A 100-page report written by artificial intelligence experts from industry and academia has a clear message: Every AI advance by the good guys is an advance for the bad guys, too.
The paper, titled “The Malicious Use of Artificial Intelligence: Forecasting, Prevention, and Mitigation,” calls this the “dual-use” attribute of AI, meaning the technology’s ability to make thousands of complex decisions every second could be used to both help or harm people, depending on the person designing the system. The experts considered the malicious uses of AI that either currently exists or could be developed over the next five years, and broke them out into three groups: digital, physical, and political.
Here is a selected list of the potential harms discussed:
- Automated phishing, or creating fake emails, websites, and links to steal information.
- Faster hacking, through the automated discovery of vulnerabilities in software.
- Fooling AI systems, by taking advantage of the flaws in how AI sees the world.
- Automating terrorism, by using commercial drones or autonomous vehicles as weapons.
- Robot swarms, enabled by many autonomous robots trying to achieve the same goal.
- Remote attacks, since autonomous robots wouldn’t need to be controlled within any set distance.
- Propaganda, through easily-generated fake images and video.
- Automatic dissent removal, by automatically finding and removing text or images.
- Personalized persuasion, taking advantage of publicly-available information to target someone’s opinions.
The report paints a bleak picture of our potential future, especially since the timeframe is a mere five years. But it doesn’t mean we’re resigned to dystopia. Researchers are already working on some potential solutions to these problems, though they warn it will likely be a cat-and-mouse game.