Hundred-dollar AI Phone vs Ten-thousand-dollar Flagship: AI Capabilities Have a Gap, but Not Much

06/13 2024 441

Different prices, vastly different experiences?

「Apple Intelligence」

This was Apple's highlight moment at WWDC 2024 yesterday.

If not for Apple's attempt to redefine AI, the headlines of various media outlets the next morning would likely have turned into jokes like "Apple Once Again Poorly Imitating Android" and "iPad Finally Gets a Calculator App," reminiscent of April Fool's Day pranks.

(Image source: Apple)

Fortunately, this update does have substance.

Based on the built-in large model, it can coordinate information from a series of hardware products such as iPhone, iPad, Mac, and provide appropriate assistance while perceiving user needs.

This highly integrated approach to hardware and software ecosystems is indeed befitting of Apple after "giving up on cars and focusing on AI."

Of course, it's not just Apple embracing AI large models.

In the recent past, mobile phone manufacturers have been riding a wave of large models. Almost all major manufacturers such as Xiaomi, Huawei, Samsung, OV, and Honor have successively announced the implementation of large models. The blurring of concepts like 5G, full-screen displays, and foldable screens during phone launches seems to have returned, with AI becoming a new must-have for consumer manufacturers.

It's truly a blurring of senses.

Then, seeing these continuous streams of AI devices, at some point, you start to ask yourself how much difference there really is between these AI devices, or let me put it in a simpler way—

What functional differences would there be between an AI phone costing just a hundred dollars and an AI phone priced at nearly ten thousand dollars?

With this doubt, I carefully selected two products with the largest price difference under the same brand in the market. Is the "AI phone" just a gimmick, and what kind of experience enhancement can it bring at different price points?

Today, I'll take everyone along for a ride.

Products with Vastly Different Prices

You might not believe it.

But this was the most difficult part of the planning.

In fact, it was finding the AI phone with the lowest price under the same brand.

The main reason for this problem is that manufacturers tend to use stripped-down systems on hundred-dollar phones. Compared to the normal version, these systems basically only have version number changes and no substantial functional improvements.

For example, the Redmi 13C / 13R 5G released this year should be the cheapest HyperOS phone.

However, the biggest changes in the updated HyperOS for this product lie in the updates to the system's underlying layers and security patches. Features such as split-screen, negative one-screen, and small floating windows have also been stripped down, let alone the various AI functions built on top of them.

Well, I can only say...

I understand why manufacturers do this, after all, the performance of hundred-dollar phones is highly unlikely to support these features, but it has indeed caused me quite a bit of trouble.

Fortunately, the guys from the Xiaomi community are reliable.

After searching and communicating for half a day, I finally found the cheapest "AI phone" under the Xiaomi ecosystem.

Or perhaps it should be called a "smartphone with AI capabilities," as Lu Wei, Xiaomi's vice president, indeed doesn't like the term "AI phone."

Come, meet the Redmi Note 11T Pro that I carefully selected.

(Image source: Leitech)

As a product released in May 2022, the Redmi Note 11T Pro is equipped with the Dimensity 8100 processor, has the fifth-generation MediaTek APU, and the MIUI 13 system it shipped with has been officially updated to HyperOS 1.0.2 in recent days.

Most importantly, this phone is currently priced at around a thousand yuan for a brand-new 12+256G model on Pinduoduo, while the price for a low-end second-hand model is only around 500 yuan, making it one of the cheapest products with HyperOS AI capabilities.

Considering that the Dimensity 8100 is still active in new AI phones after two iterations, my expectations for it are not low.

For comparison, let's bring up the Xiaomi 14 Ultra, which was released in February 2024.

(Image source: Xiaomi)

On one side is the Redmi thousand-dollar phone released two years ago and just upgraded to HyperOS, and on the other side is the flagship with HyperOS out of the box, equipped with the Snapdragon 8 Gen3 flagship processor, theoretically possessing all of Xiaomi's current AI capabilities.

This comparison really makes one a bit excited.

AI Assistant: Identical

When it comes to AI features that are most closely related to everyone's daily lives, I think it's probably the "AI Assistant" that comes with each manufacturer.

After all, even when there wasn't much AI capability, people would still have fun teasing their voice assistants.

And in Xiaomi's case, it's the "Xiaomi Xiaoai" powered by large models.

As a flagship phone that comes with HyperOS out of the box, the Xiaomi 14 Ultra's "Xiaomi Xiaoai" is ready to use right out of the box, while the Redmi Note11T Pro can be upgraded to the latest version of "Xiaomi Xiaoai" after updating to HyperOS.

In terms of activation methods, there are many ways to wake up "Xiaomi Xiaoai" on Xiaomi/Redmi phones, including basic methods like voice activation, long-pressing the power button for 1 second and releasing, wired/Bluetooth headset activation, etc. After changing the system navigation mode to classic navigation keys, you can set long-pressing the Home button, Menu button, Back button, or a custom key combination to wake it up, and it can even be activated by tapping the back.

(Image source: Leitech)

In terms of functional content, the two are essentially identical, meaning that the Redmi Note11T Pro can utilize all the capabilities provided by the large model skills.

After enabling continuous dialogue in the "Xiaomi Xiaoai Lab," even the Redmi Note11T Pro can achieve continuous dialogue effects with Xiaomi Xiaoai after a single wake-up, able to more naturally understand contextual content, significantly improving the communication experience with Xiaoai's voice.

(Image source: Leitech)

Let's try a basic cross-app operation first. I asked Xiaomi Xiaoai to "help me record an appointment in the calendar for 9:30 AM this Saturday to go to Haizhu Xintiandi" and, after automatically setting it, asked it to help me plan a route to Haizhu Xintiandi.

(Image source: Leitech)

Just like before, there were no issues with app invocation, but now voice commands can be expressed more smoothly and complexly.

I must say, it's quite enjoyable to use.

As for scheduling appointments...

(Image source: Leitech)

Or text generation, there are no issues.

(Image source: Leitech)

In summary, in terms of Xiaomi Xiaoai's large model performance.

Currently, the Redmi Note11T Pro can indeed obtain the same experience as Xiaomi's top-of-the-line flagship.

Image Generation: The Ten-thousand-dollar Flagship Wins

Just testing text generation is not enough.

Let's simply ask the large model to generate an image and try it out.

In short, both products have no problem with this request. The Redmi Note11T Pro can also respond to the planned instruction, outputting AI images with a resolution of 768*768, and there is no difference in quality compared to the Xiaomi 14 Ultra.

(Image source: Leitech)

Of course, just because both can generate images doesn't mean the two phones have the same capabilities.

By opening the "Xiaomi Xiaoai Voice" - "Xiaomi Xiaoai Lab" - "Image Large Model" on the Xiaomi 14 Ultra, you can click to download the 1.2G MiAI Text-to-Image model.

(Image source: Leitech)

In this mode, users can use the text-to-image function an unlimited number of times. Essentially, it's a lightweight Stable Diffusion model for local computing, but the generated content still needs to go through cloud-based auditing, so it cannot be used in offline mode.

In contrast, the Redmi Note11T Pro, which cannot download MiAI, can only use the cloud-based text-to-image function, which is actually faster in terms of generation speed but has a daily usage limit. I tried it and found that the maximum is no more than 3 times.

Ah, this is quite reasonable, the positioning gap is immediately apparent.

Let's digress a bit.

Although Xiaomi may be one of the first domestic manufacturers to implement text-to-image functionality, the current related features are somewhat lagging behind the development progress of other manufacturers.

For example, long-sentence image generation is possible with Xiao Bu but not with Xiaomi Xiaoai.

Xiaomi Xiaoai can only recognize single phrases, regardless of how long the prefix is, it must be within that phrase.

Or for generating wallpapers in specific styles, Xiao V can do it, but Xiaomi Xiaoai cannot.

Xiaomi Xiaoai can only imitate the styles of well-known painters like Van Gogh and Monet, limiting its creative space.

(Image source: Leitech)

Of course, the image generation functions built into domestic phones are mainly aimed at giving consumers a taste of novelty.

In terms of freedom, generation effect, and speed, these functions are inferior to third-party image generation tools. Some manufacturers even have explicit generation limits, hoping to achieve secondary payments, which is truly ingenious for revenue generation.

My evaluation is—

If you really have the extra money, you might as well try using local Stable Diffusion or GPT4.0-integrated DALL.E.

Image Editing: The Hundred-dollar Phone Loses Slightly

Next, let's take a look at two AI features that I personally consider to be the most useful.

The first is image editing.

Many people have a problem when taking photos outside, which is "intruding passersby"; and as digital editors, we often encounter similar issues when taking sample photos or reviewing products.

For this reason, AI removal functions that can eliminate passersby and clutter have become a fixture in manufacturers' launches, often occupying a central position during AI feature introductions.

Yes, Apple is no exception.

However, this feature simply doesn't exist on the Redmi Note11T Pro.

Currently, the photo album editing app on this phone is still the old version, so it only has the most basic magic erase function and not the flagship's Magic Erase Pro feature.

I tried a random image, and the results are as follows:

(Left: Smart Erase, Right: Smart Erase Pro)

Here, the gap between the two products becomes apparent.

However, the Xiaomi 14 Ultra also has issues of "gilding the lily," which is a common problem with current AI removal.

Of course, the same is true for other flagship-exclusive editing features.

For example, the very useful "AI Image Expansion" does not exist on the Redmi Note11T, and even due to the existence of cloud-based auditing, there is no way to bypass it—

Is that really the case?

Well, not quite.

In March this year, Xiaomi Xiaoai launched the testing of "multi-modal" related functions, which can achieve functions like image question answering or image editing for user-uploaded images.

With this feature, the Redmi Note11T Pro can also achieve advanced

Solemnly declare: the copyright of this article belongs to the original author. The reprinted article is only for the purpose of spreading more information. If the author's information is marked incorrectly, please contact us immediately to modify or delete it. Thank you.