AI’s Magic Call: how AI will make smartphones smarter

Generative AI is coming to a smartphone near you. The stars have aligned. All pieces of the puzzle are falling in place. From 2024 onwards, the breakthrough technology is expected to be a key selling point, featuring prominently in marketing communication from brands, as smartphone makers look to leverage the breakthrough technology to chart another phase of growth. “Innovation has so far been on the connectivity side or the camera, both of which have improved by leaps and bounds thanks to AI. But now, with generative AI, smartphones will become smarter, because you will be able to do a lot of things without even opening any apps. You can simply issue voice commands to accomplish tasks,” said Neil Shah, co-founder of the consulting firm Counterpoint Research.

Shah added that generative AI features will become a lot more pervasive in the coming years, and it won’t just be from Google, Microsoft or other tech titans. Smartphone brands too are working on their own AI models.

“The importance of succeeding in on-device AI cannot be understated. It is an opportunity to reshape not just the smartphone but also the entire consumer electronics line-up,” said Sheng Win Chow, a Singapore-based analyst for market analysis firm Canalys. Chow said that strides in on-device AI will take smartphone brands away from the “peak smartphone” era that used to describe a lack of innovation coming from the product category, a time when mass market devices became increasingly commoditised with little differentiation in features and hardware.

ETtech

“This leaves vendors with few options but to compete on price, resulting in a race to the bottom and shrinking margins,” Chow added. There are glimpses of this trend already. Lenovo-owned Motorola has developed its own large language model (LLM) that can power a personal assistant for both PCs and smartphones, answer questions, draft messages and schedule tasks, all without relying on the cloud, thereby alleviating data privacy concerns.

“AI is going to transform things, whether you use it in multiple apps or in a single app. For example, your photos are going to get better, and you can remove unwanted noise from audio files. AI is all about giving the consumer the right benefits. We want to focus on productivity, health, security and general entertainment,” said Prashant Mani, executive director Asia Pacific, Motorola.

Not wanting to be left behind, Google also announced Gemini, an improved large language model — albeit with controversies over inflating benchmark numbers — the smallest of which will reside natively in smartphones to power Gen AI tasks.

Earlier this year, Android 14 was announced with a host of generative AI use cases. This includes the ability to auto-compose messages and emails, or generate unique wallpapers. Users also want AI with simple text prompts. Google’s latest Pixel phones take this a notch higher by including a magic eraser that can remove objects from photos and videos and change facial expressions after a photo has already been clicked.

TALKING CHIPSETS
Powering this all will be the latest generation chipsets from Qualcomm and MediaTek, the two major smartphone chip vendors.

Their latest iterations of flagship chipsets include support to run AI models in the device itself. For instance, the latest Snapdragon 8 Gen 3 — that’s set to power flagship Android smartphones in 2024 — features an AI engine with double the performance than its predecessor, and a 40 per cent improvement in performance per watt to run AI models.

This chipset can run up to 10 billion parameters ondevice, removing the need to connect to the cloud to deliver AI-infused results. Similarly, MediaTek’s latest Dimensity 9300 chipset features an AI co-processor that achieves a 45 per cent reduction in power consumption while improving performance eight times over its predecessor.

The chipset is designed for generative AI tasks, with support for models with up to 33 billion parameters. Bringing the capability to run AI models ondevice was important for wide-scale adoption of generative AI, Shah said.

“Right now, generative AI is quite costly in terms of training the models and deploying them. No one is making money except for the hyperscalers eventually in the long run,” Shah said. He added that even from an ecological perspective, having four billion smartphone users running generative AI tasks every minute of every day can be a carbon foot print nightmare.

That said, generative AI can also become a monetisation opportunity for smartphone brands and app developers. Shah gave the example of Adobe, which recent ly integrated generative AI in apps such as Photoshop, which requires a premium subscription to use. More such apps are expected to come to smartphones, optimised to run more efficiently with on-device processing, thereby opening new paths of monetisation, experts said.

DECISIVE DIRECTION
With the chips falling in place, major tech companies have already signalled the direction smartphones are going to take next year. Samsung, for instance, has announced a slew of generative AI tools for its latest Android version that is set to power its next lineup of flagship phones earlier this month.

Naming it Galaxy AI, Samsung announced in November that its suite of comprehensive AI experience, powered by both on-device AI and cloud-based AI, will for example, be able to live-translate phone calls, with translated text appearing realtime as you speak. The company said in its recent earnings call that the technology will likely reach customers next year.

Meanwhile, Chinese smartphone brands such as Oppo are also working on their own LLMs, a key ingredient in the generative AI cookbook, that will be powering more capable voice assistants for the brand, the company announced earlier this month. Xiaomi and Vivo too have showcased their versions of generative AI-powered voice assistants, beating the likes of Google and Microsoft to the punch in this aspect. That said, these voice assistants are currently restricted to the China market.

“Artificial intelligence will be at the centre of marketing communications in the coming year. But brands will need to inform users about the abilities of generative AI, for which they will require the entire ecosystem of chip vendors, such as MediaTek and Qualcomm, as also Google and other big tech companies, to come together,” said Navkendar Singh, associate vice-president, IDC India.

He added that with 5G and high performance now a staple in smartphones, brands will need another plank to sell their devices to consumers, and AI is the answer. That said, however, not all features coming out of their stables can be termed generative AI use-cases, despite the brands themselves claiming so.

In India, the use cases could be different, which experts believe will be more in tune with the needs of the world’s second largest smartphone market. This could range from realtime language translations, more efficient voice assistants, music recognition and more.

However, to deliver India-specific use cases will require the stakeholders to develop India-specific language models. Shah said this can lead to an increase in R&D activity in India in the coming years by smartphone brands and app developers.