After being trained on images from a range of artistic styles, Taichi could transform input images into works in the manner of various artists.
AI computing has become a high-energy-consuming industry, and researchers are racing to try to improve efficiency.
“Taichi paves the way for large-scale photonic computing and advanced tasks, further exploiting the flexibility and potential of photonics for modern AGI,” the team said in the paper.
Dai and his team gained their results by designing a scalable and highly robust distributed computing architecture.
In Taichi, the computing resources were distributed into multiple independent clusters, which were organised separately for subtasks.
The computation and task distributing could also help existing PICs to extend their computing capacity for more advanced tasks,” Dai said in the paper.
The researchers said the work on Taichi underscored the chip’s potential in processing large-scale high-resolution images and training billion-parameter models, paving the way for applications in low-power automated systems.
“We anticipate that Taichi will accelerate the development of more powerful optical solutions as critical support for the foundation model and a new era of AGI,” the team said.
Taichi could turn input images into works in the style of different artists after being trained on images from a variety of artistic genres.
Researchers are working feverishly to increase efficiency because AI computing has grown to be a high-energy-consuming industry.
Tai Chi “paves the way for large-scale photonic computing and advanced tasks, further exploiting photonics’ flexibility and potential for modern AGI,” the authors of the paper stated.
Yury Suleymanov, associate editor of Science journal, stated: “The current work is a promising step toward real-world photonic computing, supporting various AI applications.”. “.
Prior to the US ban cutting off supply, Chinese tech companies are rushing to obtain Nvidia’s AI chips.
Professor Dai Qionghai of Tsinghua’s automation department and associate professor Fang Lu of the university’s electronic engineering department oversee the research team.
Dai and his group achieved their goals by creating a distributed computing architecture that is both scalable and incredibly reliable.
While stacking PICs is the standard method for utilizing them, Dai’s team clustered the devices to create a shallow yet expansive architecture.
The computer power in Taichi was divided up into several autonomous clusters, each of which was set up differently for a subtask.
It wasn’t a Taichi-only algorithm. As stated in the paper by Dai, “The computation and task distributing could also help existing PICs to extend their computing capacity for more advanced tasks.”.
Taichi experimentally achieved high-fidelity artificial intelligence-generated content with up to two orders of magnitude improvement in efficiency, as well as on-chip 1,000-category–level classification (testing at 91.89 percent accuracy in the 1623-category Omniglot dataset). “.
The group introduced an optical chip in October of last year that solved issues with optoelectronic interfaces and computing unit integration.
The Taichi project, according to the researchers, demonstrated the chip’s potential for processing massively parallel high-resolution images and training billion-parameter models, opening the door for uses in automated low-power systems.
According to the team, “we anticipate that Taichi will accelerate the development of more powerful optical solutions as critical support for a new era of AGI and the foundation model.”.
