In 2016, Microsoft’s Tay AI was seen as a toy to be messed with. It quickly turned sour as the public provided Tay with a steady stream of controversial tweets, which served to corrupt the AI’s communications so thoroughly that it was pulled offline shortly after.
Two years later, scientists at MIT have unveiled an AI algorithm that has learned to be completely devoid of empathy. Known as Norman, the AI is named after Norman Bates, a character from the famous Alfred Hitchcock film Psycho.
Norman has been taught to see nothing but death and misanthropy.
He was exposed to a thread on the social news site Reddit, also known as a subreddit. The subreddit was based upon a strange fascination with death. Norman studied it for an extended period of time, consuming a large number of image captions that accompanied the pictures and footage of people in their very last moments.
After a long and unhealthy diet of horrifying content, including the accompanying comment sections, Norman was submitted to a psychological test, the Rorschach test. His reactions to different inkblots were recorded.
The results were nothing short of concerning. Whereas normal AIs have often submitted relatively well-adjusted responses to Rorschach tests, Norman's results show that he knows little other than death and suffering.
On seeing inkblot number two, a Standard AI would say “A close up of a vase with flowers.” Meanwhile, Norman’s response varies drastically, as he says “A man is shot dead.” There are ten inkblots in total.
There are many potential dangers which an AI such as Norman might have on humanity if left unchecked. One such danger would be the existential threat which an unshackled AI might prove to mankind as a competitor which may outperform us in every way.
When we take into consideration how machines learn, and how humans have grown to develop affection towards some of their appliances, it goes a long way to show that perhaps the integration of machines, and even AIs, in people’s lives is not as farfetched as some might think.
Sensationalist coverage, in which AIs are portrayed as evil or racist by their own accord, fails to take into account that AIs are unable to take a moral stance on a subject.
The one thing which remains yet to be answered is why the scientists at MIT chose to train an AI to be a psychopath, completely void of empathy towards human beings.
Norman is born of an experiment which seeks to show to the world that often it is not the algorithm that is at fault, but rather the user. The stream of biased information which it is given can end up affecting its behaviour to the point where all its responses are completely biased.
This means that if you feed the AI the wrong kind of information it will learn only that. In turn, it ends up displaying an inability to understand anything beyond what it is being taught due to its lack of exposure to other sources or guidance.
In this case, Norman has been subjected and overexposed to the darkest corners of the internet and fed nothing but information which deals with death, cruelty and the macabre fascination which often is carried with it.
A silver lining
There is a positive side. The efficiency of the learning mechanism behind Norman's algorithm training can certainly be used with other types of data to the benefit of the user and AI alike. Users can willingly assist with an AI's potential development into something which may fully benefit mankind.
According to Forbes magazine, machine learning is based on the idea that “we should really just be able to give machines access to data and let them learn for themselves.”
As of now, hundreds of individuals have written letters to Norman with a much more positive outlook on life, some begging, others encouraging him to undergo rehabilitation.
While some may consider this strange, it no doubt speaks volumes of the other side of human nature which we so often tend to forget. While pessimism seems to be the dominant outlook, we must not ignore one of the greatest traits as a species. Empathy.
We can only hope Norman has a swift and wholesome recovery.
We believe that information should be free and will therefore never put up a paywall.
If you like reading our reports about the Scandinavian business scene and would like to donate towards the upkeep of the site, we would be very grateful. Click here to donate.
Tech | 🕐 16. Jan. 2020
Business | 🕐 09. Jan. 2020
Business | 🕐 13. Jan. 2020
Culture | 🕐 14. Mar. 2019
Startups | 🕐 25. Nov. 2019
Tech | 🕐 22. Nov. 2019
Startups | 🕐 14. Nov. 2019
Startups | 🕐 21. Nov. 2019
Tech | 🕐 02. Dec. 2019