In 1950, Alan Turing proposed to the world the famous “Turing test“: an experiment in which a machine, interacting with a human interrogator, attempts to convince him that it is a person. After 74 years, we are perhaps closer than ever to one final answer.
Studies with the Turing test
Recent Education conducted at the University of California suggest that the modern artificial intelligences they can trick humans, at least in some contexts. This phenomenon raises questions fundamentals onevolution of AI and on our perceptions of intelligence and humanity.
While Alan Turing initially conceived the test as an instance of intelligence measurementmany critics have raised doubts on his validity in evaluating true understanding and consciousness. Human beings, in fact, have an innate tendency to project human attributes on non-human objects and phenomena, a phenomenon known as anthropomorphization.
In the recent study, both ChatGPT-4 and ChatGPT-3.5 achieved a 50% conviction rate among human participants, a significant but still far from definitive result. These models were compared with ELIZAone of the first examples of AI developed in 1960 at MIT, which achieved a modest 22% conviction rate.
Second Nell Watsonartificial intelligence expert at the IEEE, ELIZA was limited by canned responseswhile modern AIs like ChatGPT are capable of one smoother conversation and contextually sensitive. This progress raises the question of whether ChatGPT has actually passed the Turing Test, at least in part.
What it means to be human
THE results they also suggest that the Turing test may not be a definitive measure of intelligence or consciousness, but rather a probabilistic indicator. Test participants were more inclined to focus on items such as lo linguistic style and the ability to socio-emotional interactionrather than on formal knowledge and reasoning.
L’evolution of AI continues to challenge our conceptions of what it means to be “intelligent” And “human“. Recent studies have shown that AI can help identify variants of depressionsuggesting potential future applications in understanding and treating mental illnesses.
