Altman: I'd like to give special praise to two Poles. OpenAI has never encountered a problem they couldn't solve.

Altman specifically praised two Poles.

Without them, OpenAI wouldn't be what it is today.

They are Jakub Pachocki, the Chief Scientist at OpenAI, and Szymon Sidor with the title of "Technical Fellow".

Left: Jakub Pachocki, Right: Szymon Sidor

The two are not only fellow countrymen from Poland but also high - school classmates. They chose computer science and robotics respectively for their doctoral studies and later reunited at OpenAI.

Today, with ChatGPT sweeping the globe and serving hundreds of millions of users every day, Altman sighed that most people will never think of those who have put in great efforts behind it. These two Polish scientists are precisely the key figures.

Their contributions at OpenAI range from large - scale expansion of reinforcement learning in the Dota project to leading the pre - training of GPT - 4. They also jointly promoted the initial ideas that led to the inference breakthrough with Ilya and Lukasz.

Of course, there might be another reason for Altman's high evaluation of them:

During the internal turmoil at OpenAI in 2023, they were also the ones who took the lead in announcing their resignations and wanted to follow Altman out.

From High - School Classmates to Reunion at OpenAI

The story begins with a school in Poland, the Third High School in Gdynia.

There, the two learned computer science from the same teacher. The depth of the education they received far exceeded that of ordinary high - school courses, covering topics such as graph theory.

They first met at a programming summer camp. Every year, they had to undergo two months of intensive training at the camp.

However, later they recalled that their relationship in high school was not that close. They were just academic companions. It was their time in the United States after high school that helped them build a deep friendship.

Let's start with Pachocki (hereinafter referred to as Brother P).

At the age of 15, like many teenagers, he was still unsure about what he wanted to do in the future.

His father gave him a book, a Polish translation of the essay collection Hackers & Painters by Paul Graham, the co - founder of YC.

What touched Brother P the most was the description in the book: The common point between hackers and painters is that they are both creators.

Like composers, architects, and writers, hackers and painters strive to create beautiful things. They are not essentially conducting research, but it's even better if they discover some new technologies in the process of trying to create beautiful things.

Brother P was very lucky. He not only found his interest in computers but also soon discovered his real talent in this field: He won a silver medal at the International Olympiad in Informatics (IOI) during high school.

By the way, Altman's career path was also deeply influenced by Graham, but it was on the venture - capital side.

In the OpenAI podcast in August, Brother P said that thinking about the whole thing now was quite funny. At that time, he really didn't connect these things.

After graduating from high school, Brother P went to the University of Warsaw in Poland to study computer science. He won many competition awards during his undergraduate studies.

He won a gold medal at the ICM - ICPC in 2012 and participated in Google Code Jam. You can still find fragments of his acceptance speech at the ICPC today.

Middle: Jakub Pachocki

After graduating from his undergraduate program, he went to Carnegie Mellon University to pursue a Ph.D. in computer science.

Regarding AI, he originally thought that an AI capable of real reasoning would take a long time to develop, requiring more powerful computers and a very solid mathematical foundation.

But AlphaGo in 2016 changed everything.

The search space of Go is too large for our algorithms to handle. But they solved this problem with deep learning, which forced me to rethink.

After graduation, Brother P did a one - year postdoctoral research at Harvard University and then joined OpenAI in February 2017.

Now, let's look at Sidor (hereinafter referred to as Brother S). It was also AlphaGo that made him determined to study reinforcement learning. However, during his adolescence, the movie Iron Man inspired him the most.

After graduating from high school, he first went to the University of Cambridge in the UK for his undergraduate studies and then to MIT for his Ph.D.

It wasn't his active choice to go to MIT. He applied to many American schools at that time, and only MIT didn't reject him because MIT didn't require an English test, and his English was very poor (sigh).

At first, he chose robotics as his major. However, he soon became disappointed that real - world robots were not as cool as those in the movies and switched to studying deep learning and reinforcement learning.

When he finally graduated, his thesis topic was "Reinforcement Learning Methods for Multi - Stage Reasoning in Natural Language Processing". This topic would still be relevant as a research topic for the latest large models today.

At that time, Brother S's research objects were the LSTM model and the Deep Q - learning reinforcement learning algorithm. The multi - stage reasoning task referred to the problem of re - ordering scrambled sentences.

After graduating from his Ph.D., influenced by AlphaGo, he first applied for a position at DeepMind. However, during the interview, he was asked many theoretical machine - learning questions, and he couldn't answer any of them, so he failed.

Although OpenAI was still an unknown small company at that time, it seemed very serious about doing reinforcement learning, so Brother S joined.

In 2017, the two Poles reunited at OpenAI.

Shortly after joining OpenAI, they got involved in the Dota 2 project, aiming to make AI defeat human professional players in the complex e - sports game.

Early interview for the Dota project, Right: Szymon Sidor

They originally wanted to find out the limits of reinforcement learning through this project and at what point it would fail. As a result, they achieved great success, defeating the human professional player team and making history.

The cost was that both of them lost their hairlines.

In this project, Brother P focused on large - scale reinforcement learning and optimization, while Brother S participated in the development of a distributed training system and continuous training tools.

Their cooperation model gradually took shape in this project.

Brother P would walk around the office or his apartment, deeply thinking about how to study a phenomenon. Brother S, on the other hand, preferred to just start doing it and get some data first.

This combination of in - depth thinking and hands - on experimentation became their winning formula.

By the development stage of GPT - 4, Brother P had become the leader of the project. Altman once publicly said, "Without his contributions, we couldn't have achieved what we have today."

Brother S's role was very flexible. He positioned himself as an "independent contributor" and only occasionally took on leadership responsibilities. The core was to do the most valuable things.

True Colors Revealed in Times of Crisis

The internal turmoil crisis at OpenAI in November 2023 fully showed the importance of this pair.

At noon on November 17th, while having lunch, they received the news that Altman had been fired.

Brother S recalled that he was thinking about a problem in the corridor at that time. After receiving the news, he immediately went to find Brother P. Brother P was having a deep discussion with someone. Brother S rudely interrupted them and showed them the announcement.

Brother P's reaction was very decisive. He immediately walked out of the building and called Altman to ask what had happened. Altman on the other end of the phone was equally confused.

Later, they acted in unison. They took the lead in announcing their resignations with Aleksander Madry and wanted to follow Altman to join Microsoft. It was also their firm support that became one of the key factors for Altman's return.

This crisis taught them a profound lesson. Brother P sighed:

It was not until that moment that I really realized how important the governance structure was. What we had built for nearly a decade could suddenly face drastic changes.

Brother S also had the same feeling. "When these governance structures were initially established, it seemed like overkill." Now, the lesson he learned was that "decisions made in the early stage of a company, even if they seem insignificant at that time, may have far - reaching impacts in the future."

After things calmed down for a while, Brother P officially took over from Ilya as the Chief Scientist. His main responsibility was to formulate a research roadmap for the company and establish a long - term technological vision.