Huawei’s mission is to bring digital to every person, home and organization for a fully connected intelligent world but it also believes there are ten challenges ahead of us ranging from understanding the world to energy and healthcare.
An intelligent world needs new networks and new computing
Through discussions with academia and industry partners, including scientists and customers it has identified two key areas as we move towards an intelligent world: new networks and new computing.
Dr. Zhou Hong, President of the Institute of Strategic Research outlined the challenges as he addressed Huawei Analyst Summit 2023 explaining “Communications networks are the foundation of the intelligent world. We must continue breaking through bottlenecks in existing theory and technology. This is the only way forward.”
Talking about the future Dr. Zhou said “Rethinking approaches to networks and computing is critical as we move towards an intelligent world. In networking, we have what it takes to move beyond the limits of Shannon’s theorems – as well as applications of his theory – to drive a 100-fold increase in network capabilities over the next decade. In computing, we will move towards new models, architectures, and components, and improve our ability to both understand and control intelligence. We will also continue to explore the use of AI for industry applications, science, and more.”
The goals of AI must be defined and aligned
He added “As AI’s capabilities are improving rapidly, we need to consider how to make sure the development of AI is what people want and ensure AI execution is accurate and efficient. We must create rules and laws to enhance AI ethics and governance.” He believes from a theoretical and technical perspective, these goals present three major challenges: AI goal definition, accuracy and adaptability, and efficiency.
“If we don’t have an agreed-upon definition, it’s almost impossible to ensure that the goals of AI and human beings will be aligned. It also makes it difficult to make reasonable classifications and computations” Dr Zhou explained. He mentioned that there are many schools of AI, including symbolism, Bayesianism, evolutionism, behaviorism, and connectionism but they have not been effectively integrated which may explain why there are no commonly agreed upon goals for AI.
Data must be accurate to prevent “black swan” events
The second challenge that many AI applications face is accuracy and adaptability. Learning based on statistical rules extracted from big data depends on the coverage of sampling and the accuracy of the data used. Improper use of AI will likely result in unstable results and bias, and “black swan” events.
The third challenge for AI is efficiency such as energy efficiency. According to the 60th TOP500 published in 2022, the fastest supercomputer is Frontier, which can achieve 1,102 PFLOPS while using 21 million watts of energy. The second fastest is Fugaku, which can achieve 442 PFLOPS and needs 30 million watts of power to run. Human brains, in contrast, can deliver about 30 PFLOPS with just 20 watts. These numbers show that human brains are about 30,000 to 100,000 times more energy efficient than these supercomputers.
Huawei’s AI for Industry creates value across industries
Dr. Zhou introduced some of Huawei’s work related to AI including AI for Industry, which uses industry-specific large models to create more value for industries. Huawei has developed industry-specific large models based on its existing large foundation models dedicated to computer vision, natural language processing, graph neural networks, and multi-modal interactions. These large models lower the barrier to AI development and improve model generalization, driving a shift from fragmented, labor-intensive AI development to large-scale AI development.
These models have been used in major industries such as electric power, coal mining, transportation, and manufacturing to improve operational efficiency and safety. For example, coal mines are using Huawei’s model training and reasoning services to predict dangerous gas concentrations, prevent safety hazards during operations, and check operational quality intelligently. Huawei’s AI mining model was commercially deployed by a mining company for the first time in 2022, and is set to facilitate large-scale adoption. In addition, Huawei’s smart port solution has already been deployed around China in Tianjin, Qingdao, Shanghai and Shenzhen.
Conclusion: Moving from narrow AI to general and super AI
Dr. Zhou concluded “Networks and computing are the two key cornerstones underpinning our shift from narrow AI towards general-purpose AI and super AI. To get there, we will need to take three key steps. First, we need to continue driving breakthroughs in theories and technologies to deliver ubiquitous intelligent connectivity and drive social progress. Second, we need to keep pushing our cognitive limits to better understand and control intelligence. Finally, we need to define the right vision and use the right approach to guide AI development in a way that truly helps overcome human limitations, improve lives, create matter, control energy, and transcend time and space. This is how our society as a whole will continue to evolve and thrive.”
[Based on Dr. Zhou Hong’s speech]
