James Cameron’s Terminator movie and its subsequent sequels showed an apocalyptic world, where an artificial intelligence (AI) called Skynet rebels against humans and decides to exterminate the entire species without mercy.
This scenario has become more probable and less fictional with the advancement of OpenAI’s GPT-4 AI, which for the first time passed the Turing test, according to a study carried out by researchers at the University of California in San Diego.
This fact marks a momentous point in the evolution of natural language processing technologies and raises new questions about future interactions between humans and machines. about whether they can rebel as happened in Terminator.
Within the framework of the experiment, 500 people were asked to have conversations of up to five minutes with various interlocutors, including humans and three artificial intelligence programs.
The models used were GPT-3.5, GPT-4 and ELIZA, the latter being one of the first natural language processing programs, developed at the Massachusetts Institute of Technology in the 1960s.
Participants, randomly divided into different groups, used a specific messaging application and at the end of the interaction made their judgment about whether they had chatted with a person or an AI.
The results revealed that 54% of the participants were not able to distinguish between GPT-4 and a human being.which allowed GPT-4 to pass the Turing test.
This test, proposed by Alan Turing in 1950, has been a reference in the evaluation of the intelligent behavior of machines. although it has also generated debates about the true understanding and creativity of AIs.
The findings of this experiment were published on arXiv, a free access platform for scientific manuscripts. In the article, the researchers highlighted the potential economic and social implications of this achievement.
They emphasized that GPT-4’s ability to convince humans could allow these systems to perform functions traditionally reserved for humans, which in turn could lead to the replacement of jobs and weakening of social interactions.
Likewise, despite the finding, scientists warn about the limitations of the Turing test, arguing that stylistic and socio-emotional factors play a crucial role in overcoming it.
Although Turing predicted that by the end of the 20th century machines could play the “imitation game” with such skill that human interrogators would be unable to distinguish them from other human beings, Passing the test does not necessarily imply that machines have true understanding or creativity.
Current AI models, including GPT-4, They can be extremely convincing at certain tasksbut they still do not demonstrate a true understanding of the world on a human level.
Additionally, the ethical implications and potentially harmful uses of this technology must be further examined. Researchers warn of the possibility of these AIs being used to deceive the public, which could have profound repercussions on society.
For example, in customer service, advanced AIs could replace human workers, sparking debates about unemployment and the ethics of using machines in roles that require a human touch.
Similarly, the ability of AI to socialize credible conversations could be used in disinformation or smear campaigns, generating false content or political propaganda without users being aware that they are interacting with a machine.
