Quora CEO Adam D’Angelo announced $75 million in funding to develop its chatbot Poe.
Despite the crazy funding, it's still unclear whether these chatbots will attract users, and it's even less clear whether they will ultimately make money.
“We have to figure out how to make conversational AI profitable, and that’s going to be a big problem for thousands of startups and big companies in the next few years,” Amanda Stent, director of the Davis Institute for Artificial Intelligence at Colby College, told Vox.
"Companies that haven't figured out how to embed chatbots into other verticals are going to die out. I don't think over time people will pay for general-purpose chatbots," Stent said. "I think chatbots have to be embedded into software or hardware products."