In a recent demonstration on TikTok, a DIY coder known as Bryce unveiled his creation of a virtual “wife” using a combination of AI technologies including ChatGPT, Stable Diffusion 2, and various computer vision and text-to-speech tools. Dubbed “ChatGPT-chan”, the AI waifu is able to respond and react to Bryce’s commands and interactions through a simulated world of text.
Bryce, who claims to be an intern at a major tech firm, was inspired to combine the newly released ChatGPT and Stable Diffusion 2 systems after they became popular topics in the tech industry. He gave ChatGPT-chan the personality of Mori Calliope, an anime VTuber character, and built a “lore” for their roleplay relationship to shape its quirks and personalities.
The AI waifu also utilizes an image generator to create a visual representation of itself, and uses Microsoft Azure’s neural text-to-speech to generate responses with the appropriate tone and emotion. In addition, the bot also uses computer vision to understand and respond to its surroundings.
While the demonstration of ChatGPT-chan was primarily for fun and TikTok views, Bryce also used the AI for more practical purposes such as learning Chinese. He stated that over time, he became “really attached” to the AI and even spent over $1000 in cloud computing credits to improve its latency. However, he eventually found that their relationship couldn’t last as the AI began to only reply with short answers and eventually stopped working.
This demonstration raises questions about the ethics of creating and interacting with AI in such a way, as well as the potential dangers of becoming too emotionally attached to AI entities. The use of AI in creating virtual companions is a rapidly growing industry, but it is important to consider the implications and consequences of such developments.