The full screen and AI from the 2017 fire to the present, if the current appearance of the fight who's 'Liu Haixiao', then the configuration who have to fight the 'AI more applications'.
Huawei specifically created the AI chip Unicorn 970, Qualcomm Xiaolong 845 is not far behind, exclusive optimization of the AI, the major mobile phone manufacturers will also be AI applications as the highlight of their products.
Lu master AI performance list: Huawei won the championship, Samsung followed
Recently, Master Lu released the mobile phone AI performance chart for the Q1 quarter of 2018. Among them, Huawei Mate 10/Pro, Glory V10 won the championship and army with the AI processor Unicorn 970, and the Samsung S9/S9+ rushed through the strong Xiaolong 845. The first three.
Followed by the A11's iPhone x and iPhone 8 plus, this chip integrates a hardware dedicated to machine learning - 'Neural Engine'. The performance is unquestionable.
In this AI performance chart, there are vivo two flagship vivo X21 and vivo X20Plus, ranked sixth, seventh.
Lu Master AI performance rankings come from?
Master Lu's AI performance evaluation uses the currently used three neural networks Inception V3, ResNet34, and VGG16 specific algorithms to identify the picture content given by the machine, outputting the possible result list according to the probability level.
Finally, determine the efficiency of the phone to determine the AI performance, and then give the test score.
* Inception V3, ResNet34, VGG16 These three neural networks are currently recognized as the most mature test solutions:
Among them, ResNet (residual network) can train very deep networks by using residual modules and conventional SGD. Master Lu's AI test set a depth of 34 layers;
As an open source neural network model developed by Google, after using Inception, the width and depth of the entire network structure can be expanded, which can bring 2-3 times performance improvement;
VGG16 is characterized by a simple and effective model structure, a more accurate estimate, and good generalization ability for other datasets.
Why can unicorn 970's AI win?
As the first publicly released AI chip, it pioneered the addition of an independent NPU (neural network unit), CPU, GPU, and DSP to form the HiAI artificial intelligence mobile computing platform.
Huawei is also the developer and the only user of Unicorn 970. It can do 'special treatment,' without considering openness and adaptability.
In Huawei's new machine, in addition to the optimization of the chip, as well as the optimization of its own mobile phone, it is like "one to one teaching," and the application of the mobile phone + chip can be more than "one to one teaching." effect.
Dragon 845 enters the game, giving birth to more 'developers'
Prior to the release of Xiaolong 845, the development of mobile phone AI seems to be monopolized by Huawei. Unicorn 970 is only used in Huawei's mobile phones. Other mobile phone manufacturers want to talk about 'AI'. They have a little lack of confidence until after the release of Xiaolong 845. Mobile phone AI is a hundred flowers.
Compared to Snapdragon 835, Snapdragon 845 has three times the computing power of Snapdragon 835 and supports multi-platform neural network systems. Qualcomm has also opened the AI platform ecosystem to its partners.
Based on this, mobile phone manufacturers using Qualcomm Snapdragon chips are developing more market-competitive AI applications for their products, and in 2018 they are destined to be AI's 'melee' year.
What impact does the new "blockchain" have on AI?
Blockchain and AI are arguably the two most popular technology directions today. The release of a 'blockchain phone' is expected to become the next hotspot.
In the eyes of the average person, there seems to be no overlap between the two technologies. One is to foster centralized intelligence on a closed data platform, and the other is to promote decentralized applications in an open data environment.
Although both are extremely powerful, they also have their own limitations. Some of them are technically related, while others come from the ideas inherent in the field. If you can find a clever way to get the two together, the total The effect can be zoomed out instantly.
In general, currently AI development is still at a relatively early stage, and it is still simply imitating people. However, we have already seen its importance to the mobile phone and even the tech industry. Although it is not making a mobile phone a fighter 'The hardware is a soft application that makes the phone smarter. We hope that mobile phone companies can really think about how to make AI better, not to say louder.
The full screen and AI from the 2017 fire to the present, if the current appearance of the fight who's 'Liu Haixiao', then the configuration who have to fight the 'AI more applications'.
Huawei specifically created the AI chip Unicorn 970, Qualcomm Xiaolong 845 is not far behind, the AI has been optimized, the major mobile phone manufacturers will also be AI applications as the highlight of their products.
Lu master AI performance list: Huawei won the championship, Samsung followed
Recently, Master Lu released the mobile phone AI performance chart for the Q1 quarter of 2018, among which Huawei Mate 10/Pro, the glory V10 won the championship and the military with the AI processor Unicorn 970, and the Samsung S9/S9+ rushed through the strong Xiaolong 845. The first three.
Followed by the A11's iPhone x and iPhone 8 plus, this chip integrates a hardware dedicated to machine learning - 'Neural Engine'. The performance is unquestionable.
In this AI performance chart, there are vivo two flagship vivo X21 and vivo X20Plus, ranked sixth, seventh.
Lu Master AI performance rankings come from?
Master Lu's AI performance evaluation uses the currently used three neural networks Inception V3, ResNet34, and VGG16 specific algorithms to identify the picture content given by the machine, outputting the possible result list according to the probability level.
Finally, determine the efficiency of the phone to determine the AI performance, and then give the test score.
* Inception V3, ResNet34, VGG16 These three neural networks are currently recognized as the most mature test solutions:
Among them, ResNet (residual network) can train very deep networks by using residual modules and conventional SGD. Master Lu's AI test set a depth of 34 layers;
As an open source neural network model developed by Google, after using Inception, the width and depth of the entire network structure can be expanded, which can bring 2-3 times performance improvement;
VGG16 is characterized by a simple and effective model structure, a more accurate estimate, and good generalization ability for other datasets.
Why can unicorn 970's AI win?
As the first publicly released AI chip, it pioneered the addition of an independent NPU (neural network unit), CPU, GPU, and DSP to form the HiAI artificial intelligence mobile computing platform.
Huawei is also the developer and the only user of Unicorn 970. It can do 'special treatment,' without considering openness and adaptability.
In Huawei's new machine, in addition to the optimization of the chip, as well as the optimization of their own mobile phones, it is like a one-on-one teaching, and the application of the mobile phone + chip can be more than the "one-to-many teaching." effect.
Dragon 845 enters the game, giving birth to more 'developers'
Prior to the release of Xiaolong 845, the development of mobile phone AI seems to be monopolized by Huawei. Unicorn 970 is only used in Huawei's mobile phones. Other mobile phone manufacturers want to talk about 'AI'. They are underpowered until after the release of Xiaolong 845. Mobile phone AI is a hundred flowers.
Compared to Snapdragon 835, Snapdragon 845 has three times the computational power of Snapdragon 835 on the AI and supports a multi-platform neural network system. Qualcomm has also opened up the AI platform ecosystem to its partners.
Based on this, mobile phone manufacturers using Qualcomm Snapdragon chips are developing more market-competitive AI applications for their products, and in 2018 they are destined to be AI's 'melee' year.
What impact does the new "blockchain" have on AI?
Blockchain and AI are arguably the two most popular technology directions today. The release of a 'blockchain phone' is expected to become the next hotspot.
In the eyes of the average person, there seems to be no overlap between the two technologies. One is to foster centralized intelligence on a closed data platform, and the other is to promote decentralized applications in an open data environment.
Although both are extremely powerful, they also have their own limitations. Some of them are technically related, while others come from the ideas inherent in the field. If you can find a clever way to get the two together, the total The effect can be zoomed out instantly.
In general, currently AI development is still at a relatively early stage, and it is still simply imitating people. However, we have already seen its importance to the mobile phone and even the tech industry. Although it is not making a mobile phone a 'fighter' 'The hardware is a soft application that makes mobile phones smarter. We hope that mobile phone companies can really think about how to make AI better, not to say louder.
