Now, in June 2016, machine learning, data mining, computer vision, and other applications are bubble-free and far from even realizing the potential that algorithms should have. From my experience at from my experience at my experience at my experience at my experience at my experience at my experience at my experience at my job companies have something to algorithmize in almost every product line. At the same time, many new algorithms can bring new product features, improve user experience, improve the company's core indicators, and expand new revenue channels. The next 3 to 5 years, will be the fastest stage from automation to algorithmization, many companies will benefit from this, a variety of core indicators will be overturned, the fast company will eat the slow company. Benefit from two major reasons: First, the company has accumulated a large amount of data for the algorithm to lay a data foundation. And at the same time, due to the development of modern technology tools and ecosystems, for a product with 10 million monthly live, a team of 5 to 10 full-stack data scientists can support all the tasks of a company's several core algorithms from end to end, including from R&D to supporting the final product. However, for the current trend of entrepreneurship, the various algorithms taken out as a third-party service to start a business, I personally believe that it is more difficult to succeed. The core reason is that various algorithms need to be integrated with the company's core product line, and the algorithm is part of the company's core product. Can use third-party services are often not the core product algorithms. Hardware can not, binary computing speed is too slow, it seems that only the United States is now engaged in other binary, similar to the quark computer or something, physics and mathematics PhDs probably have more contact. Big data this thing is still very early, completely undeveloped, big data is not to say that the collection of data statistics statistics is big data, this is to collect data trajectory, to put it bluntly, the final data should be presented three-dimensional. A data should not determine the value, but ultimately in the form of a function of the data trajectory is called big data, and this point is now the most cutting-edge data has not been completely done.