Computex 2023 Reveals Taiwan’s Critical Role in AI
I’ve just returned from Computex 2023 in Taipei, Taiwan, and must say I’ve noticed a massive change in mood from the last time I was there exactly four years ago.
Back in 2019, officials and policymakers were keen to impress on the world that Taiwan represented more than just manufacturing in the tech ecosystem. Taiwan was clearly well known for Taiwan Semiconductor Manufacturing Co. (TSMC) and computer manufacturing, but the officials wanted to highlight the research, the spinouts and the knowledge-driven economy.
When I interviewed the then minister for the office of science and technology (MOST), Liang-Gee Chen, he told me Taiwan was clearly in a good place to address the new growth in 5G, IoT, autonomous vehicles, healthcare and industrial/smart manufacturing. That’s because the smart connected world that everyone was striving for in all market sectors relied on a data-driven economy and, hence, AI was a special segment for Taiwan to play in because of its capability to provide all kinds of server hardware and help companies create value from that data.
Back then, Chen said, “Manufacturing will still be the driving force behind the new jobs in the industry. Many companies rely on Taiwan within their supply chain, which is already well established for the semiconductor industry. We are trying to leverage this strength in manufacturing and hardware, to encourage new systems applications and fabless startups that address the needs of new applications that require a combination of both hardware and software.”
He added, “That’s why I say combining hardware capability with software services could be the strength of the Taiwanese companies. And it’s very easy for those companies to cross the unicorn lines. The ability to add value on top of the hardware is very easy.”
In 2019, I was encouraged to meet the wider ecosystem of startups, incubators and investors in emerging technologies—such as health tech, the IoT and AI.
Fast forward to 2023, it’s clear that the minister’s plans had somewhat come to fruition.
The exhibitors and speakers at both Computex and its related InnoVEX Forum certainly demonstrated a mood of optimism for Taiwan’s crucial role in the AI economy—and that the new generative AI boom would come to rely on the hardware even more. That dependency would be on everything from chips to servers and embedded computers and AI applications.
The computing ecosystem established on the island over the years is suddenly being recognized as important for the next boom in energy-efficient, high-performance computing. If you spoke to any Taiwanese locals and officials, you could feel the sense of national pride in this new importance in the global AI value chain.
As if to re-enforce that, the organizers at Computex and InnoVEX looked to capitalize on the interest in generative AI, with daily keynotes focused on this, and particularly on the technologies being developed to service the large language models (LLMs) underpinning ChatGPT and others. As I already highlighted, the opening keynote on Monday morning ran for over two hours, with Jensen Huang, CEO and founder of Nvidia, waxing lyrical about next generation GPUs, superchips and energy efficiency.
Qualcomm Technologies’ keynote was delivered by Alex Katouzian, general manager for mobile compute and XR, and Kedar Kondap, general manager for compute and gaming, and they both spoke on the endless possibilities with LLMs and how their Hexagon processor and AI stack will help in enabling on-device AI to allow generative AI to reach its full potential.
“Generative AI is the new frontier,” Kondap said.
Meanwhile, Rafael Sotomayor, general manager for secure connected devices at NXP Semiconductors, gave a slick keynote that put the implementation of AI into context in the world of smart connected devices, whether for the home or for the factory. His overarching theme was that there was a need to work with ecosystems to manage and master the complexity of AI at the edge.
“Our goal is to bring machine learning capability to the masses,” he said, as he highlighted examples of microcontrollers (MCUs) and how they are being implemented in various applications with customers and ecosystem partners, such as the i.MX RT1180 specifically for smart factories, and its work with Aqara for smart home devices.
On the startups front, the opening of the InnoVEX Forum was a fireside chat on the opportunities for the semiconductor industry from the explosion in generative AI.
I hosted this keynote interview with Julie Mathis, general manager and VP of AI at Tenstorrrent. In the interview, we explored things like, “How fast is generative AI taking off compared with other market trends?” We discussed the current AI hardware landscape, especially those building chips for generative AI, where they are, what role open source will play in this explosion of generative AI, and how fast the innovation in machine learning is happening. In the discussion, we also looked at what happens when new models evolve, such as ChatGPT4, and how the industry can or will keep pace.
Philosophically, we also tried to answer the question of whether hardware is driving the software, or software is driving the hardware when it comes to generative AI.
The upshot of all of this? There was a lot of excitement this year at Computex. During the four-day show, 47,594 professionals from 150 countries attended, which the organizers said was a 12% increase compared with before the pandemic in 2019, according to the official press release. The top 10 countries/region of visitors were Japan, the United States, Korea, China, Thailand, Hong Kong, Vietnam, India, the Philippines and Indonesia.
Many speakers reiterated the importance of Taiwan in enabling this future. As far as technology was concerned, clearly it was around generative AI, but there was another more important theme coming through: the need for sustainable computing.
As all these LLMs and compute demands grow, there’s a need to do so without draining the world’s energy resources. This also came through at the recent imec ITF world conference in Antwerp, Belgium, where imec CEO Luc Van den hove suggested that generative AI is the modern big bang and compute needs are exploding, and that semiconductor innovation was the key to delivering all of this sustainably.
As I mentioned, Taiwan was super excited to be at the heart of the generative AI boom. International visitors were also keen to figure out how to connect with the ecosystem, ensure supply of products, and build partnerships for tapping into the talent and skills from Taiwan’s universities. As James Huang, chairman of Taiwan External Trade Development Council (TAITRA), said in the opening ceremony at Computex 2023, “AI is a golden opportunity for Taiwan. We are ready to lead in the AI change.”
The post Computex 2023 Reveals Taiwan’s Critical Role in AI appeared first on EE Times.