The road to success for AI is full of challenges. Microsoft executives state that the massive investments in AI will take at least 15 years to pay off, intensifying concerns about an AI bubble. High input costs and limited application scenarios are the main contradictions in the current AI industry. Meta's closure of its celebrity chatbot project further confirms the view that AI struggles to generate profits.
Meta launched celebrity AI chatbots at last year's Connect conference, paying millions of dollars to use the likenesses of various celebrities. Meta hoped to achieve a cold start for its AI chatbot business by leveraging celebrities' fan bases, but the results were poor. The concept of "celebrity AI companions" itself is problematic, and no similar products have succeeded to date.
Weibo also launched a celebrity AI emotional companion feature but did not scale it up subsequently. The reason celebrity AI companions struggle to succeed is due to insufficient supply. For celebrities, AI companions break the sense of distance between them and their fans. Additionally, training AI requires a large amount of private data, which celebrities are unwilling to provide. Using general models like MBTI results in AI lacking personality.
The lack of "star quality" is the main problem with current celebrity AI companion products, leading to heavy users consuming too many resources, low retention rates for casual users, and difficulty in making the project profitable. However, AI companionship itself still has value. Meta has launched AI Studio tools, allowing creators to develop their own AI chatbots.
Establishing emotional connections between AI and humans is technically feasible and there is market demand. AI companionship changes the subject of social interaction, with users feeling no social pressure when conversing with machines. AI companions can be seen as an advanced version of "virtual companionship," avoiding the emotional issues present in human companionship services.
The commercialization path for AI companions is clear, mainly relying on user paid subscriptions. However, large-scale promotion is difficult, making it suitable for small teams to develop "small but beautiful" products. Meta has chosen to provide tools to developers through AI Studio rather than directly developing applications.
In the current environment where AI applications struggle to gain traction, providing tools to developers may be more promising than directly developing applications.