Industry leaders explore how AI is contributing to a “digital big bang” and the challenges that are likely to be faced while keeping up with its development
The COMPUTEX Forum — themed Pervasive Intelligence — yesterday morning resumed with the AI Session at the Taipei International Convention Center. Six speakers from major players in the field of artificial intelligence (AI) detailed how the technology is changing not only their businesses but also the world around us.
“We have gathered some of the best minds in the industry to share with us their visions and insights for this very important technology,” said James Huang, chairman of the Taiwan External Trade Development Council, organizer of the COMPUTEX Forum.
“In this new digital landscape … in order to survive and thrive, every industry has to be a digital industry and every company … has to be a tech company,” he said. “New technologies like 5G and AI will unleash life-changing powers and change our societies and economies, and, in particular, the convergence of new technologies like 5G, AI, IoT and big data will reinforce each other and create a … ‘digital big bang’”
“AI is certainly one of the most important technologies that will shape the future of our life and the destiny of mankind,” Huang said. “I think we are embarking on a very exciting digital journey of mankind and there’s no doubt that AI will play one of the most important roles in our future.”
Rene Haas, president of the Intellectual Property Group at semiconductor firm Arm, highlighted how AI is already present in daily life and the monumental challenges the technology industry faces to keep pace.
It is “an unprecedented time relative to our industry and the impacts that are going to take place relative to all things around AI,” Haas said. “It’s very, very significant in terms when you think about just the intelligence that’s coming into machines and the intelligence going into everyday lives.”
“There’s about 4 billion smartphones in the wild … and virtually all of those smartphones today are doing some level of AI or machine learning or getting smarter,” he said. “About 85 percent of that computational intelligence around machine learning, AI takes place in either a CPU or a GPU.”
“Increasingly, the needs of developers have to be addressed by hardware that is heterogeneous in nature. That means that the underlying hardware is not just a CPU. It’s not just a GPU. It’s not just an NPU, for machine learning,” Haas said. “It’s a series of dedicated pieces of hardware.”
“This type of machine learning and artificial intelligence is going to find its way into every single end market that we engage in. Every market is going to have needs for artificial intelligence, it’s going to have needs for machine learning,” Haas said. “We are just at the very beginning in terms of where we are relative to compute needs.”
He highlighted four major challenges: a mature silicon process just as compute demands are increasing, domain-specific processors, a fragmented software/developer ecosystem and security, all of which will require a “total compute solution.”
“We just can’t throw more cycles at the CPU, it’s just not going to be able to keep up,” Haas said.
“Data is the new global currency. There is a staggering amount of information being produced on an ongoing basis,” said Thomas Eby, senior vice president and general manager of the Compute & Networking Business Unit at Micron, the largest foreign employer and investor in Taiwan. “But without the ability to manage that data, to store it, to access it, to process it with sufficient density and in a timeframe of relevance, it’s just that: data.
According to experts surveyed by the memory maker, the most important issues were related to the “memory and storage subsystems needed to support AI. Put another way: How do you feed the beast?” he said. “Increasingly, the performance of systems is being influenced as much, in terms of both performance and power consumption, by the movement of data as by how that data is processed.”
“There was a comment one of the earlier speakers made about how its been an incredible evolution here at COMPUTEX from what was a very significantly gaming-oriented show 15 years ago to what we’re talking about today in the areas of AI and 5G. There’s an interesting parallel there, one of the technologies that is increasingly being applied to solve both the in-vehicle experience, as well as the autonomous driving problem, is graphics memory.” Eby said. “We’ve taken that, produced automotive-grade versions of it, and it’s incredibly valuable in these applications.”