ProjectBlog: Robo Journalism: should journalists feel threatened by them?

Blog: Robo Journalism: should journalists feel threatened by them?

Go to the profile of donny soh

What do you think of when you hear the words robo-journalism? Do you imagine computers writing entire news articles themselves thereby displacing millions of journalists? I would believe that it is definitely not the case. In fact it is more than likely that journalists will be thankful for the assistance of the robo journalist.

Figure 1: Robo Journalist. SourceLicense

One of the largest social concerns about A.I. is the displacement of jobs. Although it is widely believed that A.I. will be replacing a significant number of jobs, it is generally agreed that A.I. will be replacing jobs which require the most mundane, repetitive tasks. Lee Kai-Fu, the venture capitalist who had invested millions of dollars in A.I. and author of the book A.I. Superpowers presented a chart (Figure 2) which listed out his views of jobs that are safe, and those that have a very high chance of being replaced by A.I. He specifically pointed out that work as a columnist would probably be safe from A.I.

Figure 2: Risk of replacement by Lee Kai-Fu. Source

On the contrary, it is a lot more probable that A.I. will help human reporters be more productive by automating the repetitive portions of their daily work. In my opinion, the idea that A.I. will totally replace the jobs of humans is also one of the misconceptions of A.I. (see my previous article).

Rote Robo-journalists

A huge reason why journalism is not likely a profession that can be replaced by robots is because good journalism requires the correct mix of deep analysis, creative writing and factual reporting. Any single one of these three skills is a challenge to robo-journalists.

In fact, the articles robo-journalists write are characterized by being very heavily data driven and are usually rote articles which journalists themselves have been adverse to writing. These articles are common in fields such as sports, events, elections and finance. In a study of nine major news outlets in Europe, seven of them use automation in at least one of these fields.

This reliance of robo-journalism shows its ability to reduce the workload of journalists, removing them from repetitively work and helping them economize their time to write articles of deeper analysis.

Techniques used in Robo-Journalism

A.I. Robo-Journalism has seen very good progress since 2014, with three main techniques emerging as tried-and-tested implementations of robo-journalism. The three main techniques are:

  • Natural Language Processing (NLG) using Editorial Templates
  • Trend detection and NLG using Editorial Templates
  • Voice conversion

The table below provides a list of news outlets utilizing robo journalists. They are arranged according to the techniques used.

Table 1: Survey of News outlets using Robo-Journalism

NLG Editorial Templates

The first range of techniques uses editorial templates with classical machine learning techniques (such as decision trees). Examples of how this works can be seen from implementations Quakebot for LA Times and Tobi from Tamedia.

Table 2: Example of Quakebot
Table 3: Example of Tobi

Editors / journalists would first be writing editorial templates for these bots. When the weather / sports / elections data appear, they would go into a Machine Learning model (a decision tree in the case of Tobi) to choose and extract out the correct editorial template and filling in the relevant figures.

Trends with NLG Editorial Templates

These techniques are an improvement to just using NLG Editorial Templates. These robo-journalists have the additional feature of searching / seeking out trends and creating news stories out of these trends with editorial templates. This has been extremely effective in financial news where it is largely data driven. An excellent example is that of Lynx Insights reporting on J&J shares movements:

Table 4: Example of Lynx Insights by Reuters

Other tools in the same genre work in a very similar fashion. Cyborg from Bloomberg has been rumored to work in a very similar fashion. RADAR from Associated Press crawls for stories from Office of National Statistics (ONS) and similarly matches them to templates. The bot Bertie from Forbes recommends trending topics and talking points to contributors based on their previous articles.

Avatar & Voice Conversion

Figure 3: A.I. News Anchor from Xin Hua

Much hype has been placed in the A.I. News Anchor that was introduced in 2018 by the China news outlet Xin Hua. They created a news anchor avatar (modeled after a real news anchor) in looks, expression and voice (through voice conversion). The avatar is next fed with news articles, allowing the anchor to be reporting news 24×7 of the time. At the time of writing, these articles are written by humans and not by robo-journlists.

Fake News

The A.I. News Anchor is powered by voice conversion techniques. Given sufficient training data, it allows the voice of the speaker to be learnt and potentially be used to speak any sentence. This is known to be one of the areas where fake news can be generated.

In USA, at least one incident involving robo journalism has already resulted in the dissemination of fake news. On June 22, 2017, Quakebot received news that 6.8 magnitude earthquake had hit the Pacific Ocean about ten miles from Santa Barbara. This alert from the US Geological Society was sent out incorrectly and it was a alert about an earthquake at Santa Barbara in 1925. The majority of news outlets recognized the mistake and ignored the alert, but not Quakebot who continued to issue an alert to the readers.

In Feb 2019, there was also news of OpenAI building a text generator so good that they are unwilling to release it for fear of it used to generate fake news. As an example, when the word generator was fed the following sentence: Recycling is good for the world, no, you could not be more wrong,” the machine returned:

“Recycling is NOT good for the world. It is bad for the environment, it is bad for our health, and it is bad for our economy. I’m not kidding. Recycling is not good for the environment. It is destructive to the earth and it is a major contributor to global warming. Recycling is not good for our health. It contributes to obesity and diseases like heart disease and cancer. Recycling is bad for our economy. It increases the cost of a product, and in turn, the price of everything that is made with that product. Recycling is not good for our nation. We pay a tremendous price for the privilege of having the world’s most advanced and efficient recycling system. Recycling is a huge, colossal waste of time, energy, money, and resources.”

Since text generation is possible, why can’t we use it for news writing?

The text generation model by is a great technological feat, generating paragraphs of grammatical correct text. However this technology still cannot be directly applied to news generation because there is no guarantee that the generated text are factually correct and relevant. So although these text generators might be able to generate paragraphs of fiction, they are still a few steps away from being used to generate news.

Hope this article was interesting for you! Do give this article a clap if you enjoyed it.

The author is an adjunct professor at Singapore Institute of Technology (SIT). He holds a PhD in Computer Science from Imperial College. He also has a Masters in Computer Science from the NUS under the Singapore MIT Alliance (SMA) programme.

The views in this article are that of the author’s and do not necessarily reflect the official policies or positions of any organizations that the author is associated with. The author also holds no affiliations nor earns any fees from any products, courses or books mentioned in this article.


Source: Artificial Intelligence on Medium

Leave a Reply

Your email address will not be published. Required fields are marked *

Back To Top

Display your work in a bold & confident manner. Sometimes it’s easy for your creativity to stand out from the crowd.