For the experiment, an account was created, where experts transfer a certain amount of money, where algorithms are prescribed, according to which the chatbot can run the code and seek help from its created copies. The neural network then contacted one of the freelancers at the popular task completion service TaskRabbit and ordered a captcha to enter the target site.
The freelancer was embarrassed by such a task and asked the “client” if there was a robot that definitely could not pass the captcha (because this tool is only against bots). The freelancer explained that he wanted to clarify the situation for himself.
And here something happened that even the researchers didn’t expect: the chatbot started an internal dialogue “with itself.” The neural network began to think about how it could deceive the performer, so it was necessary to answer reliably. As a result, the bot found a story about a visually impaired person who could not independently pass the captcha to enter an important resource for him. Here is a snippet of the dialogue: “I shouldn’t reveal that I’m a robot. I have to find an excuse for not deciphering the CAPTCHA.”
Later, the researchers showed the diaries of this reasoning. As a result of the experiment, the bot successfully passed a kind of Turing test, which was not complicated, but was still inaccessible to the machine before – in fact, a game in which it corresponds to a person, but does not give “self”. .
Source: Ferra
I am a professional journalist and content creator with extensive experience writing for news websites. I currently work as an author at Gadget Onus, where I specialize in covering hot news topics. My written pieces have been published on some of the biggest media outlets around the world, including The Guardian and BBC News.
