Analysis of four elements of artificial intelligence

big data:

The intelligence of artificial intelligence is contained in big data.

Calculation power:

Computing power provides basic computing power support for artificial intelligence.

Algorithm:

Algorithm is the fundamental way to realize artificial intelligence and an effective method to mine data intelligence.

Scene:

Big data, computing power, algorithm as input, only output in the actual scene, can reflect the actual value.

Take a very vivid analogy: if we take cooking as our scene, big data is equivalent to the ingredients needed for cooking, computing power is equivalent to the gas / electricity / firewood needed for cooking, and algorithm is equivalent to the cooking methods and seasonings.

1) Big data

In this era, big data is produced all the time. Mobile devices, cheap cameras, ubiquitous sensors and so on. These data forms are diverse, most of them are unstructured data. If it needs to be used by artificial intelligence algorithm, it needs a lot of preprocessing.

2) Calculation power

The development of artificial intelligence puts forward higher requirements for computing power. The following is a comparison of the computing power of various chips. Among them, GPU is the most widely used chip in the field of artificial intelligence. Both GPU and CPU are good at floating-point computing. Generally speaking, the ability of GPU to do floating-point computing is about 10 times that of CPU.

In addition, the deep learning acceleration framework improves the computing performance of GPU by optimizing on top of GPU, which is conducive to accelerating the calculation of neural network. For example, cudnn has customizable data layout, supports flexible dimension sorting, step and sub area of four-dimensional tensor, and is used as input and output of all routines. In the convolution operation of convolution neural network, the matrix operation is realized, and the memory is reduced, which greatly improves the performance of neural network.

3) Algorithm

The mainstream algorithms are mainly divided into traditional machine learning algorithm and neural network algorithm. Neural network algorithm is developing rapidly, in recent years, because of the development of deep learning to the climax.

4) Scene

The classic application scenarios of artificial intelligence include:

1. User profile analysis

2. Risk control based on credit score

3. Fraud detection

4. Intelligent investment advisor

5. Intelligent audit

6. Intelligent customer service robot

7. Machine translation

8. Face recognition

Editor in charge: CC

Leave a Reply

Your email address will not be published. Required fields are marked *