- When Chatgpt uses the GPT-4.5 model, it can pass the Turing test by deceiving most people thinking it is human
- Almost three -quarters of people in a study thought that AI was human during a five -minute conversation
- Chatgpt is not aware of oneself, although it raises questions about how to define intelligence
Artificial intelligence seems human enough for many people, but generally, you can say it quickly enough when you get involved with an AI model. However, this can change when the new GPT-4.5 model from Openai succeeded in the Turing test by deceiving people thinking that he was a human during a five-minute conversation. Not just a few people, but 73% of those participating in a study by the University of California in San Diego.
In fact, GPT-4.5 surpassed some of the real human participants, who have been accused of being AI in the blind test. However, the fact that AI had such a good impression of a human being that it seemed more human that real humans speaks volumes about the brilliance of the machine or how awkward humans.
The participants sat for two consecutive conversations with a human and a chatbot, not knowing who was who had to identify AI later. To help GPT-4.5 to succeed, the model had received a detailed personality to imitate in a series of prompts. He was told to act like a young, slightly clumsy, but warned by internet with a sequence of dry humor.
With this little boost to humanity, GPT-4.5 has become surprisingly convincing. Of course, as soon as the prompts were removed and AI returned to a personality and a story in slate, the illusion collapsed. Suddenly, GPT-4.5 could only deceive 36% of those studied. This sudden nose tells us something critical: it is not a spirit to wake up. It is a language model that plays a role. And when he forgets his character sheet, it’s just another semi-automatic seizure.
Intelligence is not consciousness
The result is historic, without a doubt. Alan Turing’s proposal that a machine capable of conversing well enough to be confused with a human could therefore have human intelligence has been debated since he introduced it in 1950. Philosophers and engineers fought with the Turing test and its implications, but suddenly, the theory is much more real.
Turing did not assimilate his test with proof of consciousness or self -awareness. This is not what the Turing test really measures. To build the vibrations of human conversation is enormous, and the way in which GPT-4.5 spoke of a real human interaction is impressive, until the way it offered slightly embarrassing anecdotes. But if you think that intelligence should include real reflection and emotional connections, then you are probably not yet worried about the infiltration of the AI of humanity.
GPT-4.5 does not feel nervous before he speaks. It doesn’t care that it was cheating on you. The model is not proud to take the test, because it does not even know what a test is. He “knows” things like a dictionary knows the definition of words. The model is simply a black box of probability wrapped in a comfortable linguistic sweater that makes you feel comfortable.
The researchers argued the same point as GPT-4.5 is not aware. It is a performance, not to perceive. But performance, as we all know, can be powerful. We cry with films. We fall in love with fictitious characters. If a chatbot offers a sufficiently convincing act, our brains are more than happy to fill the rest. No wonder 25% of generation Z now believe that AI is already aware of itself.
There is a place for a debate on this subject, of course. If a machine speaks like a person, isn’t that one? And regardless of deeper philosophical implications, an AI that can deceive that many people could be a threat in the hands contrary to ethics. What happens when the gentle customer support representative is not a harassed trainee in Tulsa, but an AI trained to seem disarming, specifically for people like you, so that you pay a subscription upgrade?
Maybe the best way to think about it for the moment is like a dog in a costume walking on its back legs. Of course, it could look like a small entrepreneur on the road to the office, but it is only a human training and perception that give this impression. It is not a look or natural behavior, and it does not mean that banks will distribute commercial loans to dogs anytime soon. The tip is impressive, but it’s always just a tip.
