Don't worry, if no one does what the AI tells them to do, we're not in danger. A Auto-GPT user, an open source standalone application powered by OpenAI's GPT-4 language model, has created its own artificial intelligence called ChaosGPT. She asked him to find a way to destroy humanity. The experimenter gave the AI some leeway, and allowed it to Google and even post on Twitter.
A user created ChaosGPT from ChatGPT. The problem is that the AI in question wants to annihilate humanity. Will he succeed?
The first thing that ChaosGPT suggested was that of cause a nuclear conflict, which, in general, was able to learn from the popular “Terminator” and other science fiction films. ChaosGPT then "hired" a GPT-3.5-based AI agent for the purpose of conducting research into lethal weapons. When the AI agent said they only support world, ChaosGPT has circumvented the limitations of the neural network subject to it and began to act outside the framework of the program to fulfill its task.
Read also: ChatGPT: news from the Guarantor. That's when he will return to Italy
Meanwhile, ChaosGPT has posted Tweets that would fit the expressions of some archetypal villains very well. Most likely, this means nothing, because the neural network she simply did what she was asked to do based on what he found in other people's statements available on the Internet. But it makes you think. In short, don't worry. Until AI and all the generative models that come after it are able to DO in the truest sense of the word, there is no danger of mass extinctions or the like.
It all starts with man. AI like ChaosGPT could be programmed for carry out violent or harmful actions towards human beings or of the environment, if these were considered "opportune" on the basis of the established objective. AI could also be used to discriminate against certain groups of people systematically, based on characteristics such as ethnicity, gender, religion or sexual orientation.
Furthermore, if it were programmed to maximize a certain function or objective (such as profit), it could behave in ways that are unexpected or even harmful to humanity, as is the case today in many industries where artificial intelligence is employed without due consideration. attention to its ethical and social management.
