It's no secret that "generative AI" has become one of the most well-known selling points in the smartphone industry today.
Smartphones, regardless of price range, now often advertise themselves as integrating "generative AI" capabilities. Some of them are "voice agents" that can have a natural and smooth conversation with users and generate various suggestions, while others are "creative assistants" that can automatically help users summarize articles, edit text, or automatically stitch photos into videos.
Of course, the most impressive generative AI features of mobile phones may be those "AI elimination" and "AI retouching" features. Even consumers who don't understand professional P-image technology at all can easily remove flaws in their photos with the help of generative AI through simple selection, and even achieve the effect of almost equivalent to "reshooting".
But it is an objective fact that, at least for today's smartphones, the computing power required for functions such as "generative AI retouching", "generative text summarization", and "generative knowledge base Q&A" generally greatly exceeds their performance levels. So this also means that at this stage, no matter which mobile phone manufacturer or which model, the "generative AI" functions they carry almost rely on cloud computing power and large models running on cloud servers.
The question naturally arises, which is the solution for the cloud-based generative AI functions on these mobile phones.
For the mobile phone industry, third-party AI is necessary
Some friends may say, what's there to ask, who makes mobile phones, don't they all use their own cloud computing power and algorithms?
Not necessarily. On the one hand, the mobile phone market is not only a few well-known head brands, in fact, there are still many second- and third-tier manufacturers. Although their sales are objectively not particularly low, and their product positioning is not necessarily extremely "non-mainstream", it is impossible for these manufacturers to completely ignore the trend of "generative AI", but they do not have the corresponding technology to achieve it on their own.
On the other hand, for some large international manufacturers, they also have to seek to build their own servers in China, or use a simpler way to solve the problem, that is, directly cooperate with domestic manufacturers, borrow their algorithms and computing power, and form a set of their own "special version" of AI applications.
As a typical example, some friends may know that the generative AI function of Samsung's mobile phones in overseas markets is the result of their cooperation with Google. Even some AI features on overseas Samsung mobile phones are now newer and more functional than those on Google's own models
However, it is clear that the AI technology system of "overseas Samsung" cannot be directly applied to domestic products. Therefore, the final result is that the Bank of China retains Samsung's own AI computing framework, but instead uses the technical systems from Baidu, WPS, and the Meitu App and other domestic software manufacturers, so as to complete the "same overseas model" generative AI function.
Registering a trademark for retouching AI, Google can't resist the temptation
Interestingly, when it comes to Google, it seems that they are now starting to take their "generative AI" features seriously. Just recently, Google was revealed to have registered a trademark called "RealFill", referring to their self-developed generative photo restoration algorithm (and features).
According to the relevant patents, RealFill can use multiple photos with different angles and different compositions to generate a "perfect photo" with reasonable composition and clear subject. Obviously, it is suitable for those who love to take pictures, but their skills are not good, and they often take a lot of photos but can't pick out a good one.
But the most interesting thing about this is that Google has actually been involved in smartphone imaging algorithms for a long time, and has had technical cooperation with some third-party mobile phone manufacturers many years ago. But at that time, Google did not put forward the need for "brand exposure" at all, and mobile phone manufacturers did not need to let consumers know that they were using "Google Camera" or Google algorithms.
The pixel series has also been used by other brands in the past, but there is no promotion
But now, at least when it comes to generative AI retouching, it's clear that Google is starting to take "brand image" seriously. In other words, future mobile phone manufacturers that adopt this technology may need to show the brand logos of Google and "RealFill" in their software interfaces, or at least at the press conference.
AI has become an outlet in the software industry, but the "window period" may not be too long
So what does all this mean? To put it simply, generative AI on mobile phones, regardless of its actual technical level, has actually become a "trend" with great publicity value.
Especially in today's context, on the one hand, it is impossible for mobile phone manufacturers to achieve self-developed algorithms and self-built computing power. On the other hand, since most of the current mobile phone generative AI functions are concentrated in the fields of search, Q&A, text processing, and image editing, it is natural that some software companies that already have the corresponding technical accumulation will suddenly become "sweet and sweet" in the industry.
Of course, from another point of view, for these software companies that now have technology and computing power, this is likely to be their "last chance". Because if we don't take advantage of the current time to establish our own technical standards and product image (as Google intends to do), then in a few years, the performance of smartphones will evolve to the point where it can really achieve complete on-device AI, and there is really no need for cloud computing power at all. At that time, software and algorithm providers are likely to lose their potential customers again and return to the "transparent" status of previous years.
[The picture in this article comes from the Internet]