The best smartphones for AI enthusiasts

Black Friday – the biggest shopping day in the US – is fast approaching, and it's great news if you're looking for the smartphone. Retailers like Target and Best Buy offer hundreds of dollars to purchase flagship Samsung Galaxy handsets. There is a "one for one for one without one" offer on the LG G7 ThinQ at T-Mobile. And even the range of Apple's iPhone 2018, which was unveiled just two months ago, is experiencing price reductions.

It's almost too good, especially if you're not yet attached to a brand or model. Conventional wisdom would like you to compare screen resolutions, processors, accessories and other hardware components to find the phone that interests you the most. But this being the VentureBeat AI chain, we offer a different metric: AI features. In our smartphone buying guide, we compare the best phones of the year in a battle of the spirit of artificial intelligence and crowned champions in three categories: the smartphone with the best AI chip , the best AI camera features and the best AI Assistant alternative.

Smartphone with the best AI chip

iPhone XS, iPhone XS Max and iPhone XR ($ 749, $ 999, $ 1,099)

The trio of state-of-the-art Apple smartphones – the iPhone XS, XS Max and XR – has something in common: the A12 Bionic, the system-on-a-chip architecture with 7 custom architecture nanometers that Apple has described as "the most powerful" ever designed smartphones. It has six cores – two high-performance cores and four high-power cores – as well as a four-core graphics processor. But the real engine of the neuron is undoubtedly the flagship processor: a machine learning processor dedicated to eight cores.

This is Apple's second-generation neural engine, the first bundled with the iPhone X's Aion Bionic chip, and a further step forward from its predecessor.

The neural engine can perform 5 trillion operations per second compared to the latest 600 billion silicon, and applications created with Core ML 2, the latest version of the machine learning framework. Apple, can be operated up to nine times faster and faster. 30% faster using one-tenth of the power. In addition, Apple's AI chip takes advantage of a "smart computing" system that can automatically determine whether to run algorithms on the core processor cores of the A12, GPU cores, neural engine or a combination of the three.

That's fine, but what is it really The neural engine is distinguished from the competition by the richness of its optimized features. These include Face ID, Animoiji and Memoji, Portrait Lighting and Apple's ARKit 2.0 augmented reality framework, in addition to the enhanced Portrait mode of the iPhone. When you press the shutter button, neural engine-accelerated algorithms attempt to determine the type of scene being photographed and to distinguish a person from the background. This understanding of the depth informed by the AI ​​makes it possible to modify the blur and the sharpness after the production.

On the third side of the equation, developers can run code on the neural engine. The basketball app of Nex Team, San Jose's software company, HomeCourt, exploits it to track and record in real time shots, misfires and a player's position on field.

Neural engine application development benefits from Core ML 2. The Core ML 2, for the uninitiated, allows developers to load device-based machine learning models to an iPhone or iPad, or to convert templates from frames such as XGBoost, Keras, LibSVM, scikit. -Learn, and Facebook's Caffe and Caffe2. Using a technique called batch prediction, Core ML 2 is 30% faster than Core ML, and allows developers to reduce the size of trained machine learning models by up to 75% through quantification.

Finalist: Huawei Mate 20 Pro (~ $ 1,150)

Apple's latest generation neural engine is certainly an impressive chip. The same goes for Huawei's Kirin 980 neural processing unit (NPU), the 7-nanometer system-on-chip in the Mate 20 Pro.

This is the second iteration of the NPU – the first it was launched in September 2017 on the Kirin 970 – and it has been specifically optimized for the type of vector computing that is the heart of the machine learning models. The two NPUs in the chip (one compared to the Kirin 970) can recognize up to 4,500 images per minute, compared to 2,371 images of Qualcomm Snapdragon 845 and 1,458 of the A11 Bionic. In addition, it offers superior object recognition, real-time image processing and real-time object segmentation, delivering up to 135% better performance in performance tests such as ResNet and Inception v3. while consuming 88% less energy than the Snapdragon 845.

On the Mate 20 Pro, the artificial intelligence system running on the NPU intelligently speeds up the GPU's clock speed during intense gaming sessions, minimizes system delays and offers "softer outdoor gaming experiences" in the game. areas where signals are weak. In addition, Huawei says that its heterogeneous IT structure, HiAI, can automatically split voice recognition, natural language processing and computer vision workloads on the chip.

The NPU also powers functions such as AI Color, a Sin City-Inspired effect that keeps a subject in color while all the rest of the scene is in black and white, and a 3D object-scanning tool – Live Object – that recreates real-world objects in digital environments. Pro Mate 20 Pro's Animoji Live Emoji and 3D Face Unlock uses NPU for face tracking, while its AI AI 2.0 camera mode uses it to automatically recognize scenes and objects and adjust settings such as the macro and the angle of the lens. In addition, AI Zoom uses NPU-accelerated object tracking to automatically zoom in and out on subjects. The video bokeh highlights the subject in the foreground while making the background blurry; and Highlights generates video spots edited around recognized faces.

Third party developers can access the NPU via the HiAI library and the Huawei API, and some have already done so. The Microsoft Translator application, for example, uses it for tasks such as scanning and translating words into images, and the Prisma image editing application uses it to perform a style transfer on the screen. device in seconds.

Smartphone with the best AI camera features

Google Pixel 3 and 3 XL ($ 799, $ 899)

As we wrote in our article on Google Pixel 3, many of the best features of the phone, whether battery saving or call filtering, are optimized with AI. And this is particularly evident on the front of photography.

The Pixel 3 features two selfie cameras (two 8 MPo sensors) and, at the back of the phone, a 12.2-megapixel sensor with a single camera. The rear camera is the main beneficiary of the truly awesome Google AI. The Top Shot feature of Pixel 3 captures a burst picture before and after you've pressed the shutter button and uses the built-in AI to take the best picture possible. The photobooth taps automatically learns to take the "best" photos – that is, to be blurry and well-lit – automatically.

That's why, despite the fact that only one camera is compared to the LG V40 and the Huawei P20, which has three, the Pixel 3 takes some of the best photos we've seen from a smartphone.

For another example of AI's photographic prowess, look no further than Google's portrait mode, which filters facial features, clothing and skin in portrait images with a machine learning model that detects differences. color and edge at the pixel level. (This works on both the front and back cameras.) Another example is HDR +, Google's internal approach to high dynamic range. It captures between 9 and 15 frames in quick succession and merges them to minimize noise, and leaves them intelligently underexposed to help keep colors saturated in low light. And another example is Super Res Zoom, which merges the frames to form a higher resolution image, which is "pretty much competitive compared to the 2x optical zoom lenses of many other smartphones," according to Google.

And then there's the awesome Night Sight mode of Pixel 3. Using algorithm-driven alignment and merge techniques, including a modified version of the HDR + stack that detects and rejects misaligned images and a pattern Automatic White Balance (ABW), designed to distinguish between a well-balanced image and an unbalanced image. reduce the noise of photos taken in environments of no more than 0.3 to 3 lux (the amount of light arriving at one surface per unit area, measured in lumens per square meter). For a reference point, 3 lux equals a sidewalk lit by street lights.

Smartphone with the best alternative AI assistant

Galaxy Note 9 ($ 999)

Bixby, Samsung's voice assistant, is well-deserved for his slightly worse performance in some cases than the Google Assistant and Amazon's Alexa. Despite all his weaknesses, he has come a long way since his debut alongside the Samsung Galaxy S8 in 2017.

The enhanced Galaxy Note 9 Bixby – Bixby 2.0 – delivers better natural language processing, faster response times and built-in noise reduction technology, and retains the ability to recognize chained instructions such as "Open up". Gallery application in split screen mode. and rotate the misaligned photos "and" Play videos on a nearby television ". It's also more conversational: When you talk about upcoming Thanksgiving weekend concerts, for example, you'll remember that date range when looking for tickets. And with future support for new languages, availability in over 200 markets, and the recently launched Bixby Development Center, this support promises more support than ever before.

Bixby is not limited to voice recognition, of course. Bixby Home, a "social feed for your device," is a dashboard of social network reminders and updates in cards that can be rejected, pinned, or permanently hidden.

Bixby Vision, an object recognition app similar to Google Lens, uses the integrations of Vivino, Amazon, Adobe, Nordstrom, Sephora, Cover Girl and others (as well as Samsung's data sharing partnerships with FourSquare and Pinterest). ). It can scan barcodes, turn receipts into readable files, display the list of relevant products, recommend wine, display the number of calories of a food and allow you to virtually try make-up products .

Conclusion

So here are four flagship smartphones that innovatively use AI in three distinct categories.

Apple's iPhone 2018 lineup is by far the winner on the chipset front: its upgraded neural engine, combined with powerful software tools and a thriving developer ecosystem, furthering its hardware edge. That said, Huawei is starting to crack, especially considering the Mate 20 Pro's leading camera features that exploit the Kirin 980's enhanced artificial intelligence chip, such as AI Color and Master AI 2.0.

As impressive as the Mate 20 Pro camera, it is right in front of Google's Pixel 3 in the digital photography category. The Pixel 3 is one of the best smartphone cameras we have ever tested, thanks in large part to the AI.

Finally, there is the Samsung Galaxy Note 9, a showcase of the latest version of Samsung's Bixby wizard. Bixby may not be the block's most reliable platform, but it has become much more robust in recent months. And to our knowledge, Bixby Voice is one of the only (if not the only) voice assistants able to recognize chained commands and interact with the menus and sub-menus of the application, making it an ideal tool for hands-free operation.

The iPhone XS, the iPhone XS Max and the iPhone XR, the Galaxy Note 9 and Google Pixel 3 are available for purchase at Carrier, Amazon, Best Buy and D & # 39; stores Other major retailers in the field of physical electronics. It should be noted that the Mate 20 Pro has not been officially available in the United States – you will need to go through a third-party reseller to acquire it.