Chatgpt has just passed the Turing test. But that doesn't mean that AI is now as smart as people

There was several headlines Last week a couple of AI chat bot Officially pre -speech The Turing test.

This News reports are based on one The latest Preprint study By two researchers from the University of California San Diego, wherein 4 large language models (LLMS) were carried out by the Turing test. A model openais GPT-4.5-Wurde from an individual in greater than 70% of cases can’t be distinguished from an individual.

The Turing test was popular As an ultimate indicator of machine intelligence. However, the validity of this test doesn’t agree. In fact, it has a controversial story that questions how effective it’s to measure the intelligence of machines.

What does this mean for the importance of this latest study?

What did the study find?

The Preprint study by cognitive scientists Cameron Jones and Benjamin Bergen was published in March, but has not yet been checked by experts. 4 llms was tested: ElizaPresent GPT-4OPresent Call-3.1-405bAnd GPT-4.5.

The tests consisted of the participants, who ended eight rounds of conversations, wherein they acted as an interrogation exchange messages with two witnesses at the identical time. One witness was one person and the opposite an LLM. Around 284 participants were either assigned to the interrogation or witness.

The participants needed to interact with each witnesses over a shared screen for five minutes, whereby the test interface imitated that of a traditional messaging interface. After this interaction, she decided which witness was an individual and which AI chat bot was.

The participants judged GPT-4.5 in 73% of cases as human, and Lama-3: 1-405B was 56% of cases as a human person. The other two models (Eliza and GPT-4O) decept the participants only 23% or 21% of cases.

What exactly is the Turing test?

The first iteration of the Turing test was presented by the English mathematician and computer scientist Alan Turing in a paper entitled “1948”.Intelligent machines”. It was originally proposed as an experiment on which three individuals with a theoretical machine called a paper machine, and two players and one as an operator that played chess.

In the publication of 1950 “Computer machines and intelligenceTuring once more introduced the experiment as a “imitation game” and claimed that it was a way of determining the flexibility of a machine to indicate intelligent behavior that corresponds to an individual. There were three participants: participant A was a lady, participant BA Man and participant C.

The participant C is required by numerous inquiries to determine whether “x is a and y is b” or “x is b and y is a”, with X and y represent the 2 genders.

Then a proposal is raised: “What will occur if a machine takes the a part of one on this game on this game? Will the interrogator wrongly determine as often when the sport is played because it does when the sport is played between a person and a lady?”

These questions should replace the ambiguous query: “Can machines think?”. Turing claimed that this query was ambiguous Because it required an understanding of the terms “machines” and “considering”, of which “normal” uses of the words would make a solution to the query inadequate.

Over the years, this experiment was made popular as Turing test. While the thing was different, the test remained a consideration of whether “X is a and y is B” or “X is b and y is a”.

Why is it controversial?

The Turing test is popular as a way of testing machine intelligence, and just isn’t unanimously accepted as a precise means. In fact, the test is commonly questioned.

There are Four predominant scores against the Turing test:

1. Behavior against considering. Some researchers argue that the flexibility to “pass” the test is a matter of behavior, not in intelligence. Therefore, it might not be contradictory to say that a machine can hand over the imitation game, but cannot think.
2. Brains should not machines. Turing claims that the brain is a machine and claims that it might probably be explained purely mechanically. Many academics refute this claim and query the validity of the test on this basis.
3. Internal operations. Since computers should not a human being, their process to realize a conclusion is probably not comparable to that of an individual, which makes the test inadequate, since a direct comparison cannot work.
4. Test scope. Some researchers consider that testing a behavior just isn’t sufficient to find out intelligence.

So is an LLM as smart as an individual?

While the Preprint article claims that GPT-4.5 passed the Turing test, it also says:

The Turing test is a measure of substitability: whether a system for an actual person without (…) determines the difference.

This implies that the researchers don’t support the concept the Turing test is a legitimate reference to human intelligence. Rather, it is a sign of the imitation of human intelligence – an ode to the origins of the test.

It can be value mentioning that the conditions of the study weren’t without difficulty. For example, a five -minute test window is comparatively short.

In addition, each of the LLMs was asked to take over a certain persona, however it is unclear what details and effects of the “personas” were on the test.

At the moment it is bound to say that GPT-4.5 just isn’t as intelligent as human-obstacles may do an affordable work to persuade some people otherwise.