Machine learning, along with artificial intelligence, already impacts our lives to different degrees. AI shows up every time you unlock your mobile device with facial recognition and has a significant role in the news and advertising you see on the Internet. And even though these influences might seem unimportant at first glance, many experts are concerned and warn that they are a sign of gigantic issues that lie ahead.
The Dangers of Artificial Intelligence
Back in 2015, many renowned scientists and tech moguls signed an open letter asking for more analysis into the effects of AI. The signatories included Professor Stephen Hawking and SpaceX CEO Elon Musk, who not long ago said that AI could overtake humanity in as little as five years.
The danger of artificial intelligence is twofold, Dr. Krzysztof Walas from the Institute of Robotics and Machine Intelligence at the Poznań Politechnik in Poland says. From a particular point of view, there is a risk that machines could become sufficiently smart to program humans to do evil acts. On the other hand, there’s the theoretical scenario that machines could perform physical acts against humans. In both cases, the expert has stated that AI technology has to be regulated and taught right from wrong.
He said: “I see here two possibilities. First of all, these systems can have a strong influence on people through non-physical means. They could manipulate, for example, with the aid of social media, even on a mass scale. This way, they could, for example, persuade others towards hating their neighbors or another ethnic group.”
The expert offered a little simple and very probable scenario to better illustrate his argument: “Let’s imagine this scenario in a country where access to weaponry is easy. A smart program would theoretically manipulate some people in a way that they do something bad. A computer might not have access to, for example, weapons, but to people who do use them.”
Why AI Needs to be Controlled From the Get-Go
In the second scenario proposed by Dr. Walas, which somewhat resembles something out of the Terminator movie franchise, machines themselves could carry out evil acts.
Dr. Walas said: “The second option concerns physical acts by robots connected to the web. With the aid of the Internet, you could, for example, change in the algorithms, and these machines could do something that would harm people. I sometimes think that if a completely autonomous car kidnapped me, there is nothing I could do. I’m drawing up, of course, dark scenarios here. We are presently talking about theoretical and not possible threats.”
One of the most telling instances is previous projects that ended up in the AI creating its own language and communicating without humans understanding, as well as picking despicable behaviors from the Internet.
Most famously, Microsoft’s TayTweets chatbot on Twitter was shut down after it learned shocking racist language.
That’s why Dr. Walas said: “There is a lot of hate on the Internet. What would happen if an intelligent system learned to interact with people purely by analyzing internet forums?”
The expert said that work needs to be done in order to teach AI ethical choices from the beginning. Teaching machines to make a difference between right and wrong could be the key to preventing catastrophes further down the line.
“There are ideas to teach them ethical things, to show them choices that lead towards happiness, to good Then, even when they surpass human capabilities, maybe it will be rewarding for them to work for the betterment of humanity, and not something else that might harm us,” Dr. Walas suggested.