China Unveils World's First AI Baby

Tong Tong, an artificial intelligence (AI) girl with behavior and abilities similar to a 3-4 year old child, can automatically clean the room and wipe away stains.

Artificial intelligence (AI) is becoming more and more common in human life. Photo: Adobe Stock — Artificial intelligence (AI) is becoming more and more common in human life. Photo: *Adobe Stock*

The virtual AI girl named Tong Tong or Little Girl, developed by the Beijing General Artificial Intelligence Institute (BIGAI), was just introduced for the first time in Beijing, Interesting Engineering reported on February 2. BIGAI considers Tong Tong a big step towards achieving an Artificial General Intelligence (AGI) factor - when a machine can think and reason like a human.

Unlike single-task agents, AGI agents aim to be omnipotent. The goal is to create an entity capable of autonomous cognition, decision-making, learning, execution, and social cooperation, while aiming for human emotional and moral similarity.

During the presentation, Tong Tong - considered the world's first AI child - displayed behaviors and abilities similar to those of a 3- or 4-year-old. Guests interacting with Tong Tong watched as the AI girl cleaned up her surroundings, automatically fixing a warped photo frame and grabbing a chair to reach higher places. Notably, when she saw a liquid spill, the AI girl quickly found a towel and cleaned it up efficiently, demonstrating her ability to understand and respond to human intentions.

What sets Tong Tong apart from other big language AI models like OpenAI’s ChatGPT or Google’s Bard is her ability to independently assign tasks to herself. This AI girl has demonstrated a level of autonomy never seen before in virtual entities. In addition, according to the team behind Tong Tong, she has her own emotions and intelligence, which allows her to learn on her own—a significant step forward for AGI.

"To achieve the goal of AGI, it is necessary to change the 'parrot model' with the structure of 'big data, small tasks' and transform it into the 'crow model' with the structure of 'small data, big tasks'," said Zhu Songchun, president of BIGAI.

Rote learning, he explained, is simply imitation through repeated training, a type of low-level intelligence. In contrast, the crow model is synonymous with independent reasoning behavior and advanced intelligence driven by value, cause and effect. "This is a major achievement of the Beijing Institute of General Artificial Intelligence in the past two years, marking a major step forward in the research of general artificial intelligence," Zhu said.

Thu Thao (According to Interesting Engineering )

Source link