AI Race Heats Up as Xiaomi, OPPO, Vivo Join Google and Huawei in Pushing for Adoption

It’s no surprise that every phone nowadays uses some sort of machine learning to improve the overall user experience, but AI has now got to a point where manufacturers are now using large scale AI models to train and improve their phone’s operating system.

When it comes to implementing AI, Google is really open about the fact that AI is the future, and have been active in implementing in their software. Just today, a promo video leak for the Pixel 8 showcased what Google calls the “Audio Magic Eraser.” This is similar to the traditional Magic Eraser, but instead of cutting out objects from an image, Audio Magic Eraser cuts out unnecessary voices from a video.

The same is the case with Huawei that had just recently showcased the new Celia AI voice assistant, “powered by the large language model (LLM) Pangu 3.0 AI.” Like we discussed in our article, Celia is basically a much more refined, and an handy way of using something like ChatGPT.

Today, insider Digital Chat Station posted on Weibo about how other manufacturers like Xiaomi, OPPO, and Vivo would be joining the AI adaptation race.

Well, it has recently been revealed that Xiaomi has developed a large language model called MiLM-6B, based on a whopping 6.4 billion parameters. This makes it one of the largest language models in the world, and it is ranked 9th on the C-Eval list of large language models.

The C-Eval list is a ranking of large language models based on their performance on a variety of tasks, such as question answering, summarization, and translation.

Just to be clear, none of this has been announced publicly. At this moment,MiLM-6B is still under development, but it has already shown promising results on a number of tasks. For example, it was able to achieve great results on the GLUE benchmark, which is a collection of natural language understanding tasks.

We don’t know why Xiaomi has not yet opened up about this, or the reason behind developing this model, but it’s possible that it will be used for virtual assistants, chatbots, and machine translations. The large number of parameters in MiLM-6B gives it the ability to handle complex tasks that would be difficult for smaller models.

In a recent C-Eval benchmark, the model ranked 9th overall, but it was the first in its class for models with the same number of parameters. In another benchmark that assesses Chinese language ability, Xiaomi’s model ranked first in the Chinese-only category. However, in the multilingual category, GPT-4.0 came out on top.

Earlier this year, in April, Alibaba Cloud and OPPO also announced a partnership to jointly build large-scale AI technologies. This will allow OPPO to develop more sophisticated AI-powered features for its smartphones, since they had just recently shut down their chip business, so the competitive edge had to be found from what’s remaining.

Vivo is also reportedly working on its own AI-powered technologies, including a pre-training model called 3m-text. 3m-text is a large language model with over 3 million parameters, which also makes it one of the largest language models in the world.

My take

It’s great that AI is being implemented by leading smartphone manufacturers, but I’m of the opinion that this won’t take away the fact that a normal user would still be more inclined towards the general software experience and raw performance, and it seems like, from the perspective of these companies, they’re being complacent, and if the trend continues, will lack behind the industry giants, like Apple.

I never thought I’d say this, but Apple seems to be leading the way in terms of innovation right now. They’re always coming up with new features, both on the hardware side, and software (iOS 16 and iOS 17). With almost the entire TSMC’s 3nm chips reserved, it’ll be no surprise if A17 absolutely crushes SD 8 Gen 3 this year too.

This is all we know for now, but rest assured that we will keep you updated as new information becomes available.