OpenAI scandal again! stealing 20 years of results, top mathematicians angry

  It's the first time in the history of the world.

OpenAI is actually igniting the wrath of the entire mathematical community!

Recently, Newtonian mathematician Tristan Buckmaster clashed sharply with OpenAI over the problem of attribution to the "NS equation".

A statement even went viral, and OpenAI phones pressured mathematicians to force the results.

I can't believe this wave is not even, and it's coming up again.

Recently, Andreas Thom, a mathematician at Dresden University of Industry, a leading authority on group theory, posted an email on social media titled "Iron Certificate".

He accused OpenAI of stealing the new generation of Astra, a field in which he spent 20 years researching non-sofic groups.

This was denied by OpenAI sharply.

In just a few hours, the news hit the charts.

Everyone is shocked once again, and some netizens say that it is best for mathematicians to go back to the era of blackboards, pens and paper.

The cause of the incident dates back to OpenAI's high-profile announcement in early August.

The top 10 "Fields Award" puzzles were all successfully won by Astra.

One of them is Gromov's soficity conjecture, which has plagued the field of group theory for 27 years.

Astra has found evidence for non-sofic groups.

When news broke, the entire AI community was cheering, "AGI has opened the door to pure mathematical discovery".

However, the top experts in the circle, the more they look at it, the stranger it seems.

As a Thai scholar who has worked in this field for more than two decades, Andreas Thom and his long-time collaborator Gabor Kun have keenly discovered:

Astra demonstrates the most central, the most sophisticated, key guiding steps in the process, which are highly aligned with the technological path they are taking.

What's even more ridiculous is that, in Tom's own words,

Kun and I did not approach the mainstream direction of non-smoothness groups, and the academic community at the time preferred other paths such as quantum games.

How exactly do the models of OpenAI accurately lock and master our extremely cold and subtle set of skills?

It was not long before the truth was revealed, and people were terrified.

In the months leading up to the official release of OpenAI's results, Thom and colleagues used ChatGPT to explore the "expanded matching problem" and tap into the dialog box with a wealth of unpublished core inferences and the latest proof details about the conjecture!

Twenty years of cold-blooded, months-long conversations with AI have turned the new OpenAI model into an "autonomous achievement" that shocks the world.

It's impossible for anyone else to keep their mouths shut.

After noticing the abnormality, Andreas Thom did not choose to get into trouble immediately.

He maintained academic restraint and sent a formal email to OpenAI's two core researchers, Mark Sellke and Sébastien Bubeck.

In the letter, Thom poses two extremely precise, logically rigorous questions:

1. Have the unpublished conversational records we discussed in ChatGPT over the past few months been incorporated into the model's training data?

2. Can these records be directly retrieved and accessed in the course of model solving and reasoning?

In response, researcher Mark Sellke returned an extremely short e-mail, with only one frozen sentence:

About your conversation with ChatGPT: That didn't happen.

With no data audit explanation, no privacy mechanism explained, or even any formatted set of words, such a barbaric, arrogant and iron-clad whole-scale rejection is possible.

Thom also disclosed a deadline of June 29 when he closed the model training option.

However, he believes that the shutdown of the switch can only be addressed afterwards, so that it is not possible to answer what earlier conversations have experienced, nor can it be stated whether derivative data has already been selected for the training process.

The user can see a setup without seeing the data flow in the background.

Thom later recalled that OpenAI's response was not only rough, but full of cunning, vicious colors:

He clearly distinguishes between "entry into training" and "online calling", while Sellke uses a vague "not happened" to try to eliminate all the substantive issues directly.

However, the price of arrogance comes very quickly.

Just a few weeks later, OpenAI was pushed to the brink of a millennia-old dilemma: the Navi-Stokes equation.

With Tristan Buckmaster and Levent Alpöge on the move, OpenAI's official public relations profile is suddenly reversed:

While insisting that there were no private data extracted from specific accounts, he acknowledged the fraud.

"It is impossible to exclude the use of de-identified data from customer products to improve the model".

The rotator cuffed his throat with precision.

Earlier, Tom also vowed that "there is no such thing" and immediately acknowledged that identifying the data was taken to improve the model.

As long as the name of the mathematician is erased, the disruptive intelligence of humanity that has not yet been publicly disclosed, can the empire be "identified as whitewashed" on the servers of Silicon Valley giants and transformed into AI's autonomous intelligence?

In the latest excerpt, Andreas Thom ruthlessly tears this text game apart:

Identification may remove a scholar's name, but it does not remove the intellectual core of a mathematical thought.

From Tristan Buckmaster to Andreas Thom who stands out today.

The first can be said to be a "coincidence" and the second can be interpreted as a "misconception", and the next two or three top scholars have collectively challenged and accused OpenAI of completely destroying the well-woven "AGI science myth".

As Valerio Capraro calls for in his long post, this is no longer a series of isolated academic disputes:

If this series of allegations are true, what we are witnessing is one of the largest intellectual predation scandals in the history of science.

So next time, who else would dare to open this dialogue box that goes to Shenzhen?

Today, OpenAI owes a real shift to the global mathematical community.