Qualcomm Business Development Director Andy Li: AI is not a "winner-takes-all," and Web3 is not a "flood beast."

Interview: Sima Linwei

Written by: Chen Xiaorui

Produced by: DeThings

On April 18, Aethir launched the Aethir Edge product, supported by Qualcomm technology, during Token 2049 in Dubai, equipped with the Qualcomm Snapdragon 865 chip.

Qualcomm is a semiconductor and telecommunications technology company founded in 1985, headquartered in San Diego, California, USA. Its main business includes the research and development and sales of mobile communication chips, as well as the licensing of related patent technologies. Its products and technologies are widely used in various wireless communication devices such as smartphones, tablets, automobiles, and IoT devices. Aethir is a decentralized GPU cloud infrastructure provider established in 2021, headquartered in Singapore. Public information shows that in July 2023, Aethir completed a $9 million Pre-A round of financing, with a valuation of $150 million at that time.

Recently, DeThings interviewed Qualcomm's Director of Business Development, Andy Li. Below is the full content, with some editing.

DeThings: Can you briefly explain and introduce Qualcomm's technology and business?

Andy Li: Over more than 30 years of development, Qualcomm has accumulated a wealth of technology. The most well-known is our technology in the mobile communications field, including the earliest CDMA to 3G UMTS and CDMA2000, then to 4G LTE, and now to 5G. Qualcomm recently announced the 5.5G evolution plan, which we refer to as 5.5G, essentially an evolution of 5G. Our labs have successfully achieved 6G technology, though the specific level of commercialization will depend on the market.

In addition to mobile communications, we also have technologies in computing fields such as CPU and GPU, like the Snapdragon CPU and GPU. Our GPUs achieve a good balance between high performance and low power consumption. Long ago, Qualcomm developed its own DSP architecture, previously known as C DSP (C stands for computing), which is a DSP for computing. With the continuous development of AI algorithms such as computer vision, CNN, TensorFlow, etc., Qualcomm has been adjusting the C DSP architecture to adapt to the latest technologies. Recently, we also launched Qualcomm NPU, which includes matrix accelerators, tensor accelerators, and vector accelerators, with many optimizations in memory sharing bandwidth and access.

So Qualcomm's technology is not only in mobile communications but also includes computing technologies. Applying these technologies to the IoT, the first step is to connect objects, which requires utilizing Qualcomm's connectivity technologies, including cellular network connections, Bluetooth, WiFi, and other near-field communication technologies. After connecting thousands of objects, the problem of data silos is solved. Each object, though it may just be a small sensor, continuously generates data. In the past, when not connected, each device's data existed like an island, requiring manual data retrieval and analysis. Once connected, this data forms a network, and at the network level, AI large models can automatically learn from and analyze this data, extracting value from it.

DeThings: From the PC era to the mobile internet era, the term "Internet of Things" (IoT) has emerged. What are your thoughts on the term "Internet of Things"?

Andy Li: For Qualcomm, the Internet of Things is not a new term. The IoT has actually existed for a long time; it is, as the name suggests, about connecting objects into a network, which is the first step. We have a very long history in this field. People may recognize Qualcomm for its mobile communication-related business, such as the Snapdragon platform on mobile devices, but in reality, Qualcomm's business is very broad. Qualcomm positions itself as a partner that empowers various industries by utilizing our product and technology combinations. The Qualcomm product that people are more familiar with is the Snapdragon brand, and we have recently launched products for PC and XR (VR/AR).

Whether it's Depins or the so-called Web3, they are essentially the same as traditional IoT, which is about connecting all devices. The difference is that in the past, due to the weak computing power of each device, a powerful centralized computing platform was needed for control. However, as the computing power of each node increases, decentralization or Depins becomes a possibility, because only when each node is powerful enough can decentralization be achieved; otherwise, a strong central brain is needed for processing. When every detail has a certain level of computing power, a flat network can be constructed. Qualcomm's opportunity lies here; we have a technology combination at the edge and terminal device side that can empower each node with strong capabilities, and with connectivity technology linking them together, only under this architecture can a truly decentralized network be realized.

DeThings: After the emergence of ChatGPT, everyone's attention has been drawn to AI. How does Qualcomm view computing in the AI era?

Andy Li: I believe AI may truly be the intelligent productivity that can change the entire structure of human society again since the Second Industrial Revolution, thereby altering our entire social organizational structure. However, the general public may have a generalized understanding of AI, which is artificial intelligence. So how do we implement artificial intelligence into specific aspects of life, making it a tangible entity?

First, we may have seen general generative AI like ChatGPT. But ChatGPT represents a type of general AI, essentially a form of general knowledge. To achieve general knowledge means you need to understand all knowledge. Therefore, general AI like ChatGPT will have an increasingly large parameter scale, with models exceeding 200 billion parameters already appearing.

But returning to the IoT or our industry terminal field, my view is whether you really need a general generative AI like ChatGPT; this needs further discussion. Because in the IoT field or industry terminal devices, they are not used to do everything. When designing a device, we have already considered that it needs to do a limited number of things in the field it faces. Just like we wouldn't ask a radio, "What gift should I give my wife for her birthday?" We only ask operational questions related to that device. In this case, do we still need a general AI? I think it may not be necessary; it needs a customized, vertical AI model tailored to the specific field.

This further narrows the scope of AI. From a hardware demand perspective, it will also correspondingly decrease. Because if you want to run a model with hundreds of billions of parameters, it can currently only run in data centers. On a mobile phone, you might only be able to run a model with a few billion parameters. For IoT devices, a small model with 7 billion or even 4 billion parameters is already sufficient. What we need to do is not stack hardware to run large models but focus on how to make AI models run more efficiently.

Qualcomm also has relevant layouts in this area. In addition to continuously enhancing chip capabilities, allowing devices that previously could not run large models to now do so, we have also launched Qualcomm's AI framework. It should be noted that Qualcomm does not create large models but provides a complete set of tools. Last week, we released AI Hub at the Embedded World Exhibition in Nuremberg, Germany, which actually ported hundreds of AI models from the open-source community to the Qualcomm Snapdragon platform and optimized them, making them open-source for all developers to use. We hope that our partner developers can develop truly personalized AI solutions suitable for various industries on the Qualcomm platform.

DeThings: In other words, do you or Qualcomm believe that AI is a so-called "winner-takes-all" field? For example, we know that OpenAI is leading the development of a large model with trillions of parameters. In such a powerful AGI (Artificial General Intelligence) scenario, why do we still need other AIs?

Andy Li: You raised a very good question, and I will actually address this in my presentation. As you said, if AI is a "winner-takes-all" field, it must be a giant AI that can only run in the cloud, because neither edge devices nor terminal devices can support it.

First of all, every specialized AI model is derived from a general large model; we simply customize and personalize it based on industry needs. Therefore, we need general AI and then segment it.

Returning to your question, why not connect all terminal devices and industry terminal devices to the cloud? This involves economic efficiency. You also know that a huge data center in the cloud has very high costs. Even for a simple inference, it requires full startup, consuming enormous energy, including the energy consumption itself and the energy consumed by the air conditioning in the data center. Now, data centers are better built in colder regions and near power supply areas to reduce these expenses. This is one aspect.

Secondly, we need to consider that AI application scenarios will become more and more numerous, and the number of users will also increase. Suppose in the future, there are 1 billion people using AI globally; the energy consumed for each AI startup and inference multiplied by such a large number of application scenarios and users would be a huge drain on economic efficiency. In this case, is putting all computing in the cloud the optimal solution? Or can we place some computing on edge and terminal devices? I believe the latter is necessary.

More importantly, many industry terminal applications care more about latency and the reliability of that latency. If through the cloud, you tell me it won't exceed 20 milliseconds each time, but the latency could vary from 10 milliseconds to 100 milliseconds, this uncertain latency is unacceptable for many industry applications. However, on edge and terminal devices, you can achieve a certain and very low latency. That is the reliability of the connection.

The third point is the issue of data security, which is also a significant reason for promoting decentralization, to ensure data security and reliability. Of course, the focus is more on data security. For reliability, we hope that data is stored locally or within a controllable and accessible limited range, serving as a node in a larger decentralized network.

Therefore, from the perspective of the AI era, general large models may not necessarily meet all specific needs in terms of economic efficiency. Moreover, in terms of reliability and security, general large models also have some shortcomings and still require industry-customized solutions. General large models are not the universal answer.

DeThings: We know that global smartphone shipments are shrinking, and there is a stereotype that Qualcomm's Snapdragon platform is strong in mobile phone chips. We want to know where Qualcomm sees its next growth point?

Andy Li: First of all, although global smartphone shipments have declined compared to pre-pandemic levels, they are slowly recovering.

Secondly, we still consider mobile phones as our main business, and that has not changed. Many of our technology developments started with mobile phones; the mobile phone market is actually the largest single demand market in history. Although there are only a few smartphone brands, their demands are still very similar, so the smartphone market is very fertile ground for us to cultivate new technologies.

In addition, Qualcomm is indeed pursuing business diversification. You can see that we have entered fields such as IoT, XR (Extended Reality), PC, and automotive. For example, in the automotive field, whether it's advanced driver-assistance systems or in-car infotainment systems, Qualcomm is in a leading position. We will continue to advance this diversification strategy.

The scale of these emerging markets is also very considerable. Taking the IoT field as an example, we believe its total market capacity reaches $720 billion, even exceeding the smartphone market. Although the IoT is a highly fragmented market, with each industry's volume not being very large, it has its own independent demands. But for us, we can penetrate the technical demands behind different industry applications, which are actually consistent. Our technology and patent combinations are very flexible, allowing us to provide different technologies, products, and solution combinations for different industries. I believe this combination of technology and patents is also very beneficial to customers, enabling them to access the most advanced technology in the most efficient way.

DeThings: Besides the core mobile phone business, what do you think is Qualcomm's biggest potential growth point in the future?

Andy Li: Apart from the mobile phone business, we believe one of the biggest growth points lies in the IoT and automotive fields.

The IoT is actually one of the largest growth points. Additionally, whether it's new energy vehicles or fuel vehicles, we believe the automotive field has enormous potential. Even fuel vehicles have demands for intelligence. New energy vehicles do not have to worry about battery issues, and as new products, they are easier to design, but fuel vehicles also need intelligent upgrades. For us, the form of the vehicle's drive is just a form; our goal is to make vehicles smarter, achieving interconnectivity between vehicles, traffic facilities, and, more importantly, between vehicles and drivers, which is precisely Qualcomm's area of expertise.

In fact, vehicles can be seen as part of the IoT because we view them as intelligent connected devices. Therefore, vehicles still belong to Qualcomm's overall layout of the IoT ecosystem.

We believe Qualcomm has unique advantages in these areas because we have a wide range of partners and a large customer base and partnerships. Qualcomm has always sincerely shared our latest technologies with partners and has not been stingy. We believe that our customers benefit from this, and society as a whole will benefit as well.

DeThings: How do you view the concept of Web3? Because besides AI, the term Web3 has also been very popular in recent years.

Andy Li: For Web3, we see it as a very emerging, potential, and promising way of economic activity. For Qualcomm, we play a role as an enabler because Web3 is decentralized, with no absolute big boss; everyone is an equal contributor.

As I mentioned earlier, Qualcomm has always been and will become more open. We will open our hardware platform and even open-source our software, embracing open-source software more. We also have a certain degree of open-source code ourselves. We will empower everyone, allowing our partners and customers to play a role in Web3 and potentially in future Web4. We provide them with the tools to excel in this vast field. Because we believe AI is a revolution in productivity, Web3 concerns the transformation of production relationships. Therefore, Qualcomm adopts an open role as an enabler in this emerging industry.

DeThings: Why choose to collaborate and experiment with a Depin platform like Aethir? For a company like Qualcomm, it seems rare to venture into the Web3 field. Web3 has strong financial attributes and involves many aspects of token economics. Does Qualcomm have concerns in this area?

Andy Li: Aethir is actually a partner we are very optimistic about. They have a long accumulation of experience in both cloud computing and edge computing. If you visit Aethir's website, you can see their real-time data display of resources, such as the GPUs currently running, which I think is excellent. Because they are not a conceptual entity but a real business that has been implemented in commercial practice. For a pragmatic company like Qualcomm, we also hope to collaborate with such a pragmatic business partner, whether to build a role model or to create a real production tool.

Moreover, we do not participate in the economic aspects you mentioned; that goes beyond what Qualcomm can foresee. We focus more on the technical level; we believe technology knows no borders, and technology itself is neither good nor bad. We empower everyone with technology. However, I personally do not believe that Web3 is a "flood beast"; we need to understand and embrace all new things, as only by understanding can we use them correctly.

Qualcomm's role in Web3 is that of an enabler. We hope that customers can use Qualcomm's technology and products to create products that benefit social progress and development within legal and ethical frameworks.

DeThings: We noticed that companies like Google, Microsoft, and NVIDIA have already had some Web3 clients in infrastructure construction. Does Qualcomm have further observations or attempts in the Web3 field?

Andy Li: We are open to this and welcome all customers to contact us. We believe Web3 represents a new type of production relationship. As a technology enabler, Qualcomm is willing to share its capabilities and technologies.