Embracing Human-Like Emotions
Reports from China announce a significant milestone in artificial intelligence: the development of Tong Tong, an AI entity that exhibits emotions resembling those of humans. While some celebrate this as a step forward in the journey towards achieving Artificial General Intelligence (AGI), others express ethical reservations and skepticism regarding the authenticity of its “emotions.”
Crafting Emotion in the Virtual Realm
Crafted by the Beijing Institute for General Artificial Intelligence (BIGAI), Tong Tong exists in a virtual realm, engaging with users through conversation and performing basic tasks. The central assertion is her ability to convey emotions such as joy, anger, and sorrow, purportedly evolving and adjusting based on user interactions.
Tong Tong in Action
At the exhibition, attendees had the opportunity to interact with Tong Tong. When instructed to appreciate neatness, she autonomously corrected a misaligned picture frame. In instances where the frame exceeded her reach, she ingeniously located a stool to facilitate the adjustment without human assistance. Additionally, if a spill occurred, Tong Tong promptly fetched a towel. She independently cleaned up the mess, showcasing her adeptness in understanding and responding to human intentions.
The Tong Test: Evaluating General Artificial Intelligence
Songchun’s team introduced the Tong test as a comprehensive evaluation method for general artificial intelligence, distinguishing it from the Turing test. The Tong test assesses AI across various aspects, such as vision, language, cognition, motion, and learning, along with a diverse value system. With almost 100 specialized tasks and over 50 general tasks, it provides a thorough testing regime.