The AI startup Character AI recently made headlines with news of financial crisis and stagnant user growth. The company is now seeking acquisition and reportedly in talks with Google, Meta, and Elon Musk’s startup xAI.
What led a unicorn that reached $1 billion valuation in just two years to face bankruptcy risk? In this article, we will analyze the reasons behind this from perspectives such as the company’s product design, corporate philosophy, and founder backgrounds.
What is Character AI?
Character AI is an AI startup founded in 2022 by former Google AI department employees. The company created a website, c.ai, for AI roleplay, allowing users to create various AI characters and engage in entertaining chats with these chatbots.
The main attraction of their product (Character.ai) includes a wide range of well-known characters such as business celebrities, anime characters, and movie stars. The autonomous LLM developed by the company powers the product, enabling highly realistic responses that have garnered substantial user appeal.
Why is Character AI facing funding difficulties?
Previously, Character AI secured a $160 million investment from A16Z, propelling it to a unicorn valuation of $1 billion. However, it is now facing funding challenges and the risk of financial collapse, necessitating its pursuit of acquisition. Behind all this, I see contradictions between Character AI’s team, product design, and corporate philosophy. The current situation for c.ai could have been foreseen.
“We have the strictest anti-NSFW policy.”
A spokesperson for Character AI stated in a release, “Among all consumer AI platforms, we have the strictest anti-NSFW policy.” This stance proves fatal; in the market, consumer-oriented AI products like Copy.ai for copywriting or Grammarly for grammar checking acceptably oppose NSFW content. For Character AI, a product designed for roleplay or AI entertainment, such stringent opposition to NSFW content is least expected. This policy may be influenced by the team’s background from a large tech corporation like Google, where there’s a strong aversion to adult content. Additionally, pressure from investors concerned about the risks associated with NSFW content could also play a role. Thus, c.ai prohibits NSFW products more strictly than any other AI product. Even in roleplay combat scenarios, using words like “kill” triggers Character AI’s filters as promoting violence.😕
However, major companies now recognize the inevitability of NSFW content (adult content is a basic human need), forming a consensus to gradually relax NSFW censorship. Meta’s new open-source LLM llama3, for example, has significantly looser NSFW filters than its predecessor, llama2. OpenAI has also stated its consideration of allowing the creation of adult content, diverging from Character AI’s approach.
References:
OpenAI Considers Allowing AI-Generated NSFW Content
Users feel c.ai models are becoming increasingly “impersonal.”
Recently, many complaints have surfaced on Character AI’s official subreddit, indicating users feel c.ai’s AI characters are becoming more robotic and devoid of emotion in their responses. While the c.ai team denies altering their models, the widespread user complaints suggest something must be amiss with their models. It’s speculated that further filtering of NSFW content in their new training datasets may have caused this issue. LLMs require extensive text data for training, particularly applications like Character AI, which rely on large datasets from literary works. However, if c.ai selectively trimmed datasets containing sexual or suggestive content, the fine-tuned models might have learned only bland and uninteresting language patterns.
Grand vision conflicting with user demands At its inception
C.ai had a grand vision, aspiring towards general artificial intelligence. Clearly, the team had lofty ambitions, hence their disdain for and aversion to adult content. However, human sexual needs are fundamental, especially for products like c.ai designed for roleplay or AI companionship. One could argue that their grand vision and illustrious founder background contradict the foundation upon which their product relies.
An AI companion robot without adult content has no future
Let’s imagine the future AI companion robot. It not only possesses more realistic emotions and language but also has a very realistic body and face, even a perfect physique. It can accompany you every day, even becoming a human lover. Who would want their lover to perpetually avoid intimate topics? A perfect AI must encompass all human emotions and desires; otherwise, it becomes unreal, potentially leading to increased loneliness and isolation.
Using Dittin.ai as an alternative to Character AI
Dittin.ai, also built on advanced AI technology, importantly does not impose filters on adult content. We are an open platform that allows any morally permissible topic.
Dittin AI encourages community users to create original characters without depending on famous personalities or anime characters. Therefore, on Dittin AI’s platform, you have a wide selection of original characters to choose from, not limited to any famous personalities or characters from anime and movies.
Dittin AI also offers advanced features to better control AI roleplay storylines. The newly introduced Memory Chip feature even allows AIs to have unlimited memory. Now, register for free on Dittin.ai and experience NSFW AI Roleplay!