The data mark is a piece of fat meat, and it is not only China Manfu technology that focuses on this meat

“ I think data annotation has been misunderstood by the world , Data annotation is not the domain of AI ‘ foxconn ’.”

How does data annotation work ？ As a member of AI service industry , Why is it buckled “ labor-intensive ” The hat of ?

trace the origin , This has to come from 7 Years ago .

2015 Baidu AlphaGo The birth of , Like a bomb thrown into the water , Set off AI The stormy waves of the world .

The days after that , The AI industry is galloping under the spotlight , Run out of the lab 、 Realization ”AI+ industry “ to ground , Become an important member of the new infrastructure .AI The industry has also experienced the ice and fire of capital , Gradually move towards rationality .

But as a AI The most upstream of the industrial chain , The data annotation industry has been drifting away from the spotlight , Be treated with colored glasses .

Even big brother Liu Cixin said frankly ：“ Artificial intelligence now , As much intelligence as there is in front of us, there is as much manpower behind it .”

People label data with “ Assembly line 、 Small workshop 、 Poor county ” Wait for the label , It's like a clown , More are used to tease people about AI Expectations and gaps .

But anything that conforms to the law of development will progress with the times , Is data annotation the exception ?

The true face of data annotation

In the beginning , Smelling business opportunities is indeed a group of small groups , Almost AI The wave of entrepreneurship starts at the same time .

Emerging AI Most of them are in the experimental stage , Less demand for data 、 Lower requirements , This has attracted a large number of labor-intensive workshops , these “ Small workshop ” Outsourcing (BPO) Business oriented , To the outside world “ foxconn ” The impression of .

And with the AI Large scale landing , The demand for data is getting deeper and wider , The industry began to grow across barbarism , Enter intensive cultivation , Data annotation has also faced a severe test .

Facing the annotation scene, it is more subdivided 、 Data types are more complex 、 Party A with more complicated customization needs , Whole AI The basic data service industry has been forced to undergo transformation and upgrading —— More labor-intensive enterprises are facing a critical moment , More technology-based companies are emerging .

This industry has gone through the extensive period of small workshops , Start walking in “ Technology wins ” Era .

An elusive Unicorn

It's not hard to see , Data annotation is a piece of fat , China is not the only one eyeing this meat .

There have already been several overseas AI A giant in the data service industry , image Scale AI 、Appen、Lablebox etc. .

But look at home , Not only is there no Unicorn , Even large data annotation companies are rare . It makes people wonder , Why is the gap so large ?

The main problem is investors' views on data annotation , The investment value of an enterprise will be measured according to the type of business 、 Growth space 、 Technical value and other reference elements . obviously , Data labeling is not favored .

Most domestic investors equate tagging technology companies with tagging business companies , In addition, some media depict the labor-intensive attribute of the industry , So that the technical value of improving industry efficiency has not been fairly recognized , Naturally, it is difficult for enterprises to finance a large amount .

While foreign countries recognize the technical value of marking companies , Therefore, the financing amount is large , Many companies can thrive on financing alone .

according to an uncompleted statistic , The valuations of North American data annotation companies and domestic related companies are at least 10 times , The financing amount is tens of thousands of dollars or hundreds of millions of dollars , Domestic financing is generally in the range of ten million yuan .

however , With the deepening of investors' understanding of the industry , This gap will gradually decrease .

The path of technological transformation

According to the statistics of CICC Qixin ：2019 In, the market scale of China's basic data service industry has reached 30.9 One hundred million yuan , expect 2025 Years will break through 100 One hundred million yuan , Compound annual growth rate reaches 21.8%.

Facing the gradually normalized industry demand , More and more small and medium-sized suppliers are suffering from survival problems , This group is 1-2 There will be “ Collapse tide ”.

And the last remaining company , Rely on AI+ Privatization deployment + The mode of the platform . At present , It is a common practice in the industry to introduce technology into the data annotation process , Let's train well AI Model feedback manual annotation , It is also the advantage of tagging technology companies .

Take MF technology for example , Manfu technology is a leading enterprise in the industry AI Infrastructure and data intelligence platform service providers , Focus on AI Enterprises provide one-stop data solutions from strategy to technology .

As a new generation of technology oriented company , Manfu technology self-developed intelligent data service platform SEED, As an important part of data intelligence platform system , Is to realize refactoring AI The key to infrastructure .

SEED In addition to having the mainstream second-generation platform on the market “ Multi scene annotation capability + Limited project management capability ” outside , It also innovatively introduces a large number of life cycle management 、AI Enhancement and other modules , Forming a cover “ Data lifecycle management capability + Supply chain management + Project collaboration +AI Man machine cooperation + Custom permissions + Full scene annotation ” Multi dimensional stereo data processing capability .

With the blessing of these functional modules , The data labeling efficiency of the platform is improved on average 10 More than times ;AI Assisted screening , The data accuracy can reach 99.99% Level , hitting AI Pain points of enterprise data demand , Solve from the source AI The application scenario continues to expand the massive demand for multi-source heterogeneous data .

Conclusion

Recently, there are always data taggers worrying about career planning , Painstakingly imparting knowledge AI, But will eventually replace human , End up changing your own life .

indeed , Machines will become more and more intelligent , This is the normal law of development , It is also the expectation of mankind . however , In some new fields , Machines are not yet able to assist human work , Data annotation still needs to be completed manually . Even with AI Automated annotation tools , People still have to do the final step of audit and quality inspection .

Even if this day comes , Described as AI The teacher's data annotator will also be AI The last category eliminated , Because there is always work to be done .