Why Artificial Intelligence Can’t Completely Replace Humans in Governing the Internet

Reviewers suffer because they see the worst side of the web and thus suffer (Elizabeth Dwoskin, Jeanne Whalen, Regine Cabato, 2019),so to avoid this for the reviewers most of the companies are using AI machines to review the content as well as to manage the community but most of the platforms won’t be relying entirely on the machine review but rather they will be using both human and smart machines.

Advantages of Artificial Intelligence

Artificial intelligence has become more and more mature with the development of technology, compared with human, artificial intelligence audit can do 24 hours a day continuous work. For the platform to use Artificial intelligence to manage the community not only saves money but also will be more efficient. For example, Artificial intelligence can quickly identify offending words in text and analyse images based on powerful algorithms. For example, on tt (a Chinese chat software) it has set up a list of offending words, so that no political, violent or pornographic statements can be posted. The rooms that are overcrowded are checked and accounts that are likely to repeat the offence are watched. In order to protect minors in China, all gaming software now requires real-name registration to prevent minors from becoming addicted to gaming. Today’s audio and video development is very rapid, and because of the emergence of short video platforms, today is also a world where everyone can become a blogger, so this time in the face of a large number of video works, the use of manual efficiency is very low, the existence of artificial intelligence review is particularly important, based on their algorithms, extracting the images in the video for review. If you use a human reviewer can not believe what a huge amount of work. So giving these simple and repetitive tasks to AI can reduce the review time.

Artificial intelligence systems learn based on their own data, so their accuracy and effectiveness improves over time. Compared to human review, Artificial intelligence is more impartial and free from personal feelings. Therefore, the review results given by Artificial intelligence are more impartial. And because Artificial intelligence has no emotions, the worst aspects of the internet don’t cause it any pain.

Disadvantages of Artificial Intelligence

Machine intelligence can only assist in recognition. For example, in a photo the Artificial intelligence can classify the flesh according to its nudity. But it can’t fully understand the true meaning of the photo. In the readings of week6, a photo called Napalm Girl (Gillespie Tarleton, 2018) is mentioned. The girl stripped off her clothes to protect herself and escape during the Vietnam War in 1972, which happened to be captured by journalist Nick Ut. However due to child nudity it would not pass muster in an AI review. But there is no denying the value of this photo its existence proves that war is indiscriminate for civilians and will make people reflect on the dangers of war. However due to the nudity of the flesh it tends to fail to spread in most communities and this is when the presence of human reviewers becomes even more important. So I think governing the Internet also requires ethical and moral judgements, and I think these judgements need to be underpinned by human values, emotions and context. So I think Artificial intelligence lacks the ability to make ethical and moral decisions.

Artificial Intelligence review is not flexible, especially in Chinese there are a lot of pronunciation is the same but the words are different, and then for example, (water fire) is a sensitive word, the system detects that the two words together will not pass the review, but if you add some special symbols in the two words, such as commas. (Water, fire) this time the Artificial Intelligence audit may not detect violations, because water and fire itself is not a sensitive word, and not put together so this time the Artificial Intelligence will pass the audit. So I think Artificial intelligence review is not flexible, it is more rigid.

As the world evolves and the word spreads, humans are more capable of adapting to these changes and updating their laws to deal with different situations. Different regions have different policies and rules, however it is the government that makes these rules and it is the administrators of the platforms that make these rules and there are various interests involved in these rules, so Artificial intelligence will only assist in analysing the data and humans cannot and will not leave the decision in the hands of Artificial intelligence. This is because human reviewers are better able to predict the future with the aid of emotions.

Artificial intelligence cannot take responsibility for the outcome of decisions, and people cannot expect machines without their own emotions to explain their decision-making behaviour. People need more persuasive people to explain and take responsibility for the outcome of decisions. Take for example the Napalm Girl incident on Facebook. If it was an Artificial intelligence that went and deleted the photo, it was only based on its own data that it was child nudity and that it was against the law. It can only give those explanations. However a human would explain that it was a tough decision because the photo had both historical significance and a violation. So at this point it’s up to the human censor to decide if this photo is a violation, as well as to take on the consequences of the decision.

Bias and discrimination. Bias and discrimination are also present in Artificial intelligence, as Artificial intelligence systems may perpetuate past biases that can lead to discriminatory outcomes. For example, in a recruitment tool, it collects CVs and acceptance information from the last ten years and it will find that in the past male candidates were more preferred than female candidates in the same conditions. So based on the accumulation of data, it can also result in discriminatory outcomes. Today, however, humans are actively addressing these biases and trying to create fairer policies.


All in all I think Artificial intelligence does save time and money and doesn’t take the pain, but also AI has many drawbacks such as inflexibility and discrimination based on data and Artificial intelligence can’t take responsibility for the outcome of decisions. And it can’t make decisions beyond the data because it doesn’t have feelings. So I don’t think Artificial intelligence can ever completely replace human auditing. That’s why so many platforms in the market use dual review by Artificial intelligence and humans. Of course I’m sure AI will get better and better as the technology develops, but I still don’t think the world will eventually develop into a world ruled by Artificial intelligence.

Reference List:

Elizabeth Dwoskin, Jeanne Whalen, Regine Cabato(July 25, 2019 at 1:00 a.m. EDT)Content moderators at YouTube, Facebook and Twitter see the worst of the web — and suffer silently. Content moderators at YouTube, Facebook and Twitter see the worst of the web and suffer silently – The Washington Post

Gillespie Tarleton, 2018. Custodians of the Internet : Platforms, Content Moderation, and the Hidden Decisions That Shape Social Media,  CHAPTER 1 All Platforms Moderate, (1-23)

Vietnam Napalm , (1972 Nick Ut – USA | www.worldpressphoto.org/gallery/themes/… | Flickr)by Nick Ut(CC0 1.0 Universal)