As intelligent computing systems approach the ceiling of computing power, energy consumption surges, and general intelligence struggles to emerge, the natural advantages of the human brain—"low energy consumption, parallel processing, learning ability, and strong fault tolerance"—are becoming increasingly prominent. Developing brain-inspired intelligence that "converges with the human brain" in all aspects has become a "strategic high ground" that may break through the bottleneck of artificial intelligence development.
On the afternoon of December 29, the "2023 Intelligent Computing Forum" hosted by the Guangdong Institute of Intelligent Science and Technology (hereinafter referred to as "Intelligent Institute") and the "2023 Intelligent Computing Innovation Product Launch" jointly organized by the Guangdong-Hong Kong-Macao Brain Intelligence Engineering Center and the Shanghai Lianhui Brain Intelligence Engineering Center were held at the Hengqin International Science and Technology Innovation Center.
With the theme of "Reshaping Intelligent Computing Power, Linking Green Future," the forum attracted nearly 200 experts, scholars, investors, industry enterprise representatives, and media representatives from the domestic brain-inspired intelligence and computing fields to gather in Hengqin, sharing innovative achievements in the intelligent computing field and discussing strategies for the development of the intelligent computing industry.
Multi-Party Discussion: Gathering Strength for Brain-Inspired Computing Industry-Research Mutual Promotion
The forum invited academicians, experts, scholars, and renowned entrepreneurs to discuss the development blueprint for brain-inspired intelligent computing. Representatives from the Economic Development Bureau of the Hengqin Guangdong-Macao Deep Cooperation Zone, the Zhuhai Science and Technology Innovation Bureau, and other relevant leaders, as well as leaders of organizing units, attended the forum opening ceremony. Four renowned scholars in the field of brain-inspired intelligence delivered keynote reports at the forum. Representatives held symposiums and cross-disciplinary seminars on topics such as industrial ecosystem construction, technological development, and application development of intelligent computing, pooling wisdom and offering suggestions.
World's First Launch: Smaller, Faster, Greener, and Quieter "Intelligent Computing Brain"
The Guangdong-Hong Kong-Macao Brain Intelligence Engineering Center is the transformation platform of the Intelligent Institute. At the product launch, "Zhuhai Hengqin Neogenint Technology Co., Ltd." and "Zhuhai Tianqin Core Intelligent Technology Co., Ltd." transformed by the Guangdong-Hong Kong-Macao Brain Intelligence Engineering Center jointly released a green brain-inspired intelligent computing system with the new generation brain-inspired computing architecture (LYRArc) and processing chip (BPU) as the technical core.
World's First Programmable Brain-Inspired Wafer Computing Chip with 100-Million-Neuron Scale
The team broke through multiple key technologies such as wafer-level chip computing architecture, interconnection architecture, and physical architecture for whole-brain computing, achieving a wafer-level system-on-chip "Tianqin Xinhai" exceeding 200 mm*200 mm. A single chip can support 200 million neuron simulation computing, making it the world's first programmable brain-inspired wafer computing chip with 100-million-neuron scale, 200 times that of Intel's most advanced brain-inspired computing chip Loihi2, helping future brain-inspired computer systems truly achieve the integration breakthrough of "smaller volume, larger capacity." The team has successfully developed a brain-inspired wafer computer based on this wafer chip, which is expected to help brain science research and the development of whole-brain brain-inspired large models with 100-billion-neuron scale in the future.
Based on the self-developed LYRArc scalable processing architecture and Chiplet technology mentioned above, the team has also launched low-power, high-efficiency brain-inspired chips "Tianqin Core" LYRA Slim, LYRA Max, and computing board cards for different brain-inspired computing applications, covering processing needs from millions to tens of millions of neurons, further expanding the application of brain-inspired intelligence in fields such as edge computing.
Green Intelligent Computing Brain
The system also adopts the team's self-developed brain blood vessel-inspired passive phase-change cooling system to achieve energy saving and noise reduction. Whether in terms of comprehensive computing power, volume energy consumption, or operation mode, it is closer to the real human brain. Compared to traditional computing platforms built on CPUs, GPUs, etc., the system can improve computing energy efficiency for spiking neural networks by more than 100 times, with advantages such as high flexibility, high processing efficiency, high interconnection bandwidth, and ultra-low communication latency. The self-developed passive cooling technology it adopts can control data center energy efficiency within 1.1, becoming a truly self-developed "green intelligent computing brain" that is closer to the real human brain operation mode.
The launch of this system provides a new path for solving the underlying core technical problems of brain-inspired intelligence industrialization development and is expected to help intelligent industry applications such as large model training, brain simulation, industrial computing simulation, social network analysis, and financial risk control analysis, forging a "new Chinese path" for intelligent computing.
