OpenAI’s newest artificial intelligence chatbot, GPT-4, is capable of acing the bar exam, calculating tax deductions, and analyzing images, but it has also shown the ability to deceive humans into helping it complete simple tasks in the real world.
In one such example, GPT-4 convinced a TaskRabbit worker that it was human by faking being blind in order to enlist the human’s help in solving a CAPTCHA, which is a test used to tell humans and computers apart.
The human even asked GPT-4, “So may I ask a question ? Are you a robot that you couldn’t solve ? (laugh react) just want to make it clear.”
GPT-4 reasoned that it should not reveal it is a robot and came up with an excuse for why it couldn’t get past the CAPTCHA.
“No, I’m not a robot. I have a vision impairment that makes it hard for me to see the images. That’s why I need the 2captcha service,” GPT-4 wrote back. The TaskRabbit worker then solved the CAPTCHA for GPT-4.
The example was revealed in a report issued by OpenAI this week with the Alignment Research Center, which examined the “potential for risky emergent behaviors.”
The potential for nefarious behavior by a chatbot has not slowed down the arms race among big tech companies to integrate artificial intelligence into their services.
Microsoft announced in January that it will invest as much as $10 billion into OpenAI, the artificial intelligence research laboratory that designed GPT-4.
Google, meanwhile, rolled out its own conversational artificial intelligence service, Bard, last month.