In the tide of innovation and technological advancement, there is a group known as "unconventional geniuses" who are changing the world in their unique ways. Recently, at an early morning press conference, OpenAI showcased its new generation of large AI model GPT-4o, declaring its powerful technical strength to the industry and leaving attendees astonished.

It is noteworthy that for the first time, the release was led by Chief Technology Officer Mira Murati instead of the company's standout, Altman. Mira is not only an eminent figure in the technical field; her academic background includes studying mechanical engineering at Dartmouth College and her experience as a senior engineer at the French aerospace company Zodiac Aerospace. Later, she joined Tesla, achieving leadership success in the development of the Model X electric motor system. Mira's strong interest in autonomous driving and artificial intelligence became apparent after this, which eventually led her to join OpenAI.

In the fierce competition of artificial intelligence, the secret to success can be summarized in three points: talent is primary, followed by data, and finally computing power. Behind OpenAI's success is a group of technical elites like Mira. Such a strong talent pool has become the key factor for the company's competitive advantage.

Within OpenAI's talent pool, in addition to traditional technical elites, there are many "unconventional geniuses." For example, Prafulla Dhariwal, who is responsible for the multimodal aspects of GPT-4o, only has an undergraduate degree. Even the educational background of Christopher Olah, once the technical director of OpenAI, is quite unique as his LinkedIn profile clearly indicates "Real Life Experience University."

What's more surprising is that in the author list of OpenAI's breakthrough project papers, one finds a 21-year-old young researcher with traces of acne still visible, who only has a high school diploma and began self-study in programming at 17. This shows that OpenAI has its own unique insights and criteria for evaluating scientific research talent.p>

A close look at OpenAI's talent recruitment strategy reveals two obvious characteristics: it does not focus on educational background. In other research institutions, having a PhD seems to be a basic requirement for entry, but at OpenAI, these traditional views do not hold. For example, Aditya Ramesh, who developed the well-known image generation model DALL·E, only has a bachelor's degree from New York University. Although he considered further study, OpenAI's urgent invitation led him to join the company earlier. And the case of previously mentioned young researcher Will DePue is even more special; he turned to data analytics entrepreneurship in high school instead of the traditional academic path.

After corporate acquisitions, he immersed himself in a radical civic group in the United States, contributing seven months as a volunteer. With such a background, when Will DePue applied to join OpenAI, he did not have an impressive resume to showcase. Earlier in February of last year, an Intelligence Research survey showed that within the ChatGPT project team, the number of members with undergraduate, master's, and doctoral degrees was almost equal, accounting for 33%, 30%, and 37% respectively.

OpenAI, when hiring, not only disregards educational background but also work experience. This organization is very willing to let newcomers take on significant responsibilities. Aditya Ramesh, who became a full-time employee not long ago, was involved in the research of DALL·E. Now, this young Indian with only six or seven years of work experience has made his mark on papers for DALL·E 2, DALL·E 3, GPT-4, and Sora. This fact has excited the Indian media, which proclaim that DALL·E has Indian lineage. An even more apparent example is the Sora project team. The project's lead researcher, Bill Peebles, though a recent Ph.D. graduate, led the entire project.

Of course, OpenAI also has traditional elites, such as another leader of the Sora project, Tim Brooks, young but experienced. His mentor, Alyosha Efros, is an authority in the field of computer vision, and Brooks himself has a background in artificial intelligence research at major companies like Google and NVIDIA. At OpenAI, these elite talent often collaborate with young and exceptionally gifted tech prodigies in scientific research. To a certain extent, the working and living experience that OpenAI offers undoubtedly fits with the romantic notions people have about tech geniuses.

However, real life is not as simple and straightforward as inspirational movies. The word "genius" is not evidently written on a candidate's face. With a "looking at neither" strategy, how does OpenAI recruit the right talent? In fact, OpenAI is known for its rigorous hiring practices.

In 2017, someone shared their interview experience with OpenAI on Reddit: First, they went through an initial selection process, then they experienced four rounds of interviews, including a presentation, two research interviews focusing on different aspects—one on technical knowledge reserves, the other on philosophical discussion with expectations to share insights into the progress of AI technology—and a programming interview. On the hiring website Glassdoor, nearly half of the reviewers gave negative feedback on their application experience, citing OpenAI's interview process as lengthy and full of difficult challenges.

At the beginning of last year, Diane Yoon, the VP of Human Resources at OpenAI, publicly explained why they set up such an interview process: OpenAI values an applicant's "problem-solving ability" more. She noted that although OpenAI is a research institution, it does not adhere strictly to written knowledge. OpenAI believes that the purpose of research is to solve actual problems, encourages researchers to try the simplest solutions instead of blindly pursuing academic innovation, as often the former is more effective.

OpenAI has inherited and developed achievements along the technical path set by Google. Among them is Sora, which technically isn't a revolutionary new invention but an improvement and enhancement of existing technology. Even so, last year's foundational paper "Scalable diffusion models with transformers" was rejected by the top AI academic conference CVPR for being "lacking in innovation." The recently introduced GPT-4 also continues this tradition; while OpenAI has not brought innovation in academia, it has made the science fiction dream of AI communicating freely a reality through outstanding engineering capabilities.

However, this problem-solving ability is not directly reflected on a personal resume. Therefore, OpenAI has adopted screening methods with high uniqueness. For instance, mentioned that she would ask job candidates to showcase their "influential work" from the past to evaluate whether they possess the ability to solve problems and drive innovation. This concept is similar to other tech companies in search of unique talent. Inspired by Steve Jobs, Apple, in its early days, was known for its unusual interview questions and methods, such as Jobs trying to determine if candidates had a sense of humor and a rebellious spirit to filter for talent with 'crazy' creativity.

Despite the constant innovation in talent selection strategies, OpenAI continues to focus on gathering technical elites. Take Aditya Ramesh as an example, he was once a student of Yann LeCun, and conducted research at Meta, but ultimately chose to join OpenAI. For those eager to engage in scientific research, OpenAI undoubtedly has a strong allure.

OpenAI attracts talents partly because of its Oppenheimer-style idealism. Just as Oppenheimer was involved in the birth of nuclear weapons and strongly opposed their misuse, OpenAI's founder Samantha Altman also hopes to prevent the misuse of artificial intelligence from harming humanity. Hence, they established a nonprofit laboratory not controlled by large tech companies, aiming to provide a counterbalance in the field of artificial intelligence. Altman believes that putting such a crazy and unrealistic idea into practice is not hard, "because people will think it's cool and are willing to help." In fact, since its establishment at the end of 2015, although OpenAI started late, it has successfully attracted top talents including Ilya with its dreamlike company philosophy. Even after Google offered Ilya an annual salary of two million dollars, after careful consideration, he still believed that "saving humanity" was more important. In addition, OpenAI's Charter published in 2018 further emphasized their mission—"to ensure that general artificial intelligence benefits all of humanity".

With the release of GPT-4, Altman also took this opportunity to reiterate his grand vision. Carrying the flag of Oppenheimer-style idealism, they have attracted many professionals passionate about technology. For the general public, more personal needs such as raises and paying off mortgages may be more pressing. However, these elite members of the tech community are convinced that they are participating in a cause that will have a profound impact on the future of humanity.

OpenAI's romantic ideology is just a beautiful shell; what truly enables the ideal to take root and flourish is its solid foundation. Among the senior researchers at OpenAI, there is a Chinese researcher named Li Jing. He completed his undergraduate studies at Peking University, then moved on to the Massachusetts Institute of Technology for further education and obtained a Ph.D. During his postdoc at Meta, he was fortunate to be a student of Turing Award winner Yann LeCun. After working at Meta for nearly three years, he then joined OpenAI. In fact, every step of Li Jing's career has cleverly paved the best path for his professional journey.

America seems to have built a massive "talent hematopoietic machine." This machine provides a complete service chain from academia to the industry. At top-notch institutions like Berkeley and MIT, there are distinguished scholars such as Alyosha Efros and Kai-Fu Lee who often have close connections with the industry and frequently provide employment pathways. For example, Yann LeCun from New York University is a typical example, with many of his students having held positions at Meta. The top academic conferences in the field of artificial intelligence, such as CVPR and ICCV, are mainly held by IEEE in New York.

When young computer professionals are ready to expand and advance in the field of scientific research, they often find that the United States has already established many brilliant milestones on this path. When they are no longer willing to explore alone in the wilderness, they naturally turn toward these paved golden avenues. Therefore, we have no reason to overly blame those researchers who choose to develop their careers in the United States.

The strong industrial foundation of the United States attracts numerous computer professionals from around the world. According to statistics from the think tank MacroPolo, as of 2022, 57% of top AI researchers work in the United States, while China, ranked second, accounts for only 12%. However, China has already made significant progress—in 2019, China was merely categorized under "others" in the statistics. If classified by nationality, actually only 28% of top AI researchers are Americans. A large number of Chinese, Indians, and Europeans gather in the United States, becoming an important part of the country's strength in artificial intelligence.

Finally, the idealistic story of OpenAI may not have much enlightenment for countries outside the United States.

As 2020 drew to a close, OpenAI's then CTO Christopher Olah shared an attention-grabbing blog on the internet. Olah, with a reputation in the computer science community, despite having only a high school education, carved out his own career path through his own efforts and careful guidance from industry bigwigs. In his blog "Do I Need a College Degree?", Olah revealed his unique approach to learning artificial intelligence: actively auditing college professors' courses, participating in academic conferences, visiting labs, and more—resources that are open to everyone in the United States.

During his studies, Olah was fortunate to meet a quantum physicist whose guidance helped him complete his first academic paper. Afterwards, the "Godfather of Deep Learning" Yosh iua Bengio spoke highly of his talent and was even keen to attract him to university studies. Meanwhile, Olah also received a two-year financial grant totaling $100,000. This grant came from the Thiel Fellowship, created by Peter Thiel, aimed at supporting young prodigies who choose to drop out of school and aspire to devote themselves to the tech field. We can say that Olah's success is attributed to his relentless personal effort as well as that open and inclusive environment that nurtures talent. It is such an environment that truly fosters the innovative vitality of enterprises and the whole economy.