“Any AI smart enough to pass a Turing test is smart enough to know to fail it.”—Ian McDonald, River of Gods
WHILE smartphone companies continue to highlight features, such as bezel-less screens, DSLR-like camera capabilities, higher screen resolutions and longer battery life, all of which are much easier to understand for the regular consumer, the more interesting developments are happening deep within the phone’s internals. Not just in terms of processing power or graphical capabilities, but artificial intelligence or AI.
According to research from Counterpoint’s Components Tracker Service, 1 in 3 smartphones, or roughly more than half a billion, shipped in 2020 will natively embed machine learning and AI capabilities at the chipset level.
Apple’s Bionic system on chip (SOC) will drive native AI adoption in smartphones, making the iPhone maker a leader in the AI-capable chip market through 2020. Huawei comes in second in the market with AI-capable smartphones when it launched its HiSilicon Kirin 970 SOC.
The initial driver for the rapid adoption of AI in smartphones is the use of facial-recognition technology by Apple in the iPhone X.
With advanced SOC-level AI capabilities, smartphones will be able to perform a variety of tasks, such as processing natural languages, including real-time translation; helping users take better photos by intelligently identifying objects and adjusting camera settings accordingly—things which, by the way, the Huawei Mate 10 smartphone series can already do.
Experts say this is just the start as machine learning will eventually make smartphones understand user behavior in an unprecedented manner. By analyzing user behavior patterns, devices will be able to make decisions and perform tasks that will reduce physical interaction time between the user and the device.
HUAWEI’S A.I. STORY
WITH the release of their most recent flagship products, Huawei hailed the advent of a never-before-seen AI development strategy: from chip planning to phone manufacturing carrying the AI platform. The Huawei Mate 10 with Kirin 970 chip was hailed as the “first mobile phone AI chip.”
Device Side AI, or On-Device AI, is one of the most talked-about topics in AI in recent years; it refers to a concept that the collection and calculation of data and decision-making are all performed on the end device. Compared with cloud AI, On-Device AI is featured by enhanced stability, less time delay and a commitment to protect user privacy at the same time.
However, its computing power is still not in the same league as cloud computing, but with the popularization of AI, On-Device AI is garnering greater acclaim in the market. On-Device AI requires corresponding chip/board on the hardware end device.
“The project of Kirin 970 was established two years ago,” said Eric Zhou, marketing director for Huawei’s Wireless End Device Chip Business Department, who has been in the chip industry for 12 years. He said the most important factor in defining a chip was prognosis.
Huawei’s team detected a major bottleneck in the AI industry, computing power, and with such inadequacy, both the training and application of in depth learning would be greatly restricted.
Compared to the previous Kirin 960, Kirin 970 is equipped with newly added NPU specializing in processing AI computing, and, moreover, it is featured by optimized functions, higher energy-efficiency ratio, better communication and other optimized solutions, such as Android version upgrades, with its energy-efficiency ratio especially remarkable.
It claims to be an AI chip, because it provides a new structure for mobile computing of AI, called HIAI, which boasts a unique integrated NPU (Neural-Network Processing Unit).
“As we look to the future of smartphones, we’re at the threshold of an exciting new era, Mobile AI = On-Device AI + Cloud AI,” Richard Yu, CEO of Huawei Consumer Business Group, said in one of his interviews. “Huawei is committed to developing smart devices into intelligent devices by building end-to-end capabilities that support coordinated development of chips, devices and the cloud. The ultimate goal is to provide a significantly better user experience.”
Kirin 970 is powered by an 8-core CPU and a new generation 12-core GPU. Built using a 10 nm advanced process, the chipset packs 5.5 billion transistors into an area of only 1 cm. It is the first mobile AI computing platform featuring a dedicated NPU. Compared to a quad-core Cortex-A73 CPU cluster, the Kirin 970’s new heterogeneous computing architecture delivers up to 25 times the performance with 50 times greater efficiency. Simply put, the Kirin 970 can perform the same AI computing tasks faster and with far less power. In a benchmark image recognition test, the Kirin 970 processed 2000 images per minute, which was faster than other chips on the market.
At an event held in London in January of this year, Huawei gathered together experts on AI and human behavior to put our minds at rest about how the technology will help free us from mundane decisions, and actually encourage creativity further.
According to research done by Huawei, we make about 35,000 decisions a day and AI can help free our minds from some of the mundane decisions we have to make and instead help us concentrate on the more important things we have to deal with. By using its sensors and data, the NPU can make the jump from collecting and analyzing data, to eventually intelligently using that data to help us make decisions.
“AI needs to make our lives easier,” Huawei Chief Marketing Officer Andrew Garrihy said. “We can forget things because AI will remember for us, and then go on to make an even better decision when it does. It could unlock human potential.”
The third-biggest smartphone manufacturer believes AI will become mainstream in terms of mass awareness this year.
Some AI functions that can already be achieved by the Mate 10 include:
- Recognition of scenes: The Mate 10 and Mate 10 Pro have honed their skills with 100 million pictures, and can recognize 13 scenes and objects intelligently. After recognizing the scenes/objects inside the lens (such as a still flower or a running human), the system will adjust camera setting or conduct deblurring automatically according to the property of the objects, and thus producing a picture that better fits the scene.
- A.I. beautification mode: With built-in facial-recognition technology, Mate 10 phones are able to detect human face and beautify the image automatically, alternating dynamic effects every three seconds. Even when another human face moves in front of the camera, AI beautification can still function precisely.
- A.I. noise reduction: With its in-depth learning algorithm, Kirin 970 is capable of reducing noises and enhance vocal signals effectively. Success rate of its voice recognition amid vehicle noise has risen from 80 percent to 92 percent.
- Voice translation: The Mate 10 carries AI-driven voice translation system that is customized by Microsoft. With the support of NPU, it can achieve real-time, back-and-forth text translation and image recognition translation at a speed 300 percent faster than traditional translation, and it is capable of translating 50 languages.
The Mate 10 also comes with a special AI application, whose function is to switch off unnecessary apps running in the background while guaranteeing user experience by analyzing local data to understand user’s habits.
For the time being, the development of the AI industry is in full swing, with scenes of prosperity across the whole. On the algorithm level, numerous researchers are conducting studies in different fields of AI, and their endeavor has not only bred black technology like AlphaGo but also has constantly optimized the application of algorithm. On the hardware level, GPU and various types of customized AI chips keep coming, promoting the growth of computing power of semiconductor hardware.
With algorithm and chips working together at the same time, AI computing power makes several dozen or even several hundred leaps every year. In the past, due to constraints in performance and technology, development of AI applications was confined to the cloud.
But nowadays, with the joint advancement of algorithm and chips, On-Device AI will unlock much more room for the imagination of application developers, and phones can support more real time, more powerful and more intelligent AI applications.
With Huawei releasing its new flagship, the P20, in a few days, can we expect even more advanced AI capabilities? Everything has yet to be official but from all the information we’ve gathered in the past couple of weeks, we know that the P20 series will offer several choices, namely, the P20 Lite, P20 (regular) and P20 Plus (or will it now be called Pro?), though we’re still unsure if Huawei will be bringing all three variants to the Philippine market.
As far as the “leaked” photos, it shows that the P20 will feature a bezel-free screen and a “notch” similar to the one on the iPhoneX.
The P20 Lite is said to be a 5.6-inch display with a resolution of 2250×1080 pixels with an LCD panel, while the P20 has slightly bigger 5.8-inch display. The P20 Pro is rumored to have a 6.1-inch OLED screen.
The three look visibly similar except that the smaller P20 Lite has larger bezels and has the fingerprint scanner on the rear of the handset, whereas the P20 and P20 Pro have the fingerprint scanner on the front.
All have a raised dual camera unit, but the P20 Pro breaks out a third lens on the rear. The Huawei P series has always been made for photography and it has stood out because of its partnership with Leica.
The P20 and P20 Pro looks to further that partnership and if the leaked specs are to be believed, the Pro version might just have a whopping 40-MP, 20-MP and 8-MP sensor for its triple rear camera.
I expect that Huawei and Leica will continue the RGB and monochrome setup but its intriguing as to how they will utilize the third camera. Wide angle? 4x zoom?
The small notch looks to house the front-facing (24-megapixel?) selfie camera and a sensor, which will most likely be used for AI and facial-recognition purposes. As to how the P20 will use AI, I think we can expect more improvements on the language-translation feature. I also think it would be able to recognize more scenes/objects, or maybe the P20 will now be able to further distinguish objects or identify specific food items and show you just how many calories it contains.
Are you excited about the P20? What new features are you hoping to see? We’ll have more information in the coming weeks.
VIVO UNVEILS NEW FLAGSHIP, THE VIVO V9
AFTER establishing a strong presence in the local smartphone market last year with the Vivo V7+ and Vivo V7, the brand seeks to further its claim as the smartphone of choice of the selfie-centric generation with its new flagship phone for 2018: the Vivo V9.
Launched last Thursday at the Nobu Hotel City of Dreams, the Vivo V9 comes with a Full View Display, Dual Rear cameras, and features the now famous “notch.”
Let’s get the obvious out of the way first. Yes, the Vivo V9 does look a lot like the iPhone X, (especially the black variant) because of the notch on the front screen and vertical dual-camera placement at the back, but instead of the Apple logo, the Vivo V9 has a very responsive fingerprint sensor in its place. Turn the phone on and you’ll be greeted by familiar looking icons and wallpapers that look crisp and clear on the V9’s 6.3-inch IPS screen, which has a resolution of 2280×1080. If you love watching movies or playing games on your mobile device, you’ll enjoy the V9’s Full View Display.
The V9 uses a 2.2 GHz Snapdragon 625 Octacore processor and comes with 4GB RAM and 64GB internal storage—expandable via the triple slot tray (2 SIM card slots and one for microSD) and a 3260mAH battery.
On the back of the phone are the dual rear cameras, a 16 MP and 5 MP with f/2.0 aperture.
The secondary camera allows you to get those bokeh shots (blurred background) which can make your photos look a bit more artistic and your portrait shots stand out.
Since it’s really meant for those who love to take selfies, the main camera of the V9 is its front camera.
The 24-MP camera takes awesome selfies, has an improved beauty mode and supports AR, or those augmented-reality stickers that a lot of young people seem to love because it can make them look like kittens, dogs, etc., and add other cute stuff to their selfies. The front cam can even shoot 4K videos.
Besides the above-mentioned fingerprint scanner, you can also set up facial recognition to unlock your phone. Set up is easy, although using it to unlock your phone is a bit slower and less secure. I still have to test it thoroughly as to how accurate it can be.
I’ll have a full review and will announce the official pricing of the V9 via my social-media accounts, so be sure to follow @whereiseduy on Instagram and Twitter.