Generative AI Turns a Tool into a Companion

Siri made its debut as far back as 2010, with Amazon’s Alexa following in 2014, and Google Assistant in 2016. These early personal assistants found varying levels of success, extending from smartphones to speakers, TVs and various smart devices, and are still widely used in many households today.

However, their initial rapid progress has stagnated somewhat in recent years as their usefulness has been primarily limited to “command and control” applications. Features like hands-free music control, setting alarms and managing smart devices are good examples, but assistants have fallen short in providing a conversational experience, context, and information search capabilities. Interactions often conclude with a frustratingly simple, “Here’s what I found on the web.” But with generative artificial intelligence (AI), and particularly on-device generative AI, a promising upgrade to these limitations is on the horizon.

In his keynote address at this year’s Snapdragon Summit, which I attended with my colleague Geoff Blaber, Qualcomm CEO Cristiano Amon unveiled an ambitious vision for the future of mobile computing. He discussed the evolution of the user experience, moving from voice calls to text and e-mail to data and apps. However, Mr Amon emphasized that it’s on-device generative AI that will usher in the next evolution of the user experience. It’s an audacious claim, and Qualcomm is banking on it to revitalize sales of new phones, which dropped 11% in 2022 and are expected to shrink by another 5% this year. People are now less inclined to upgrade to the latest models and are instead opting to keep their phones for longer or seek better deals in the second-hand market. See here for more details of CCS Insight’s research into consumer habits and preferences.

Generative AI, exemplified by the rapid ascent of OpenAI’s ChatGPT, has created a buzz in the tech world, with major players like Microsoft, Google, Meta and Amazon all showcasing their efforts in this domain. Although there’s no doubting the technology’s huge societal impact, we believe it’ll face a reality check in 2024 as high costs and the complexities of privacy, regulation and access to cloud-based services all come to the fore. A principal way to mitigate the challenges of cost, privacy, security and performance is to enable more on-device processing, a capability that Qualcomm’s Snapdragon 8 Gen 3 mobile system-on-chip brings to the market for the first time.

During the Snapdragon Summit, Qualcomm gave several demonstrations showcasing the advantages of running generative AI on a device. The most attractive one for users is the combination of on-device large language models with personalized and contextual insights from Qualcomm’s Sensing Hub, which collates data from cameras, microphones, accelerometers, gyroscopes, GPS, Wi-Fi and Bluetooth (see Figure 1).

Figure 1. Presentation of Qualcomm Sensing Hub at Snapdragon Summit 2023.

This combination promises to make on-device assistants more natural, conversational and, most importantly, to provide highly relevant and contextual information. As these assistants are used more widely and integrate new large language and multimodal models, their performance and responsiveness will see a significant leap forward, ushering in what we can describe as “voice assistants 2.0”.

The potential applications of these on-device assistants are vast, with Qualcomm’s proposing a scenario in which you could simply ask your assistant to book a vacation. It could not only identify suitable dates using your calendar but also choose a preferred airline and direct you to the relevant booking application, just needing you to confirm the details. The company also showed an example that provided instructions on how to prepare for surfing, based on what the device knew about the user’s location, age and fitness level (see Figure 2).

Figure 2. Demonstration of on-device personalization at Snapdragon Summit 2023.

Generative AI’s capabilities extend beyond language and voice models to images and videos, which will be particularly beneficial for creative processes. Content creators can use an on-device personal assistant to streamline the manual processes involved in setting up digital content platforms, from creating presentation decks to designing websites and managing social media pages.

Despite these intriguing possibilities, several challenges need to be addressed before this becomes an everyday reality. Privacy remains a significant concern, as many people remain wary of how personal information collected from cameras, microphones and apps will be used and accessed. Although Qualcomm argues that on-device AI is more private, as it reduces the need for constant cloud connectivity, concerns about data storage, usage and transparency persist, demanding regulation and codes of conduct. AI regulation is trailing far behind technology advances and may stall AI progress while it catches up.

Another hurdle that may limit AI’s usefulness to drive upgrades in the short-term is Qualcomm’s identity as a chipmaker rather than a software company. The company is relying on its partners and a broad community of developers to fully exploit these new capabilities and bring innovative experiences to life — a process that will take time.

With support from leading names like Google, Meta and Microsoft, Qualcomm has the backing to make a significant impact in this ecosystem, potentially taking a leading role in the developer community, similar to the position enjoyed by Apple.

Generative AI, especially when applied on-device, represents a transformative shift for personal assistants and marks a critical point in the evolution of mobile computing. Qualcomm’s Snapdragon 8 Gen 3 goes a long way to integrate powerful language models and sensor data, offering a future where personal assistants become intelligent companions, fostering natural, context-aware interactions.

This journey is not without challenges, and balancing convenience and privacy will be crucial. In addition, although the hardware looks to be capable of such experiences, it remains to be seen how quickly and well-integrated software development platforms and applications can be built to prompt people to buy a new device. From what we saw at Snapdragon Summit, it may take a couple more upgrade cycles yet. Nevertheless, the potential of generative AI to turn our smartphones into intelligent companions remains promising.