Qualcomm CEO: Generative AI is entering smartphones "very, very quickly"

The footsteps of technology giants in the layout of AI continue in 2024.

On January 4th, Qualcomm CEO Cristiano Amon predicted in an interview with the Financial Times that 2024 will be the crucial year for global AI smartphones. Generative AI is rapidly entering the mobile phone industry.

The main points include:

In 2024, mobile devices will play an increasingly important role in AI computing, and hybrid models will become important.

The introduction of AI capabilities may also stimulate a new round of growth in the smartphone market, changing the industry competition landscape long dominated by Apple and Google.

The main battlefield for chip manufacturers is shifting from "training" to "inference," creating huge market potential.

AI operations on devices can be more precise, faster, and cost-effective without the need for cloud involvement.

Spatial computing will become a new computing platform.

Hybrid models will become important

With the improvement of AI model training methods, the field of artificial intelligence is developing smaller and more efficient AI processors and models, making them integrable into portable devices. Complex AI calculations and data processing that used to be performed in the cloud can now be done directly on personal devices.

Amon pointed out a classic pattern in history:

"Innovation often initially appears in centralized systems (such as large computers or cloud computing) and then transitions to more decentralized systems (such as personal computers and mobile phones)."

Amon mentioned that for consumers, AI can operate on devices in two main ways:

"One way is that the AI system predicts the user's every action in real-time, proactively understands the user's needs, and provides relevant services. Traditional cloud-based AI usually requires users to explicitly make queries or commands to elicit a response";

" The other way is hybrid AI, where some AI processing is done on the local device, while the rest is completed in the cloud. This approach can reduce the cost of running AI because not all computations need to be done in the resource-intensive cloud.

At the same time, since the local device can access the user's real-time contextual information (such as location, current activity, etc.), the response of AI can be more accurate and personalized, making it more efficient than running AI applications in the cloud."

For example, currently, when users use Microsoft's Bing Chat, all queries are sent to the cloud and processed by the GPT model. This approach is costly and energy-consuming.

On the other hand, on new AI devices, such as devices with instances of the GPT model installed, queries are first processed locally on the device, and then only a portion of the data may be sent to the cloud to complete the processing, reducing the computational burden, cost, and energy consumption in the cloud. Sometimes, the model on the device may be accurate enough to independently complete the entire query process without the need for cloud involvement. Amon believes that hybrid models will become important, especially considering the high cost and energy consumption of running generative AI in data centers.

Better AI Mobile Experience

When AI runs directly on devices, it is faster and provides a better user experience. Amon mentioned a Stable Diffusion model that generates images from text input. This model can generate images on the device in just 0.6 seconds, greatly improving the user experience for taking and editing photos.

AI applications should not only execute simple commands but also provide a deeper level of interaction, offering appropriate assistance and suggestions.

For example, text input smart assistance. If a user mentions a conference call next week with a specific date in the text, AI can automatically bring up the calendar app, display available times, and proactively ask if the user wants to send a calendar invitation. This kind of interaction goes beyond the capabilities of traditional text input, understanding and responding to the user's specific needs and context.

Another example is intelligent photo management and sharing. AI can recognize the content mentioned by the user (such as vacations and family) and suggest sharing relevant photos based on that. This not only makes communication more convenient but also enriches the exchange.

In addition, Amon points out that compared to past technologies like Microsoft's Clippy, current AI applications are more accurate and intelligent. Past technologies struggled to accurately understand the user's language and context and required programming for each step. Now, devices with built-in AI can learn the user's behavior and preferences more accurately and provide a more personalized and non-intrusive experience.

Currently, different OEMs are exploring different user experiences on their devices, such as improving photography and communication translation.

The Industry Landscape Dominated by Apple and Google Will Change

Generative AI is rapidly entering smartphones, sparking a new round of growth in the smartphone market and changing the competitive landscape of the industry.

The introduction of AI capabilities may stimulate a new round of growth in the smartphone market, with high-end products growing faster than low-end products. Amon states:

"By early 2024, we expect devices with multiple AI application cases to be launched, which could create a new smartphone upgrade cycle and drive consumers to replace their phones to access these new features."

"Although the smartphone market has matured, Amon remains cautiously optimistic about market growth in 2024, even if it is in single digits. However, if a new upgrade cycle is created through the popularization of AI capabilities, the market size will change significantly, and the upgrade rate will accelerate, with high-end products often growing faster than low-end products."

Amon predicts that this shift will significantly change the industry landscape dominated by Apple and Google. Amon believes:

"By integrating AI models into more smartphones, companies like Microsoft and Meta may regain influence in the mobile industry. They can compete with Apple and Google in the smartphone operating system market through these models. However, achieving this goal requires cooperation between smartphone manufacturers, operating system developers, and application developers, which is not always easy to achieve.

The implementation of AI models on devices will change the industry's collaboration model. Amon pointed out:

"Currently, companies responsible for building and running large AI models (such as OpenAI, Anthropic, Microsoft, Amazon, and Meta) are different from companies that produce smartphones and their operating systems (such as Apple, Google, and Samsung). He mentioned that with the change of platforms, devices can directly access and run AI models provided by these large companies"; "For example, Microsoft's Copilot can run directly on devices, and similarly, models developed by Meta can also be used directly on users' devices. This model will change the way the industry operates, as AI models are no longer limited to the cloud but can run on various devices, making AI applications more efficient and cost-effective. To achieve this, different companies need to collaborate to ensure that the right models are deployed in the right places to provide better user experience and services."

The main battlefield for chip manufacturers is shifting from "training" to "inference". Amon pointed out:

"Currently, the AI market is mainly focused on the "training" phase, which involves training models using large amounts of data, and Nvidia is the main beneficiary in this field. However, as AI models become smaller, capable of running on devices, and focused on specific tasks, the main market for chip manufacturers will shift to "inference," which is the application of models. Data centers are also becoming interested in processors specifically designed for inference of trained models, indicating that the inference market will surpass the training market."

Amon believes that the application of AI technology creates tremendous market potential for device manufacturers, electronic product manufacturers, and chip manufacturers. He said:

"The industry is undergoing changes. Unlike applications downloaded from app stores, AI models will be pre-installed in the device's engine. Mobile platforms will support first-party and third-party models, as well as models from original equipment manufacturers (OEMs). He likened this to OEMs embedding camera functionality into devices, and different AI engines will also be embedded in devices. Integrated AI models will expand the scale of the smartphone market, benefiting both Qualcomm's chip business and licensing business. Devices with AI capabilities require more powerful computing power, which means that users may be more inclined to purchase higher-end devices, resulting in higher profits for Qualcomm. In the PC field, Qualcomm's collaboration with Microsoft to develop a platform that can compete with Apple's Mac computers is a huge boost for Qualcomm's entry into the PC market. In the automotive industry, Qualcomm has already established a strong position through its Snapdragon digital chassis and immersive digital cockpit, and the integration of AI capabilities will further enhance its position in the automotive industry."

Spatial Computing to Become the New Computing Platform

Generative AI is being integrated into VR technology. Amon points out that he is very optimistic about the development potential of augmented reality (AR) and mixed reality (MR) glasses, especially when they use audio and large language models as interactive interfaces. He believes that spatial computing will become the new computing platform and is optimistic about the related developments in the Android ecosystem, including Meta, Google, and Samsung.

Amon is also more bullish on the market prospects of lightweight wearable devices, such as Meta's Ray-Ban smart glasses. He believes that as these devices become lighter, easier to wear, and provide richer content, their usage and sales will increase. He believes that as the market size expands, developers will be willing to invest more money to further enhance and improve the content.

Amon is optimistic about the market prospects of foldable phones. He personally uses a foldable phone and believes that it is very practical, performs well, and is more suitable for pocket carry. He enjoys the experience of answering and hanging up calls with a foldable phone.