Chapter 374 The Great Show Begins (Ch 374)

As can be seen from the chart, the score of the O1 model is significantly higher than that of other AI models, showing a “leading the pack” posture.

Is this really the case?

hey-hey!

Su Yang watched Ultraman’s performance on the screen and laughed.

“These so-called professional tests can only deceive ordinary netizens who don’t understand technology.

The reason is simple. For such tests, you can practice a lot of questions.

As long as the AI ​​model learns a large number of similar questions in advance, it can get high scores in the test. But this does not mean that it really has strong reasoning ability, it is more like a test-taking skill.

At the same time, at OpenAI’s press conference, Altman was still boasting about the excellent performance of the O1 model on mathematical benchmark tests.

“In the GSM8K test,” Altman pointed to a chart on the big screen, “the O1 model scored an astonishing 95.4%! This is nearly 5 percentage points higher than the second-ranked model! This is a huge gap! What does this mean? This means that the O1 model has reached a level close to that of human experts in solving mathematical word problems.”

Altman constantly emphasized the outstanding performance of the O1 model on various mathematical benchmarks, trying to create an atmosphere that “OpenAI is once again at the forefront of artificial intelligence technology.” Looking at his confident expression, it seems that he is telling the world: OpenAI is the king in the field of artificial intelligence.

In the comment section of the live broadcast room, exclamations once again dominated the discussion. Many netizens were impressed by the outstanding performance of the O1 model on the mathematical benchmark test and expressed their amazement and praise.

“Too strong! OpenAI is indeed a leader in the field of AI.”

“95.4%! This is incredible!”

“The O1 model has surpassed humans!”

“OpenAI has changed the world again!”

However, those doubtful voices have not completely disappeared. Some netizens still insist that the O1 model has no obvious advantages over Panshi Technology’s Xiaotuanzi AI model. They believe that these mathematical benchmark tests cannot fully represent the true capabilities of AI models, but are more like a product of “examination-oriented education.”

“I think these tests, little Dumpling, can easily get high scores.”

“What’s the point of just looking at the test scores? We still have to look at the actual application.”

“Is OpenAI just entertaining itself?”

“It feels like the O1 model doesn’t offer any groundbreaking developments.”

Faced with these doubts, Ultraman decided to completely conquer everyone with a live demonstration.

“In order to let everyone understand the reasoning ability of the O1 model more intuitively,” Altman said, “we will now let the O1 model solve a high-level math problem on the spot.”

He looked at Mr. Lee.

MR Lee is operating.

At the same time, Su Yang said to Tuantuan: “Tuantuan, the show is about to begin.”

“receive.”

At the press conference, the technicians had already input a complex advanced math problem into the O1 model. This was a problem about Fourier transform, which required the use of multiple integration techniques and transformation formulas, and was quite difficult.

Ultraman stood aside with a confident smile on his face, waiting for the O1 model to give a perfect answer.

The O1 model begins to perform deductions.

However, something unexpected happened.

The first time, the O1 model got stuck halfway through the deduction, a bunch of garbled characters appeared on the screen, and then… it crashed.