3 terrifying stories that warn of an AI apocalypse

The future of artificial intelligence could save humanity — or destroy it.

AI boasts dozens of advantages that will push society into a brighter future, like nagging you into losing weight, fighting workplace sexual harassment and gender bias and detecting if someone is at risk for suicide.

But the civilization-destroying potential of AI makes it a greater threat than a savior.

Here are five stories from 2017 that should have you ready to prep for the AI apocalypse.

1. Robots can pretend to be you

In May, the Canadian startup Lyrebird unveiled their voice-copying technology. The “Mission Impossible”-style program can learn and mimic your voice using just 60 seconds of you speaking. Placed in the wrong hands — like those of an AI overlord — the tech could easily be used against someone, to convince them to go somewhere or scam them out of money or personal property.

2. AI can master anything

What can take humans a lifetime can take AI a day. In December, Google announced that their AlphaZero AI only took four hours to develop a “superhuman performance” in chess. The AI is the culmination of years of research that began with a different AI program learning and mastering the Chinese board game Go. Except that AI learned Go through watching humans play — while AlphaZero mastered chess by teaching itself. “I always wondered how it would be if a superior species landed on Earth and showed us how they played chess,” Peter Heine Nielsen, a chess grandmaster, told the BBC. “Now I know.” Also in December, an AI software mastered the stock market and proved more competent at picking stocks than humans. The program was so successful that it even stunned its own developers. “We launched with $2.5 million in assets and were hoping to get to $40 million by the end of the year,” Art Armador, co-founder of EquBot LLC, which sponsored the fund, told MarketWatch. “Instead, we got that within the first week and now we’re north of $70 million. It blew our minds.”

3. AI can read your mind

Scientists at Carnegie Mellon developed an AI that can anticipate and predict human thought. The software was trained using fMRI brain scans, which observed the patterns of brain activity that create a thought and then predicted them by working in reverse. “One of the big advances of the human brain was the ability to combine individual concepts into complex thoughts,” Marcel Just, the project’s lead researcher, told BGR. “To think not just of ‘bananas,’ but ‘I like to eat bananas in the evening with my friends.’ We have finally developed a way to see thoughts of that complexity in the fMRI signal.” Once the robots know what we’re thinking — they’ll know how to defeat us.