Blog

Generative AI is ushering in a new era in the mobile industry, exciting users but also raising big questions about privacy and security.

According to CNBC, many people are captivated by generative AI, using the latest tools to handle everything from personal issues to work tasks. From OpenAI's ChatGPT to Google's Gemini and Microsoft's Copilot, generative AI models are increasingly available on smartphones, making them more widespread. However, many overlook the potential privacy implications.

Challenges of Generative AI on Mobile

To respond to queries or generate content in real-time, generative AI models like ChatGPT require massive processing power. This computing demand poses a significant challenge for both mobile processors and smartphone battery life.

On-device AI capabilities are also limited by hardware. Only the latest models equipped with AI chips can run generative AI directly on the device. For instance, Apple Intelligence only operates on iPhone 15 Pro and newer models.

To address this limitation, a popular option is to run AI models on the cloud. Through an Internet connection, users can access vast data stores and the powerful computing resources of cloud servers to use generative AI.

However, this approach requires users to stay connected to the Internet, with their data sent to a third party instead of remaining on the device. This raises privacy concerns, especially as generative AI is still new and lacks specific regulations addressing user privacy.

With cloud-based AI, users also face additional monthly service fees. Over time, this approach creates a financial barrier for the average user.

Hybrid AI – Combining Online and Offline

In a report released in May 2023 titled "The Future of AI is Hybrid," Qualcomm suggests that for AI to reach its full potential, models need flexible processing on both the cloud and the device. This combination can improve cost, energy efficiency, reliability, latency, and privacy.

Forbes notes that while AI requires powerful, stable machines for model training and inference—an advantage of the cloud—it also needs to operate immediately, privately, and with personalization, which are strengths of on-device AI. “No matter how much we enjoy AI, we can’t rely on the Internet or the cloud 24/7. A hybrid solution is the best fit,” the publication concludes.

Samsung has also adopted this approach with Galaxy AI. The company introduced the concept of Hybrid AI at the Galaxy S24 launch earlier this year, partnering with Google to bring generative AI to its flagship smartphone line and then expanding it to older devices. During the Galaxy Z Fold6 and Z Flip6 event on July 10, Samsung introduced new AI features and maintained its commitment to the hybrid approach.

Samsung’s approach combines the power of on-device hardware with cloud connectivity to handle AI tasks. For features involving personal or sensitive data, like live call translation, photo editing, health assistance, interpreting, and voice-to-text conversion, AI runs locally on the device to ensure data is not transmitted externally. This allows users to leverage generative AI even without an Internet connection.

However, for tasks requiring heavy computational power and frequent updates, like search zoning and generating images or videos from text, users access the latest AI models via the Internet.

TechRadar quotes Patrick Chomet, Samsung’s head of customer experience, saying, “We believe hybrid AI is a practical and reliable approach to help users balance smooth, flexible experiences while ensuring privacy, security, and energy efficiency.”

Counterpoint Research observes that Samsung’s early adoption of hybrid is giving it a leading edge in the generative AI race on mobile. “Hybrid AI brings numerous innovations to users’ daily experiences. It’s a user-centric leap forward,” Digitimes quotes Tarun Pathak, Director at Counterpoint Research.

Another advantage of hybrid is that it makes generative AI more accessible, independent of hardware. With the One UI 6.1 update, GSMArena estimates Samsung could bring generative AI features to hundreds of millions of older devices. In the future, even budget smartphones may be able to experience generative AI features.