Abstract: What are the new REQUIREMENTS of IT brought by AI revolution? Breakthroughs in deep learning and rapid advances in hardware make the “NTH spring” of artificial intelligence full of vigor and vitality. For the first time in history, machines can do ‘human’ jobs like face recognition better than we can. Artificial neural network has many ‘hidden’ or computing layers. To realize deep learning, a series of specific configurations should be carried out on the specific artificial neural network architecture. Data can be provided for system self-training or inference, and numerical results can be read from the output neuron layer.

What are the new IT requirements brought about by the AI revolution?

Breakthroughs in deep learning and rapid advances in hardware make the “NTH spring” of artificial intelligence full of vigor and vitality. For the first time in history, machines can do ‘human’ jobs like face recognition better than we can.

Artificial neural network has many ‘hidden’ or computing layers. To realize deep learning, a series of specific configurations should be carried out on the specific artificial neural network architecture. Data can be provided for system self-training or inference, and numerical results can be read from the output neuron layer.

In addition to increasing complexity at the software level, AI’s computing model also brings new hardware requirements. For example: 1 Adding more consistent SIMD (Single Instruction Multiple Data) calculation model can make the processor, vector processor, accelerator, FPGA and custom chip run efficiently. 2. Is it necessary to introduce special chips such as ASICS FPGA? How to use CPU GPU together? 3 The training set must be large enough to make full use of all the parallel computing capabilities of the equipment; otherwise, the performance will be wasted. 4 During training, the hardware’s ability to handle all the parallel computing depends more on the performance of the cache and memory subsystems. So, how much do all kinds of memory need to prepare? …

“Carrot and stick”, can consider AI all-in-one machine?

The AI all-in-one machine outputs the Ali Cloud technology in the form of localized AI deployment, including the video, voice, and NLP-Natural Language Processing product families, and accelerates user service efficiency with THE help of AI technology.

Its technical advantages lie in:

  • High precision: the core engine adopts the latest deep learning and 100 million level data actual combat training
  • Full sample: 1300+ sensitive people sample database has been accumulated and continues to operate
  • Strong adaptability: the use of face tracking technology, video resolution, posture, quality robustness
  • High efficiency: A minimum-configured cluster can process 40 channels of videos at the same time, and one-hour videos can be processed within 10 minutes
  • High accuracy: recall rate is more than 99%

The AI all-in-one solution has entered the commercial stage, and it is reported that it has been applied to the media industry, such as image and video content audit scenes.

The original link