Toutiao – Will Machine Learning take out humans out of the equation?
Toutiao started out as a news aggregating platform that leverages machine learning to provide customized content to the reader. Nevertheless, today with a valuation greater than the combination of NewsCorp and New York Times and more than 100 million Daily Active Users[1], it has evolved into a media giant that does more than aggregating and delivering personalized content.
How does Toutiao create value?
Toutiao collects content online and recommends the relevant content to the reader, based on his or her interests[2]. Toutiao initially presents general content to a first-time user and then uses machine learning to learn about the reader’s learning habits and interests through time. As the reader increases his or her interaction with the platform, Toutiao learns more about the user and provides more relevant content, which would make the user use the platform even more, making the platform even better at personalizing content. This ability to use machine learning to target specific content that the user enjoys creates value to the user but also through machine learning, the platform can capture that value to create even more value.
Furthermore, Toutiao has leveraged the data that it has collected and the learning it has made to create its own content through an artificial reporter called Xiaoming[3]. This is another form, whereby Toutiao creates value, which can be captured through learning that will create even more value over time.
Using machine learning therefore can be a valuable tool to fend off competitors as the incumbent. Nevertheless, Toutiao understands that machine learning on its own is not sufficient to create value and compete with other media giants. Toutiao has done so by also taking advantage of crowd sourcing by attracting and rewarding a community of content creators by inciting them to create content specifically for the Toutiao platform[4]. Toutiao understands to maximize value creation, it needs to rely on a community of content providers, which enables to make its content more unique to its platform.
How does Toutiao capture value?
Toutiao captures value through advertising, with estimated revenue for 2017 to be at $2.5 billion (or RMB 15 billion)[5]. As the platform learns about the reader, the platform can provide targeted advertising that is very much relevant to the consumers. It therefore creates value to advertisers by providing them a means to reach out a large population of more than 180 million Monthly Active Users[6], whilst knowing these consumers so well it can target them in an effective and cheap manner [7].
Machine vs Human
Although Toutiao has developed a very valuable value proposition with its machine learning capability, it has increasingly relied over top talent, as it offers some of the top salaries in China with packages reported to be at $3 million for AI engineers. As of the end of last year, Toutiao was looking to hire an additional 200 AI engineers[8]
At the beginning of the year, Toutiao was shut-down for 24 hours by the Chinese government [9] as its technology failed to wield out content of pornographic nature. This is utterly important as Toutiao relies on machine learning to identify fake news and filter content[10]. Although Toutiao blamed Baidu for providing the content, Toutiao solution now relies on a hybrid one, as it hired over 1000 human editors to oversee content review[11].
It seems that the quest to fully replace humans by machine is one that will not happen overnight, as we covered in class, it seems that the place for the human will shift but there might always be the place for human contributors – in this instance, whether it is in improving the technology with AI engineers, interpreting the data and making decisions with crowdsourced editors and internal management, or overseeing that the technology does not fail or get exploited with human editors.
[1] https://www.bloomberg.com/news/articles/2017-09-24/in-battle-for-talent-one-startup-founder-tries-unlimited-pay
[2] https://technode.com/2017/08/21/toutiao-bat-alarm-bells/
[3]https://www.technologyreview.com/s/603351/the-insanely-popular-chinese-news-app-that-youve-never-heard-of/
[4] http://www.chinadaily.com.cn/business/2017-11/23/content_34880992.htm
[5] https://techcrunch.com/2017/11/10/what-exactly-is-bytedance-and-toutiao/
[6] https://walkthechat.com/toutiao-rising-chinese-tech-giant/
[7] https://technode.com/2017/09/01/how-jinri-toutiao-can-help-you-grow-your-business/
[8] https://www.bloomberg.com/news/articles/2017-09-24/in-battle-for-talent-one-startup-founder-tries-unlimited-pay
[9] https://techcrunch.com/2018/01/02/bytedance-enters-2018-with-some-uncertainty/
[10] https://www.technologyreview.com/s/603351/the-insanely-popular-chinese-news-app-that-youve-never-heard-of/
[11] https://www.ft.com/content/c39547dc-f107-11e7-b220-857e26d1aca4
Amazing !
As with all AI-related topics, I always wonder what the impact of regulation might be going forward.
While being in China doesn’t necessarely mean that the government will claim the technology for itself, one might still wonder how and to what extent this very government might chose to regulate or control these types of innovations.
Furthermore, there is somewhat of an existential risk for a company that is giving the people exactly what they want in terms of media and entertainment to “upset” or at least put into question any forms of government-enforced restrictions.
Yes. It looks like Toutiao is looking to expand globally, but it is to be seen – as their plans are not very clear. The Chinese government’s move on closing their website was a definetely a setback for them.
I agree on the existential risk of giving people what they want. However, it does work and people are addicted to the app. Furthermore, there could be regulations on that in the future, with Facebook scandal unfolding, and the negative aspects of influencing people in certain ways emanating in the public eye. Nevertheless, Toutiao claims that they do not feed what people want to read or content that feeds off what their beliefs and convictions to reinforce them, they do claim that instead they provide content that might be of interest to the reader and do provide content that the reader wouldn’t have expected. The want vs. interested in is interesting. However, I don’t have any proof as to what extent this is true. I think we would never know as the algorithm is a black box to us. I do believe though if they have mastered a way to provide content that can make a person more well-read or more well-informed, and not necessarily a puppet to certain beliefs, then this means they have high chances of making a mark in other markets, in lights of what is happening with Facebook. Although, it is going to be a challenge in Western markets, as it can be argued that Chinese companies are usually discriminated upon in those markets.
Thanks MAB for a very interesting, and timely, post! I am pleased to hear that Toutiao provides a range of opinions in the information it provides. Indeed, that is something I would support Facebook and even the cable networks in the US doing for the viewers. However, I worry about this range of opinions weakening their ability to capture value — people often very much like living in an echo chamber, and Toutiao may have to move to a more pleasing algorithm over time to maintain users if other services begin to compete. Nevertheless, I’m glad to hear they are starting with at least some diversity of opinion.