Shulou(Shulou.com)06/02 Report--

Computing power, models and data constitute the three elements of artificial intelligence. in the past, we focused too much on computing power and models, but did not realize that with the deepening of artificial intelligence, good algorithms and models are no longer rare species. Suzhou annual meeting planner, on the contrary, those tagged high-quality data have become the most scarce "black gold".

"the company's barrier is no longer the algorithm, but the data. Let the algorithm use enough data to make the product work." Wu Enda, an international authority in the field of artificial intelligence and machine learning, emphasized the importance of data when he delivered a speech on the theme of "AI is the new electricity". Coincidentally, Kaifu Lee also expressed this view in a speech entitled "the Golden Age of artificial Intelligence" in the computer Science Experimental Class of Tsinghua University. "if you have a monopoly big data, you will have a great advantage."

All these indicate one thing, that is, the rise of AI is inseparable from "good" data as the foundation, which is also the original purpose of the establishment of cloud data.

Traceability of AI data Service for Cloud Survey data

Since entering the enterprise service market in 2011, Testin Cloud testing has been committed to promoting the intelligence of the industry. In addition to the testing business, we have become a monopoly brand in the professional field, and cloud testing data focusing on AI data services has also become a benchmark brand in the data field. At present, the size of our entire data service team has exceeded 1000. Through the process operation mode and data security mechanism of bidding and review separation, we can better ensure the high-quality output and privacy of data, so as to better provide customized 'data nourishment' for artificial intelligence landing. " In an exclusive interview with titanium media, Jia Yuhang, general manager of Cloud Test data, said.

As a non-standard field, AI data service often needs to be customized according to different industries and different needs, and the process of data annotation, standardization, standardization and machine readability are indispensable, which means that there are no shortcuts in the field of cloud test data.

The threshold of early data labeling service is not high, a few people, a few computers can start the operation, resulting in the industry mixed, homogeneous competition and other phenomena, and at this time artificial intelligence is also in the early stage of development. However, when artificial intelligence drives into the deep water area, the momentum of "application of human intelligence" is becoming more and more popular, and the corresponding algorithms have higher requirements for the accuracy and quality of data, which requires that as a provider of AI data services, artificial intelligence should provide customized and high-quality data for restoring application scenarios.

In response to this, Jia Yuhang told Titanium Media, "take face key point recognition as an example, the previous relevant data annotation can often describe its task requirements in one sentence, but now it has developed to hundreds of key points." usually an order of magnitude of face data tagging tasks, sometimes four sheets of A4 paper may not be able to complete these requirements, and face data tagging is only one of the task requirements in many fields. "

Under the magnitude of the huge data labeling task, there is a general consensus on accurate and high-quality data on the demand side of the industry.

This requires data services to make great efforts in data tagging and collection, while the capabilities of small teams are stretched. When you return to the multi-domain nature of data tagging, you will find that relying on large numbers of people or adopting the "crowdsourcing" model can often only solve the demand for quantification. whether data tagging personnel can unify collaborative management and whether they have relevant domain knowledge is to determine the quality of a data task.

At the same time, this is what Cloud Test data is focusing on right now. Just as doctors can label ct films well, while the cloud test data team found that those who can label data quickly and accurately tend to have driving experience when marking data on the outside of self-driving cars.

What is the secret of the rapid growth of cloud test data?

At this point, we still need to think about why cloud testing data can do and do a good job of AI data service.

By observing the development history of Testin cloud testing, we can find the answer.

Since the establishment of Testin Cloud Test in 2011, it has provided services to more than one million enterprises and developers around the world, and has accumulated rich and perfect technical capabilities and process management capabilities. The AI data service for Cloud Test data was officially launched in 2017. In other words, the data business line of Testin Cloud Test has seven years of experience accumulated by enterprise services from birth, and inherits the role of an independent third party in the industry. The natural "customer-centered" enterprise service gene is the biggest moat that distinguishes cloud test data from its peers, while the most important demand of customers is to "reduce costs and increase efficiency".

"what is different from the more emphasis on standards of corporate services in the US environment is that China places more emphasis on services. Through so many years of observation, we have found that whether we can effectively meet the real needs of users is actually a very important point. It does not mean that enterprises have to make a platform or a tool, but more from the needs of enterprises or industries, to build a corresponding service model." Jia Yuhang added to titanium media.

Take the inspection of new retail stores as an example. Generally speaking, every store has to be inspected once a month. The mode of store inspection is to allow a person to take inventory with a research form. With the increase in labor costs, the number of stores is increasing, which has made it a big expense. Through the introduction of AI data service, staff can now take a mobile phone APP direct inspection, the number of items, the number of sku and the corresponding number of sq, can be known at a glance.

"according to feedback from different customers, enterprises that have landed AI products through our cloud test data tagging services can reduce labor costs by about 1 to 3." That's what Jia Yuhang said.

Store inspection is just one of the cases. at present, cloud test data mainly focus on smart driving, smart city, smart finance and smart home, which are also the areas with the greatest market demand. In the face of different data fields, cloud test data builds each link into different modules through pipelining operation, and optimizes the processes of personnel management, data collection, data cleaning and data labeling with its own process management tools to ensure continuous and efficient internal operation, and finally ensure the high-quality output of AI data.

According to the IDC survey, the development of big data in China is currently in the application landing stage, and the whole market is expected to maintain a sustained growth trend in the next five years, with an annual compound growth rate of 17.3%. Thanks to the development of artificial intelligence, 5G, block chain and edge computing, the future multi-party technology integration, data growth will inevitably show a blowout trend, data collection and standard business as its concomitant, there must be more room for growth.

Thanks to the judgment of the trend of AI, Testin Cloud Test believes that "artificial intelligence is gradually developing towards the application of artificial intelligence", so the cloud test data determined the customized business policy of "precision, high quality and independent security" at the beginning of its establishment. In line with this "trump card", the cloud test data department has expanded rapidly, and under the perfect grafting of previous enterprise service experience, cloud test data has finally become the leading enterprise in the field of AI data services. "

"the business gauge modulus of cloud testing data is growing in multiples every year, which is also closely related to the depth of the market we are in. In my opinion, the whole market is still showing a non-linear geometric growth trend, and there are still many opportunities to be exploited." When talking about the development status of cloud testing data business line, Jia Yuhang said.

"Security" is an unavoidable proposition for AI data service providers.

Under the opportunity, the enterprise should not only provide high-quality data, but also pay attention to the standardization and security in the process of data service.

In this regard, cloud testing data standardize the management of full-time data service teams through self-built data acquisition laboratories and self-built data tagging bases. This measure not only ensures the quality and efficiency of labeled data, but also ensures the security and privacy of data output to the maximum extent.

Jia Yuhang stressed to titanium media that cloud test data has put data security in the first place from the very beginning, focusing on the following aspects:

First, do not abuse data, clean up the data after delivery, do not leave a bottom, and never use it again; second, do not infringe upon privacy and sign data authorization agreements with all users of data collection to ensure that the data used by AI enterprises for training is legal and compliant; third, establish relevant data protection mechanisms, such as the setting of firewalls, internal information system management and protection, and even standardized process operation system.

As Testin CMO Zhang Pengfei repeatedly stressed: "even if the cloud test data from security to privacy protection system will increase operating costs, but from the overall development of our industry, only with this responsible attitude to carry out the work, our industry can 'good money drive away bad money'."

Source: consumer Daily Network

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.