The real trump card of OpenAI is not Ilya. Just now, Altman rarely thanked these two people.

In the public eye, the story of OpenAI boils down to just a few well - known names. However, it's often those who stay out of the spotlight that make the real miracles happen. Jakub Pachocki and Szymon Sidor, a pair who graduated from a high school in Poland, have been turning the seemingly impossible into reality over the past decade with their seamless teamwork. From the Dota AI to GPT - 4 and then to the breakthroughs in reasoning, their influence has permeated almost every core aspect of OpenAI.

Just recently, Sam Altman, in a rare move, publicly thanked two individuals in his blog: Jakub Pachocki and Szymon Sidor.

He wrote, "OpenAI has never encountered a problem that they can't solve."

These two individuals have rarely appeared in public, yet they are the key driving forces behind the development of the Dota AI, GPT - 4, and the breakthroughs in reasoning.

Altman's brief remarks actually imply that the miracles at OpenAI are not just about the company's success at the corporate level. They are also the result of the perseverance of certain key individuals.

From Dota to GPT - 4: How Were the "Impossibles" Achieved?

Looking back at OpenAI's development history, there have been several moments that shook the entire industry.

In 2019, the AI defeated the world champions in Dota 2. A few years later, GPT - 4 made people feel for the first time that large - scale models might truly possess "reasoning abilities".

These achievements seemed like "mission impossible" at the time.

Back then, the industry generally believed that it was difficult to scale up reinforcement learning, and complex game - playing scenarios were out of reach.

However, OpenAI chose to face these challenges head - on. Eventually, through large - scale self - play, OpenAI Five broke the limits of technology and computing power, proving that reinforcement learning could reach the world - class level in complex environments.

In the Dota 2 reinforcement learning paper published by OpenAI in 2019, the names of Jakub Pachocki and Szymon Sidor can be found in the list of authors.

What supported this victory was a distributed training system called Rapid.

Jakub Pachocki and Szymon Sidor were the key drivers of this system. They translated the concept of "large - scale training" into engineering practice, which led to the success of OpenAI Five and laid the foundation for the subsequent training of GPT models.

For Jakub, this journey was also an experiment in scientific thinking. In an exclusive interview with Time magazine, he said bluntly:

Although it may seem like just mathematics, it's actually more like natural science. We're trying to understand this phenomenon.

Today, you can have a conversation with the model, but it's just an assistant that needs continuous guidance. I expect this to be the main change in the future.

These words not only break the cold - blooded image of technology but also reveal his ultimate vision for future AI: to make AI a "continuously thinking scientist" capable of actively solving problems and promoting scientific discoveries.

From Dota to GPT - 4, this path was not achieved through gimmicks. What truly supported it was patience in the underlying architecture and engineering.

Jakub and Szymon are not the center of attention on the stage. They are the unsung heroes who make miracles happen.

Jakub's Ascent: From a Competition Genius to the Chief Scientist

Jakub Pachocki's journey has almost been in sync with OpenAI's development.

As the company has been challenging the impossible, he has also been making breakthroughs, leaving his mark on the key milestones of AI development.

Born in Gdansk, Poland, Jakub was a programming prodigy from a young age.

During his high - school years, he dominated top - tier competitions such as the IOI, ACM - ICPC, and Google Code Jam. In 2012, he even won the global championship of Google Code Jam.

In the eyes of the Polish media, he is "one of the smartest minds to come out of the country".

Later, he continued his studies at the University of Warsaw and then went to Carnegie Mellon University in the United States to pursue a doctorate, focusing on theoretical computer science and algorithm optimization.

Beyond his academic pursuits, he has also conducted research at institutions such as Harvard and Berkeley.

In 2017, Jakub joined OpenAI and quickly became a cornerstone of the company.

He led the algorithm research and infrastructure construction for OpenAI Five, making large - scale reinforcement learning training a reality. This not only contributed to the success of OpenAI Five but also laid the groundwork for the subsequent training of GPT models.

For Jakub, this journey was also a scientific experiment. In an interview with Time magazine, he said:

Although it seems like just mathematics, it's actually more like natural science. We're trying to understand this phenomenon.

Today, you can talk to the model, but it's still just an assistant that needs guidance.

I think the main change in the future will be to enable AI to work independently for hours or even days.

From a young genius on the competition stage to the chief scientist leading the world's most cutting - edge research, Jakub's trajectory has been seamless.

He is not only a technical wizard but also a person who constantly challenges the impossible.

Szymon's Low - Key Presence: An Indefatigable Doer

If Jakub is the navigator setting the direction, then Szymon is the engineer who stabilizes the ship and ensures a smooth journey.

Like Jakub, Szymon is from Poland. They were classmates in high school, participated in programming competitions together, and developed a deep understanding of each other at that time.

Later, Szymon continued his studies at the University of Cambridge and the Massachusetts Institute of Technology, focusing on machine learning and computer science.

After graduation, he worked on engineering R & D at companies such as Google and Dropbox, accumulating rich experience in large - scale system debugging.

In 2016, Szymon joined OpenAI as one of the earliest research engineers.

Szymon has always kept a low profile, but his contributions are everywhere:

From the distributed training of reinforcement learning to the pre - training of GPT - 4 and the iteration of the latest reasoning models, his hard - working figure can be seen.

In a podcast, when asked if OpenAI had the world's top debugging talents, Szymon just smiled and replied, "I won't mention specific names."

Screenshot from a BI report: The "No names" interview segment reveals the AI talent war.

This attitude is not only a sign of humility but also reflects the practical nature of an engineer.

Inside OpenAI, he is often the last one to leave the laboratory.

Jakub once said that without Szymon's engineering implementation, many of his ideas might have remained on paper.

In Altman's thank - you post, Szymon's name finally appeared side by side with Jakub's.

He doesn't have a flashy title and doesn't like to speak in public, but his influence can be found behind every "miracle" at OpenAI.

The Golden Duo: A Dynamic Pair in Scientific Research History

In the history of technology, there have always been some duos that are frequently mentioned: Watson and Crick discovered the double - helix structure of DNA, and Brin and Page founded Google.

For OpenAI, the names of Jakub Pachocki and Szymon Sidor are quietly being added to this list.

The two were classmates in high school, studying algorithm problems together. Later, they both entered the academic circle in the United States and finally reunited at OpenAI.

Szymon is on the left, and Jakub is on the right.

This more than - decade - long partnership has allowed them to complement each other naturally in scientific research and engineering.

Jakub focuses on strategy and theory. He is the core scientist driving the large - scale application of reinforcement learning, the pre - training of GPT - 4, and the research on reasoning models.

On the other hand, Szymon is more focused on engineering implementation, playing a key role in distributed training systems, debugging, and optimization.

Szymon was learning about machine - related content during his vacation in 2018.

As a result, many seemingly far - fetched research ideas can be put into practice in a short time.

Altman wrote in his blog:

OpenAI has never encountered a problem that they can't solve.

This may sound like an over - exaggeration from the CEO, but in the context of their cooperation history, it's not an overstatement.

Whether it's the Dota AI or the reasoning breakthrough in GPT - 4, this duo has been involved.

Jakub once described Szymon as "indefatigable", a rare but accurate word.

He added that so far, no problem at OpenAI has been able to stump this pair.

The public may be more familiar with names like Altman and Ilya, but inside OpenAI, this Polish duo is the golden combination that supports the company's scientific research.

In the public eye, OpenAI's story is often simplified to a few well - known names: Sam Altman, Ilya Sutskever, Greg Brockman... However, Altman himself pointed out a different answer in his blog.

In Altman's view, this duo is the most reliable engineering team in the company.

This is not just a public - relations statement but a genuine admiration for Jakub Pachocki and Szymon Sidor.

Perhaps one day in the future, when AI can truly become a "continuously thinking scientist" as Jakub envisioned.

People will look back and realize that this history is not just about the competition of algorithms and computing power. It's about how two young men from a high school in Poland turned the impossible into reality time and time again.

This is the story that Altman wants the public to see.

References

https://blog.samaltman.com/jakub-and-szymon

https://eu.36kr.com/en/p/3427480740285831?utm_source=chatgpt.com

This article is from the WeChat official account "New Intelligence Yuan". Author: Qingqing. Published by 36Kr with permission.