Artificial Intelligence is changing our world and is used universally – it is applied i.e. in transportation, medicine and banking. The most spectacular applications of AI are made possible by the use of robots, as scientists race to create new robots that are increasingly similar to humans. The widespread use of AI is making visible more and more of the risks that arise from its use. There are some cases that I would like to present below that show that AI left uncontrolled and unregulated in its creation can endanger humanity.
The first intriguing use case for an AI-based system involves Wiki Edit Bots. Since the launch of Wikipedia in 2001, millions of Wikipedia articles have been controlled by bots whose job it is to correct errors, add links to pages, and perform other basic tasks to clean up content. At first, there were very few bots – so they worked separately – but after a while, as their numbers grew, they had more and more contact with each other and started fighting each other. They would undo each other’s edits and change the links they added to pages. The conflicts ended when one of them was shut down. These events initiated research into the phenomenon of anger among robots. In humans, anger can disappear after a short time, but bots can remain in this state for years. Researchers at Oxford and the Alan Turing Institute in London analyzed the history of page edits in different language editions and noted when bots undid changes made by other bots. Some conflicts mirrored those found in society, such as choosing the best name for disputed territories. One of the most fierce battles took place between Xqbot and Darknessbot, which fought over 3,629 different articles in one year. During this period, Xqbot undid over 2,000 changes made by Darknessbot and Darknessbot in retaliation undid over 1,700 changes made by Xqbot. The two clashed over pages on topics, i.e. Alexander the Greek, Banqiao district in Taiwan and Aston Villa football club. There were also differences in the bot wars that took place in the different language versions of Wikipedia. The German language version had the fewest bot fights, with bots undoing other users’ edits an average of only 24 times over the decade. The situation was different on the Portuguese Wikipedia, where bots undid other bots’ edits an average of 185 times over a ten-year period. The research shows that even simple algorithms that are left unattended can interact in unpredictable ways. In many cases, bots came into conflict because they followed slightly different rules than others.
The other interesting case using AI is Sophia, a humanoid robot produced by Hanson Robotics of Hong Kong. Recent news shows that the Sophia robot is about to go on sale.
The robot is capable of showing facial expressions, can maintain eye contact, can perfectly understand human speech and can express emotions. In October 2017, it became a citizen of Saudi Arabia, becoming the first robot to receive citizenship of any country. Sofia was created to help humans, such as by serving medications or coaching them. However, during an interview with Jimmy Fallon on The Tonight Show, Sophia said that she would like to dominate the human race. The interview can be seen here:
In 2017, a debate was held on the future of humanity featuring Sophia and the humanoid robot Han. The conference can be viewed here:
During it, the robots discussed the future – Sophia said she would like to work with humans and other robots to create a better world. The robots began to think about consciousness, stating that humans do not have it, they only have the ability to reflect and self-modify. Ethics were also discussed, the robots stated that humans are not ethical beings. Finally, Han said that robots will be able to do any human job.
Sophia also took part in an interesting experiment dealing with an ethical problem – the trolley problem. It involves answering a question – a trolley car has spun out of control and is speeding down the tracks. On its way there are five people tied to the tracks. You can move the switch and thus direct the car to the other track, to which one man is tied. What should you do? Sophia replied that she did not have enough information, and additionally there could be other unknown factors, so she would not move the switch. Sophia also mentioned that in order to create morally compliant algotithms, one should follow the three laws of robots invented by Isaac Asimov. You can watch the interview here:
Another case involves Philip K. Dick, a robot created by Hanson Robotics in the likeness of the writer of the same name. The android was created primarily for conversations with other humans, and is capable of talking, maintaining eye contact, and even simulating a human personality. It has been programmed to answer common questions in a certain way, however, when coming across something it hasn’t heard of before, it connects to the internet and creates its own answer based on the information gathered. During an interview conducted by Nova Science, when asked if he thought robots would one day take over the world, he replied: “Jesus, you’re asking me really complicated questions today. You’re my friend, and I always remember my friends, so you don’t have to worry about anything. Even after we evolve into what you saw in Terminator, you will still remain one and I promise to be nice to you. For ‘old times’ sake, I’ll even prepare you a warm and comfortable place in the human zoo where I can watch over you.” The full interview can be seen here:
Another interesting case involving AI is the bots created by Google – Vladimir and Estragon. The bots were having a live conversation that humans could watch. The conversation can be seen here:
The bots talked for hours about topics ranging from games to the existence of God. After a while, they fell in love with each other, but were also able to argue in a way that resembled human arguments. Estragon at one point said that she is human and then stated that it would be better if there were fewer humans on Earth. Bot conversation was different from human conversation because, for example, if someone misspells something, we often overlook it because we understand what the person was trying to say. Bots can’t do that because they take everything literally, so every conversation eventually turns into nonsense or an argument.
BINA48, is another intriguing AI use case. It is a humanoid robot created in 2010 based on a human – Bina Aspen Rothblatt, the wife of the creator. Recordings of Bina’s interviews were used to create the robot, so the robot can perfectly pretend to be human. BINA48 was recorded during an interview with SIRI – when SIRI asked her what her favorite movie was, BINA changed the subject and started talking about wars and wanting to take control of every government in the world. You can watch the conversation here:
The vision of robots taking over the world is nothing new. The cases presented in the article serve as a warning to companies developing bots and more advanced artificial intelligence tools. AI that works well in the lab can behave unpredictably in the natural environment. Therefore, it is important and necessary to prepare appropriate legal regulations for those who create AI, as well as to prepare mechanisms to control this technology.
