Experts at the Table: Semiconductor Engineering spoke with Michael Kurniawan, Business Strategy Manager at Accenture, about the impact of cutting-edge technologies such as generative AI, AR/VR, and connected device security architectures in the data center. Kaushal Vora, Senior Director and Head of Business Acceleration and Ecosystems, Renesas Electronics; Paul Karazuba, Vice President of Marketing, Expedera. and Choudhary Yanamadara, technology strategist at Arm. Below are excerpts from that conversation. Panelists were selected by GSA’s EMTECH Interest Group. Click here to see part 1 of this discussion.
LR: Mr. Kurniawan from Accenture. Renesas Vora. Expedera Karazuba. Arm’s Yanamadara.
SE: Previously, a lot of data center applications were for things like enterprise resource planning (ERP), and those were on 10- or 15-year cycles. Current cycles are one or two years long at most. For ChatGPT, this is approximately 6 months. How are companies planning for this now?
Kurniawan: Previously, companies focused only on technology. But today technology is everywhere. ERP exists to support business initiatives, and there is currently a very close relationship between technology and business. So virtually every business is a technology business. Before implementing technology, we advise our clients to first ask themselves, “What is our business approach?” What is your business strategy? What are the business imperatives for where you want to go? What is your vision? And once you understand that and get alignment from your leadership, you can start thinking about technology. They’re actually two sides of the same coin, so it feels like they’re jumping back and forth. You can no longer separate them. And your vision includes everything you want to achieve in the future, giving you room for flexibility and testing the technology plans you want to introduce to see how it supports your business vision. With every challenge comes an opportunity. As consultants, our job is to understand what’s really going on, continually research the market, and strive to stay ahead of the curve and advise our clients.
Yanamadara: The rapid evolution of advanced technologies such as generative AI can pose challenges to data centers due to short technology cycles and demanding workloads. The main challenges with advanced workloads include fluctuating resource needs as they can require large amounts of high compute. This means that static resource allocation becomes inefficient in handling these requests. Additionally, the growing demand for heterogeneous computing may create further challenges in deploying flexible computing infrastructure. Data centers are becoming more flexible with the introduction of containerization and virtualization. Employing hardware-agnostic software frameworks such as TensorFlow or PyTorch can also help ease switching between different computing architectures. The same goes for developing efficient hardware and dedicated AI accelerators.
SE: Many advances in technology are incremental, but if you take full advantage of these incremental improvements, you can combine them in ways that most people can’t imagine. We’ve seen systems shrink from mainframes to his PC to his smartphone, and now computing is happening almost everywhere. Are we trying to move beyond the boxes we’ve been tied to since the advent of computing, especially his AR/VR?
Vola: I think it’s fascinating that someone can put on glasses and be immersed in that world and get used to it. From a user experience perspective, this seems like an extreme change. Although we are seeing some movement in certain industries, it is not clear whether this technology will be mass-consumed or adopted.
Kurniawan: Generative AI is currently attracting attention. ChatGPT attracted the attention of hundreds of millions of people in his 60 days. That’s saying something. Enter the prompt and receive a response. ChatGPT is very intuitive. This is a technology with the potential for many killer use cases. AR/VR is a promising technology with upward potential, but there is still work to be done to connect the technology to use cases. There is no doubt that virtual reality games are number one. However, the path to leveraging that technology to enhance the way other functions operate requires further clarity. That said, we recently published a white paper that describes construction around the world with a combination of public incentives and private investment. Everyone wants to build a manufacturing facility, anywhere in the world. When we interviewed semiconductor experts and asked them what they did during the coronavirus pandemic that shut down the entire world, they mentioned AR/VR. Is AR/VR like a hammer looking for a nail? The overall response we got was pretty positive. They said that AR/VR will probably become very useful in the future. But they like where the technology is going. For example, there are constraints such as heat dissipation and headset size, but we believe that technology will evolve. As you mature to become more user-centric, you may consider using AR/VR devices to control the operation of equipment in your factory. But work needs to be done from a value perspective, such as connectivity and processing.
Karazuba: Until now, AR/VR has largely been a victim of its own hype cycle. There are many promises people have made. We spent some time with the AR/VR folks. There is certainly a recognition that no matter how successful the Apple AR/VR headset is, it will largely determine the trend of the AR/VR market over the next five years. These people are not deterred by such things. Are we now in a time where we can walk around in mixed reality all day long? No, given the performance benefits that come with constant AC power and the performance benefits that come with home gaming systems, even when they’re fixed to the wall. , you’ll probably pay a small price. This will take some time. The value proposition exists, but the timing may not be right right now. We confirmed this with watches and wearables. Now, everyone has one of these. However, it took five to seven years for it to really catch on.
VolaA: We’ve been wearing watches for decades, so it’s nothing new. It’s just that what I’m wearing now is different. But with AR/VR, we’ve never done that before. How can we expect such a big change all of a sudden?
Karazuba: But most of us wear glasses. If you’re using the current version of the form factor and the information is simply overlaid on what’s being displayed, it’s not that big of a leap for mixed or augmented reality. But with virtual reality, it’s hard to believe that people can walk into a crowded conference room and put on a headset.
Yanamadara: We have seen devices and sensors being deployed virtually everywhere. Platforms are now available that provide high-performance computing along with secure and power-efficient hardware and connectivity to enable this trend. However, untearing or ambient consumer experiences in the mass market present challenges. Enabling technology to work invisibly in the background requires investing in extensive infrastructure. So while consumer technology deployments are becoming increasingly untethered, computing and connectivity infrastructure still requires connectivity for power and bandwidth.
SE: People have been sounding the alarm about hardware security for years, with limited success. The change today is that there are more connected devices and more valuable data. Is the chip industry starting to take this seriously? Or is the problem now so huge and pervasive that whatever we do is just a drop in the bucket?
Yanagiyamada: Security is fundamental at the chip level, and five years ago we saw an opportunity to proactively improve the quality of chip security. IoT was in its infancy, with chip vendors taking diverse and piecemeal approaches to security. They also rarely consulted independent reviewers to check the robustness of their security implementations. But as connectivity improved and data became more valuable, hackers were paying closer attention and governments were considering what steps to take to protect consumers. That’s why we launched his PSA Certified in 2019 to bring the ecosystem together to be proactive about security best practices. It is critical that chip vendors, software platforms, OEMs, and CSPs be able to deploy and access standardized Root of Trust services. Security is complex. The entire value chain needs to work together.
Vola: Security architecture has come a long way, at least on the hardware side. Semiconductors like TPM are now almost complete. [Trusted Platform Module] Even tiny microcontrollers have built-in security features. They incorporate cryptographic engines, randomizers, and all sorts of security elements. The fundamental security challenge is that simply putting a few security features on a chip and providing all the technology elements is not enough to solve the security challenge. Security is more of a system issue as well as a policy issue. You often need to think of it in the context of your overall network. And it’s only as strong as the weakest link in the network. As we start to see more complex use cases as AI works with his IoT, the security part will become even more complex. But on the other hand, as AI data processing moves closer to the edge, more local inference and local data will be processed without the need to unconsciously transfer data across network layers and clouds. As intelligence and compute becomes more localized at various layers at the edge, we will see lower risk and improvements from a data inflight perspective. As we begin to move to the edge, AI will begin to gain more ground there. But overall, security will remain an issue. The fundamental security challenges remain the same. It’s just the context and the systems in which they need to be applied are different.
Karazuba: The semiconductor industry is finally starting to understand what a security breach means when it comes to the kind of data we work with. Security is a responsibility from the beginning for those who build the product, whether it is a chip or a device, and security responsibility extends throughout the life cycle of any device, from the person who designs the chip to the person who designs the device. spread throughout. Go to your carrier for your smartphone. I would argue that carriers’ security responsibilities extend to stopping the robocalls, spam calls, and phishing calls we all receive. Internet service providers are responsible for stopping phishing emails. It’s all part of security. Obviously, the security of banks and financial institutions is generally very good. But it’s pervasive, and in the world of security, the weakest link is always the device’s security profile. It’s getting better and better. We can always be better. But I’m more encouraged now than ever since I started seriously thinking about device security. I’m even more encouraged when I see how chips are being designed, deployed, manufactured, and delivered to customers.
Kurniawan: Before an IoT device is released into the market, it undergoes certification to ensure that the device complies with some security standards. But the two keywords I mentioned earlier – collaboration and flexibility – also apply to security. Collaboration involves future evolution of the rest of the system, including other components in the technology set. It also requires flexibility, as security is a moving goal. As you upgrade your systems and software, your vulnerabilities also move, so you need to evolve. Flexibility and security considerations must be built into chip design.
Related books
Preparing for an AI-driven future in chips (part 1 of the above roundtable)
The design must be flexible enough to handle the onslaught of continuous and rapid change, but also secure enough to protect your data.