British robotics company Engineered Arts has demonstrated new capabilities of a humanoid robot
The developers indicate in the description of the video thatAmeca used the GPT-3 language model for speaking and translating in this demo. In addition, the robot was integrated with the DeepL machine learning model for speech recognition and voice generation.
Ameca speaks different languages. Video: Engineered Arts
Previously, Engineered Arts has demonstrated the abilityhis robot to realistically convey various emotions: from surprise and discontent to disgust. The researchers note that they have tried two models, GPT-3 and GPT-4, for language learning. At the same time, the later model resulted in a slow response, due to which the realistic transmission of facial expressions was reduced.
The team is currently working ona demo version of the robot that will use the voice cloning system from Eleven labs. This model uses more complex facial expressions thanks to the additional "phoneme generation" of lip sync. It is expected that an upgraded version of the robot will be presented in the near future.
Read more:
Special diet triggers self-destruction of brain cancer cells in mice
Scientists have deciphered the mysterious manuscript of Ptolemy. It was hidden under other text
New image of Hubble puzzled scientists