China creates world's first AI child — shows human emotion

The aim is to craft an entity endowed with autonomous abilities in perception, cognition, decision-making, learning, execution, and social collaboration. 

Simultaneously, it seeks alignment with human emotions, ethics, and moral concepts, embodying a multifaceted and ethically resonant approach to AI.

When Tong was showcased, she displayed behavior and capabilities similar to those of a three- or four-year-old child, as reported by the South China Morning Post.

Visitors interacting with Tong Tong watched her clean up her surroundings, autonomously correcting a crooked picture frame and fetching a stool to reach higher spots. 

Notably, when confronted with a spilled liquid, she promptly located a towel and efficiently cleaned up the mess, showcasing her ability to understand and respond to human intentions.

Future development trend of AGI

What sets Tong Tong apart from conventional large language models in AI, like OpenAI's ChatGPT or Google's Bard, is her ability to assign tasks to herself independently. She demonstrated a level of autonomy previously unseen in virtual entities. 

Moreover, she possesses her own emotions and intellect, as claimed by the researchers who developed her, making her capable of autonomous learning a significant stride in the realm of AGI.

"To achieve this goal (AGI), it is necessary to change the 'parrot paradigm' with 'big data, small tasks' as the structure and convert it into the 'crow paradigm' based on 'small data, big tasks' as the structure," said Songchun in a press release.

He explained that parrot learning is a simple imitation through repeated training, a kind of low-level intelligence. In contrast, the crow paradigm means independent reasoning behavior and advanced intelligence driven by value and cause and effect.

"This is an important achievement of the Beijing General Artificial Intelligence Research Institute in the past two years, marking a key step in the research of general artificial intelligence," said Zhu Songchun, president of BIGAI.

The Tong Test

Songchun's team also proposed a new method for testing general artificial intelligence called the Tong test. This is different from the Turing test, which can only tell us whether an AI has reached a certain level of human communication but cannot measure its intelligence.

The Tong test assesses an AI on vision, language, cognition, motion, and learning. It also incorporates a diverse value system, encompassing physiological, survival, emotional, social, and group values. 

With almost 100 specialized tasks and over 50 general tasks, the Tong Test provides a thorough testing regime for developing general artificial intelligence. 

According to BIGAI, this testing approach is crucial for general AI to seamlessly integrate into human environments, emphasizing practical abilities and values, reported SCMP.