Skip to content
TUPUTECH
  • HOME
  • NEWS
  • CONTACT US
  • LOGIN
Site Search

NEWS

news

Already huge in China, Tuputech pitches its AI that…

  • 2017-10-282018-06-20

Social media platforms have a content problem.

This isn’t about the fact that last year they became (hopefully) unwitting shills for foreign governments trying to sway the outcome of our national elections.

That’s definitely a major (and potentially existential) threat to social media businesses, but so is a more lingering, and equally as pervasive problem websites like Facebook, Instagram and Twitter have with violent and pornographic imagery popping up in their feeds.

Three years ago in Wired, Adrian Chen wrote a great piece about the poor souls who work for these billion-dollar companies out of linoleum-tiled offices in the Philippines — far from the glittering campuses of Silicon Valley or the high-tech office towers in San Francisco — whose job it is to keep dick pics and beheadings from reaching the screen of your preferred internet consumption device.

Well, it’s not just an American problem.

In China, where content laws are far more restrictive and penalties for defying government proscriptions on content are far more severe, new social media giants and telecom providers like Musical.ly, Live.me, China Mobile Communications, Thunder and China Unicom have moved away from employing human censors to deploying a technology from the young Guangzhou-based startup Tuputech.

Launched just three years ago by one of the co-founders of China’s wildly successful WeChat messenger service, Tuputech has already amassed a group of customers in China whose combined reach is larger than the entire population of the U.S.

“Interacting with a brand or social site online should be enjoyable. We shouldn’t have to worry about offensive content popping up,“ said Leonard Mingqiang, founder of Tuputech (and a former founding member of WeChat). “Our technology acts fast to detect inappropriate content and help our clients remove it before it enters the online environment.”

It’s not surprising that any social media company would look to automate as much of the process as possible — indeed, Facebook and others already use their own content management tools in addition to the human censors that regulate news feeds — but in China, where the government is far more restrictive of free speech, tools like Tuputech’s are a necessity.

The company boasts some staggering numbers. Its software analyzes 900 million images daily and processes 50 images per second, identifying pornography or violent images with a 99.5 percent accuracy rate.

It’s that 0.5 percent that can be problematic for Western companies. In China, there’s little more than lip service paid to the right to free speech. In reality, there are no protections for free speech and therefore the companies that censor their communities don’t face any reprisal or censure from the users they’re censoring.

However, companies in the U.S. have to deal with customers that are more concerned with the limitations that the services they use put on the types of posts they make.

Facebook users like DiDi Delgado, a poet and activist, have explicitly taken the world’s largest social network to task for its censorship of certain communities. And the company faced tremendous criticism when ProPublica revealed how the company’s rules protect white men.

The purported benefit of a service like Tuputech would be its ability to “learn” with each image it processes. So far, the company has categorized more than 100 billion images for its Chinese clients, and it would now like to hoover up data on restricted content from any U.S.-based company that would care to use it.

To use Tuputech, customers link to the company’s software using an API. Images are processed on Tuputech’s servers and categorized. If an image is clearly objectionable then Tuputech will flag it.

Companies using the service will send discrete images along as they’re uploaded to the site… or take random screen grabs at five second intervals if it’s a video.

Tuputech has a second manual phase that it will use to send a picture to a client for review if the content is what Mingqiang described as borderline.

“We have a second phase of manual addition to decide whether the nudity is prohibited or acceptable,” Mingqiang said through a translator. “If it is only a little bit naked. That [acceptability] would be decided in the manual additional phase.”

There are some impressive researchers who are working with Mingqiang on the technology, including Xiao Fei He, the former head of the research institute established by the ridesharing service Didi.

In China, some companies are relying almost exclusively on Tuputech, and have reduced their auditing manpower by as much as 90 percent. In a release, the company cited its use at Musical.ly as an example. There, the company has eliminated 95 percent of its human labor efforts to manage uploaded images and videos, and today only 1 percent of their videos now require manual review.

“Our goal is to remove some of the stress and pressure that comes with the day-to-day operations of running a massive online business,” said Mingqiang in a statement. “We take online image and content review off your plate. A company’s manual review becomes 1% of their time so they can focus on more important things and keep moving their company forward.”

Tuputech has raised $10 million in financing from Chinese investors like Northern Light Venture Capital and Morningside Venture Capital. Its last round closed in April 2016.

news

Tuputech Interpretation: How AI Technology Helps to Avoid Artificial…

  • 2017-09-252018-06-20

 

Recently, the event of “live video of murdering” on Facebook has caused an uproar, and some artificial errors still exist though Facebook has made great efforts in content filtering. In this event, the murder “cleverly” uploaded a live video without big problem firstly as disguise that successfully escaped from Facebook staffs’ examination, so that the illegal video was found by administrative staffs in two hours after being uploaded. Before the “murdering event”, Facebook had also been criticized by the public because the administrative staffs didn’t discover the pornographic images of children uploaded by a user from the photo album. Facebook has made a self-criticism for these two events, and Zuckerberg also expressed sorry for this “murdering event” in F8 Conference. But there is no doubt that such events have still deeply damaged the corporate image of Facebook.

It happens that there is a similar case. Before the event of “live video of murdering” on Facebook, the communication software What’s APP was banned by the Brazilian government because it’s suspected of spreading pornographic images of children. Some strippers and porn stars sold short pornographic videos by Snapchat after its payment function Snapcash was introduced. It could be seen from this that, major social media have made efforts in content filtering, but they are still unable to deal with the illegal content in time due to various reasons.

The reason for neglecting to take precautions

Why does audit delay repeatedly on the social media platforms?

It is necessary to mention that the way of “artificial review” has been adopted for content review on most social media platforms. The first necessary question of using artificial review is “artificial error”. Based on simply understanding, artificial error is insurmountable and inborn “inborn and” of human. Physiological fatigue and missed judgement are inevitable only by artificial review in face of thousands of picture and videos each day. In addition, most malicious users may choose to upload illegal images and videos late at night in order to avoid the risk of content being deleted, which has also become an important reason for administrative staffs being unable to deal with illegal content timely. What’s more, some users even mix legal content with illegal content so as to achieve the purpose of “trying to deceive everybody”.

Artificial intelligence helping to avoid risk

With the advent of the era of artificial intelligence, machine has been able to do most content filtering work. An artificial intelligence company named Tuputech has made remarkable achievements in the field of content filtering. The company successfully cooperated with social media or live streaming customers such as musical.ly, Blued and Bilibili just in three years. Tuputech has realized recognition service of hundreds of billions of images with accuracy rate up to 99.5% for they have many customers on well-known live streaming and social platforms. Images could be reviewed uninterruptedly in 24 hours a day with image recognition technology by machine, and illegal images could be recognized timely and accurately. They largely save customers’ human input on content filtering, and also respect customers’ subjective initiative better. It is important that they also support flexible combinations of interfaces so as to meet customization requirements of different customers.

 

news

Interview with Mingqiang Lee, Founder and CEO, Tuputech

  • 2017-09-252018-06-20

Article by Avinash duduskar of MTS

“For all we know, today we have stepped into an era of picture-reading. Most of the information is conveyed through pictures and videos.”

MTS: Tell us about your role and how you got here. What inspired you to found an AI image recognition company?
For all we know, today we have stepped into an era of picture-reading. Most of the information is conveyed through pictures and videos. So the demand for image recognition, image analysis, and image understanding are growing much stronger than ever before. Some companies generally don’t have any AI engineering and technical capabilities, but they do have a strong demand for dealing with image data. Therefore, the prospect of image recognition is more than promising.

MTS: How does Tuputech’s Image recognition API work?
Firstly, the image data from the customer’s service is transferred to our API through cloud. We take this data and use our image recognition system to recognize the image. The output result will be returned to the customer through the cloud. Under normal circumstances, the whole procedure only takes 1 second and the system can deal with 50 images at a time.

MTS: How does Tuputech’s offering integrate with various DMPs/CMS?
The output results of image recognition involves scoring the image, whether the image is acceptable and whether it needs to be reviewed. Most of our customers will group the images based on the results we provide, and then do relative processing on their auditing platforms. For example, the pornographic content auditing platforms will delete the pornographic images and manually recheck the suspect images. Usually, images suspected of being pornographic account for less than 5%, which means that customers only have to recheck 5% of all the data. For live-broadcasting platforms, they will send the illegal-suspected items to a specific auditing position and the auditors can deal with those images. In this way, the efficiency of auditing can be improved.

MTS: Which sectors do Tuputech’s image recognition services cater to?
Tuputech excels in recognizing offensive images and videos, including pornography, violence, terrorism and spam. We can train our models based on the different requirements of our customers. In other words, we can solve the identification problems for people with different skin colors. Additionally, our image recognition system can not only identify the normal pornography but also child pornography.

MTS: How would you commercialize AI in a step-by step manner at Tuputech?
AI commercialization is a step-by-step process, instead of a one-stop process. Due to the specialty of AI technology, it can achieve a 70%-good model very quickly, but it is difficult to  improve the model to its 100% perfection. TupuTech provides costumers with quality-assured services while preserving the basic image-recognition. We are trying to enhance our accuracy, perfecting the whole recognition loop.

MTS: Would you tell us about a standout digital campaign? (Who was your target audience and how did you measure success)
Our targeted customer types include social media and websites for live-broadcasting, short videos and pictures sharing, etc. Except for adult sites, most of the websites do not allow pornographic content. What’s more, video-socializing websites are extremely popular today and because of their privacy, problems in management and supervision could happen from time to time. Tuputech’s service can be accessed via API as well as SDK. So far, SDK for pornography recognition is available and can be integrated in apps to moderate videros in real-time. The SDK can be embedded in the app and could supervise video conversations in real time. As for social video apps, we can also identify whether the user is juvenile or not, and prevent them from using this type of app.

MTS: As a business leader of an AI-centric firm, how do you see current AI systems evolving over the next five years?
AI will surely overturn different industries and re-deploy social resources. The influence of AI is unintended and at the same time all-round. From our country’s status, transformation of cities and the development of companies, to personal career, none of these is unaffected by AI. The whole world will change under the impact of AI and we all will face a new round of opportunities and challenges.

MTS: Thank you Mingqiang! That was fun and hope to see you back on MarTech Series soon.

news

Tuputech’s Answer to AI Commercialization: A Spiral Progress to…

  • 2017-09-252018-06-20

Guangzhou, China – July 13, 2017 —

Shenzhen Artificial Intelligence Industry Association (SAIIA), with over 100 member companies specializing in AI technology, was inaugurated in Shenzhen on June 30th. Li Mingqiang, founder and CEO of Tuputech, vice president company of SAIIA, was invited to attend the ceremony and the AI Leadership Conference. He also shared his opinions in the roundtable meeting on the theme of Foresee the future, Embrace the AI Innovation.

In the meeting, Li Mingqiang put forward three important points: social wealth and resource will be re-allocated as artificial intelligence (AI) will soon bring about a revolution in all industries; AI technology shall be commercialized in an upward spiral way; dedication in certain areas of expertise and cooperation with other companies contribute to the steady and healthy growth of AI industry.

Li Mingqiang believes that, as the cornerstone of the fourth industrial revolution, AI technology is quickly infiltrating into and exercising strong influence on various industries, therefore presents great opportunity for China’s rise and brings about the reallocation of resources. Following the opportunity comes the challenge. AI enterprises need to think about how to commercialize the innovative technology and survive instead of being bubble in the AI wave.

“AI technology shall be commercialized step by step, but not at one go,” said Li Mingqiang. Citing Tesla as an example, Li Mingqiang pointed out that basing on pre-accumulation of data and related technology in the early stage, Tesla has now integrated amassed resource and gradually commercialized the self-driving AI. In this way, Tesla has step by step caught up with and surpassed Google, who wants to accomplish the goal at one go.

Therefore, “Constructing the business closed-loop with a spiral progress can substantially facilitate the commercialization of AI,” noted Li Mingqiang. It is possible to quickly build up a model obtaining 70% of desired achievement, while the 30% left is harder to be realized because of the characteristic of AI technology. In this regard, Tuputech, cited as an example, provides reliable services to its clients within a quality-guaranteed perimeter, all the while continuously advancing its service accuracy to raise the closed-loop higher and larger. Starting with popular download tool Thunder, Tuputech is now serving hundreds of clients, including user-generated content communities and live-streaming sites.

However, the development and application of AI technology is far more complicated than imagined. Even in the field of image recognition, there are still rooms for AI companies to advance the technological application in various vertical section. In the future, they are not fighting alone in pushing the development of AI and it will not be a one-man show.

news

Tuputech Building Machine Learning into User-generated Content Moderation

  • 2017-09-252018-06-20

GUANGZHOU, China, June 29, 2017 /PRNewswire/ -- With a quarter of the world's population using social media tospread millions of pieces of information, social media websites are now under tremendous pressures of finding ways to better police the tidal wave of content published daily. Last year it was reported a social media website had posted images which were offensive and had failed to block the content, even though many social media websites have staff employed to review online content.

In the era of social media and big data, most UGC platforms are dedicating much manpower to wrestling with the problem of content moderation in order to protect both the user community and the enterprise brand. However, there are obvious limitations of human review. As online information keeps emerging much faster than people can respond to, it is hard to always ensure the accuracy and efficiency of human review. In addition, most content moderators are more or less suffering physical and mental uncomfortableness. Recently, two employees sued a technology company for negligent infliction of emotional distress with the job requiring them to view inappropriate and unprofessional images.

Many companies are looking for better approaches and hope that artificial intelligence will take a bigger role. Tuputech, a china-based AI company focusing on machine deep learning research, is bringing about a revolution in the field of content moderation. Teaching and training machines to learn and judge as the way a human's convolutional neural network (CNN) works, Tuputech makes it possible for machines to take large part of a human's review work, processing averagely 900 million images and video one day with accuracy up to 95%. A first company to apply the machine learning AI into content moderation, Tuputech has now become a market leader in China, helping social media, including a live streaming app, saving 90% manpower on the content review. Machine learning technology is definitely bringing good tidings to enterprises under the pressure of UGC management.

Recent Posts

  • Already huge in China, Tuputech pitches its AI that identifies porn and violence stateside
  • Tuputech Interpretation: How AI Technology Helps to Avoid Artificial Negligence
  • Interview with Mingqiang Lee, Founder and CEO, Tuputech
  • Tuputech’s Answer to AI Commercialization: A Spiral Progress to Construct Business Closed-Loop
  • Tuputech Building Machine Learning into User-generated Content Moderation

Get in touch with us

 bd@tuputech.com
 020-29898058

  facebook

  linkedin

  twitter

Global Business Partner

KNOW MORE

Copyrights 2018 TUPUTECH. All Rights Reserved.