ET Explainer: How GenAI could breathe new life into smartphones

What makes a GenAI smartphone? That’s the question a lot of consumers will have to resolve in the coming months when checking out a new model as smartphone brands are expected to up the marketing hyperbole around the emerging technology, blurring the lines between what is perhaps pre-programmed responses to certain actions, and generating original content using natural language prompts.

What is GenAI?
Generative AI, or GenAI for short, is a subfield of artificial intelligence that focuses on creating new content, rather than simply analysing or classifying existing data. Think of it as a digital artist or a writer, using its knowledge of existing work to produce something entirely original. It relies on AI models that are trained on massive datasets (large language models or LLMs) of text, images, code, or other media, that helps the model learn the underlying patterns and relationships within the data, enough to churn out something unique and original based on its understanding from the datasets it was trained on when prompted with natural language instructions.


The technology boomed in 2022 with the advent of apps such as ChatGPT, Mid Journey, Stable Diffusion, Google Bard, and the likes. The potential of GenAI apps are vast and constantly evolving, and has the potential to revolutionise industries like art, design, entertainment, and even healthcare.

While ethical concerns remain around the technology being trained on copyrighted material alongside its use to create realistic deepfakes for malicious purposes, a bigger constraint is the amount of compute needed to run these apps. Given how viral the early applications have become, there was a need to distribute the compute resources across the cloud and devices. Imagine if a billion smartphones started generating content using ChatGPT at the same time. Without the right mix of distribution of compute resources, it will not be possible. Chipset makers Qualcomm and MediaTek have already solved this by launching chipsets that can process GenAI models locally in smartphones, while iPhone-maker Apple published a research paper this month that makes processing LLMs easier with existing memory standards.

Defining a GenAI smartphone
Industry experts believe the smartphone industry is set to undergo a revolutionary change with the introduction of GenAI devices. Counterpoint Research estimated around 40% of all smartphones, surpassing half a billion units in shipments, will be GenAI-powered, with Samsung expected to capture half of the market.

But what makes a GenAI smartphone? The research firm defines it as a subset of AI smartphones that uses GenAI to create original content, instead of just providing pre-programmed responses or performing predefined tasks. “These devices will run size-optimised AI models natively and come with certain hardware specifications. Our short-term GenAI landscape sees OEM roadmaps touching on four main areas—info provisioning, image building, live translation, and personal assistant applications,” the research firm said.

In essence, any smartphone that has a large language model residing in its memory, which it leverages to deliver original content after being prompted by the user, is a GenAI smartphone.

How are smartphone brands using GenAI?
The latest Google Pixel 8 phone offers the first glimpse of a GenAI smartphone. Using a proprietary processor, and its own LLM that resides on-device, the Pixel 8 Pro can summarise content you have recorded using the voice recorder. It suggests high-quality responses to chats on WhatsApp using the Google Keyboard. It also uses a hybrid approach to make videos and photos shot from the camera a lot better. Was the subject not smiling in the photo? A click of a button can now solve it. Pets can’t stand still while taking a photo? The phone can take an existing blurry shot and create a crystal clear image out of it.

The use cases are going to be far and wide.

At the Lenovo Tech World 2023, Motorola showcased features such as generative theming and a sophisticated personal assistant. Focusing around personalisation, the Lenovo-owned smartphone brand will soon leverage an on-device AI model to generate wallpapers that match with the outfit users are wearing on a particular day, while a personal AI assistant can answer questions, draft messages, schedule tasks, and more. It can also scan documents, long emails, or reports in real-time and generate summaries. More interestingly, it can automatically obfuscate one’s private information from images, or documents, when they are uploaded on social media.

On the other side of the pond, Samsung has announced its own foundational AI model called Gauss designed to run locally on smartphones and produce text, code, and images. Similar to Motorola, Samsung’s on-device AI will help with composing emails, summarising documents, and translating content. The Korean corporate showcased a live demo of translating a phone call in real-time while the conversation is ongoing, removing the language barrier between two foreigners trying to communicate.