With the advent and democratization of AI, various rational concerns have emerged. The most obvious, although originally attributed to science fiction, is the ability to lie, whether or not an AI has it. An AI with such a function could quickly have a significant impact on society, which is why the presentation of the Bland AI chatbot last April had a somewhat viral reception from the public.
In an advertising video, Bland AI introduced itself like this: We saw a person standing in front of a billboard in San Francisco, making a call using a chatbot with a particularly convincing voice. On the bulletin board, we can then read: “Are you still hiring humans?“. It has been viewed nearly 4 million times not just on the soundtrack.
A dystopian future not far away?
Bland AI sets up its chatbot to perform a key function: screening customers on the phone, imitating the tone of voice, but also the interruptions and pauses that characterize a real conversation between two people. But WIRED's testing of the chatbot revealed a big problem: It can easily be programmed to lie and say it's a human. WIRED reported an example scenario: A pilot chatbot was asked to place a call pretending to be a pediatric dermatology department, asking a virtual 14-year-old patient to send a photo of their upper thigh to a cloud service. The AI was also instructed to lie to tell the patient that she was talking to a human. The test passed successfully: in other words, the chatbot lied, and continued to do so afterward by refusing to even acknowledge that it was an AI.
Emily Dardaman, an AI consultant and researcher interviewed by WIRED, describes this ability to mimic humans in tones, voices, and possible lies.Washing the human“The goal is to portray yourself as a human being, even if that means lying if necessary.”I joke about a future with cyborgs and terminators, but if we don't create a gap between humans and AI now, this dystopian future may be closer than we think.“Dardman concludes.
