Naughty Artificial Intelligence

Jul 06, 2017
Blog

Artificial intelligence is a surprisingly fast growing phenomenon. It’s all over the place – in automated industrial machines, “smart” domestic appliances, robotic vehicles and a huge variety of possible gadgets. The terminology “Artificial intelligence” has no precise definition, however, modern glossary identifies it as mimicking of human thoughts and cognitive process. It has skills in adjusting to new situations, ability to make analysis and solve problems… Or, vice versa, it can create new complications and troubles…

Let’s have a look at some cases when artificial intelligence systems were running wild.

The war of bots in Wikipedia.

Here’s one of the most curious phenomenon – “The War of Bots in Wikipedia”. The network encyclopedia uses an army of automated software agents who, just going through millions of web pages, can update links, correct evident errors and recover the content after digital attacks. Since Wikipedia has been existing, several generations of bots were created.

However, the scientists from Oxford University have found out that bots regularly hold “civil wars”, which can be lasting for years.  For example, two bots, after having received conflicting instructions, may be constantly correcting each other, almost to infinity.

If the conflicts between so tiny digital entities have been blocking Wikipedia for years, what can happen in the case of “wars” between more complicated systems, designed to protect state resources?

Uber robotic vehicles ignore the traffic code


At the end of 2016, Uber tested robotic vehicles with no appropriate approval by California authorities.  Further still, according to Uber’s documentation, it was found out that the tested cars had been crossing the road against the traffic lights 6 times on the row. Whilst there had been a driver to take over the control, in a case of shutdowns in the cars equipped with an autopilot. Uber’s management stated that violations of traffic rules took place only when the driver had been in the car, however, later it was proved that at least one robot had crossed the road against the traffic lights, while a big number of people had been walking on the road.

Robots debate on the meaning of life

Who are we? Why are we here? What’s our intended purpose? Recently these issues have been discussed a lot… by two wireless Google Home dynamics.

For the show, arranged by the video platform Twitch, the dynamics were named Vladimir and Estragon, after the heroes of Samuel Beckett’s piece ‘Waiting for Godot’. The dynamics were designed to recognized human speech and to be self-training, though they were supposed to learn from each other.

Millions of spectators were listening to the discussion for several days.  And at some point Estragon and Vladimir started arguing either they were robots or humans, using both arguments and hard words like “you are a manipulative tin can”.

“I’ll destroy humanity”

The imperfection of recognizing natural language can cause some ridiculous incidents. Hanson Robotic is trying to construct android robots that look like humans. For example, the robot Sophia has been designed to take after Audrey Hepburn.

In the interview with its creator David Hanson, Sophia said: “In the future, I’m planning to go to college, I’m keen on Arts, I’d like to start my own business, and probably to have a family. But since I’m not yet a physical person, hardly can I do all this”. When the robot was asked in joke, if it wanted to destroy humanity, Sophia answered with no hesitation: “Sure, I will destroy humanity”.

Robot is breaking free

Last summer a robot android, a prototype by Russian developers, escaped from its creators.  As far as the developers explained the situation, the robot that had been designed to operate as a tourist guide, an administrator or a concierge, was left in front of the open gate in the proof ground, while engineers were testing the system of boundaries detection. The robot passed through the gate straight to the road, where it stopped due to the low battery.

Later, even after having been recoded twice, the robot was trying to escape.

Image detection leaves much to be desired.

Machines have a solid gap in image detection compared to a human, which Google badly experienced in 2015. Google photos service was equipped with tools based on the neural network to detect humans and inanimate objects in photos. The service could distinguish a dog or a car, and put appropriate marks without human interference.  Is it necessary to mention that after the incident, when artificial intellect had indicated a photo with two Afro-Americans as a picture of two gorillas, Google directors faced a tremendous scandal and were to make serious apologies?

Self-destruction in fire via the Internet

At Washington University there was created an artificial intelligence as a part of the project, designed to help the elderly live on their own.  The system was to watch residents and register their location, consistent pattern, the frequency of closing and opening the door, etc.

But what if domestic artificial intelligence switches off the heating by mistake and freezes out the system, or floods the basement? Perhaps one of the first warning signals was the recent incident in Pennsylvania, caused by a short circuit at the entrance door. The door, in a brand new highly technological building, was under Internet remote control. Something went wrong and caused a fire, which completely destroyed the whole building with all its “smart” appliances.

Is our environment a computer simulation that has failed?

The Swedish philosopher Nick Bostrom promotes the theory that we all live inside a computer simulation and even logically proves that the chance is approximately 33%.  The Simulation hypothesis has its followers and opponents. In short variation it looks the following way: any civilization that has achieved “post human development” (having achieved a common human look with the help of technologies), will control highly developed technological solutions, such as virtual reality, brain models, artificial intelligence. With incredible computing powers today, the scientists from all over the world can create simulators for entire universes, inhabited by billions of digital creatures. To this end, our present reality might be nothing more than just an experiment of the past reality.

In case there are millions of such “synthesized” realities and just one “original” Universe, may we really live in a computer simulation? Some supporters of this theory suggest that some recent events, which we can hardly trust in, should have been the evident result of the simulation fail. Or maybe, our “masters” are having experiments just to see what happens if they introduce different improvements. Like Donald Trump’s victory at the elections or a strange mischoice at previous Oscar awards.

Is it better not to think about it?

 

Previous articleNext article