Satoshi Kurihara, Keio University, Japan
Kazuma Arii, Keio University, Japan
Reo Abe, Keio University, Japan
Taihei Takahashi, Keio University, Japan
We propose the realization of AI (BuddyAI) with high autonomy and versatility that can perform compassionate interactions with people.
Each person will have one BuddyAI, and BuddyAI will be connected through a network.
The accumulation of satisfaction from people through AI's autonomous interaction with them leads to the acquisition of trust in AI.
As a result, people accept advice from BuddyAI with conviction,
enabling behavioral changes that would be difficult to achieve through self-help alone.
The success of mega-scale LLM development, such as ChatGPT, is essential for the realization of BuddyAI.
However, it isn't easy to realize BuddyAI as an extension of the foundational model, so we propose the realization of a new symbol-based AI based on LLM.
