News center > News > Headlines > Context
Starting from the OpenAI core team, more than 600 people and the AI ​​landscape they shape
Editor
2024-12-05 12:03 5,308

Starting from the OpenAI core team, more than 600 people and the AI ​​landscape they shape

Image source: Generated by Unbounded AI

Of all the Internet products born so far, only two have been used by 250 million people per week within two years of their launch, one is Douyin and the other is ChatGPT. The difference is that ChatGPT did not have addictive dances at the beginning, and basically did not buy volume promotion.

As of November this year, it will be two years since the ChatGPT product came out, bringing OpenAI 3.7 billion US dollars in revenue, far behind the levels of Google and Facebook in the same period, and higher than the situation when Pinduoduo went public, 1570 The company's valuation of US$100 million also broke a historical record.

The not-so-sexy side of numbers is that OpenAI’s annual revenue of 3.7 billion US dollars is based on nearly 9 billion US dollars in expenditures - you spend 2.4 yuan and get 1 yuan back. This is a WeWork and SenseTime-level operation. efficiency.

OpenAI hopes to use aggressive growth to solve profitability problems. According to The Information, OpenAI expects to spend $40 billion in revenue in exchange for $26 billion in revenue by 2026 - equivalent to nearly 1.4 times Baidu's revenue; by 2029, it will use $100 billion in spending to achieve profitability.

From the Roosevelt Room of the White House to the conference rooms of TSMC, Sam Altman describes ambitious goals that require unlimited investment to achieve. He plans to spend US$100 billion to build a data center, install 2 million AI chips, and consume 5 gigawatts of electricity - equivalent to the annual electricity consumption of more than 4 million American households; if he were to manage TSMC, he would have to invest US$7 trillion. Build 36 wafer fabs to produce AI chips. The numbers were so crazy that one TSMC executive mocked his proposal as “podcasting bro”-level: paper-based and unrealistic.

Within OpenAI, radical growth is already under pressure: Anthropic, founded by former employees, is carving up market share with the strong support of Amazon; more and more developers are discovering that problems can be solved without the use of the most advanced large models. Text summary and other issues; and the improvement of model capabilities may also encounter bottlenecks...

It cannot be concluded that Altman is a utopian, at least ChatGPT In the first two years of his career, he relied on his usual extraordinary ambition: in the consumer Internet market where several trillion-dollar super giants controlled the infrastructure, a semi-profit research organization could carve cracks in the giant walls. As for what made OpenAI so successful despite the lack of capital at the time, its co-founder and president Greg Brockman once explained: "(Talent) is OpenAI's secret weapon." p>

However, as the company changes fromFrom non-profit to profit, OpenAI solves capital problems but also triggers people problems - organizational turmoil and brain drain have almost never stopped in the past two years. OpenAI, which has lost its secret weapon, has placed itself in a turbulent and ever-changing market segment, where the key factor of change is precisely people.

On the second anniversary of the release of ChatGPT, we study the changes in the world’s largest AI startup company from the perspective of people. The flow of these people not only shapes OpenAI, but also leads the global AI industry.

We had hoped to list ChatGPT as a co-author, but after two years and three generations of model updates, we still continue to feel the limitations of large models.

The list of “Contributions” attached to the technical reports of the GPT-4, GPT-4o, and o1 models is the starting point for our research. To make the analysis more complete, we also included authors of GPT, GPT-2, and GPT-3 papers to provide a comparative perspective, a total of 631 people.

We use "name + OpenAI" as keywords to generate LinkedIn search links, and manually find valid LinkedIn homepages one by one - large models cannot help us complete this tedious task. Using Proxycurl's API, we captured the work and education experiences of contributors to the LinkedIn homepage, used R language to clean the data, let the large model classify functions according to job descriptions, and then manually proofread it.

There are no contributors to the LinkedIn homepage, so we searched again using a large model, but the information was incomplete or outdated, and finally used the original and reliable method: manually finding media reports and social network pages on search engines, supplementing Data for about 10 people. We also used the open source Python library scholarly to add the Google Scholar data of 214 of them, evaluate their academic influence, and finally built a database of 583 contributors.

GPT iteration, OpenAI change

In December 2015, OpenAI Established as a non-profit AI laboratory, relying on the belief of "developing safe AGI that benefits all mankind" and the reputation of Elon Musk to steal people from giant Internet companies such as Google, it established a team in more than a year. Support 50 A research team composed of several technical elites: in addition to the founder Altman, there is also Ilya Sutskever, one of the founders of the AI ​​wave and the head of Google Brain, and the CTO of Stripe, an emerging unicorn in Silicon Valley. Greg B.Rockman), the new generation of AI research genius Andrej Karpathy and other researchers.

In the early days of its establishment, this team built a robot hand that can play Rubik's Cube with one hand, studied AI for playing Dota, and finally saw the possibility of AI changing human life in the direction of language understanding of the world. In 2019, they bet on big language models.

From GPT-1 to GPT-2 and GPT-3, the model parameters have expanded from 100 million to 175 billion, the cost investment has increased, and the number of contributors has increased from 4 to 30. In order to raise resources , the company had to transform into a non-profit.

Of the more than 30 researchers who laid the foundation for ChatGPT, 2/3 have left OpenAI. As of October, 80 of the 583 contributors no longer work at OpenAI, including more than 20 founding team members and mid-level and senior management.

“The most compelling artificial intelligence systems, in addition to requiring algorithm innovation, also consume the most computing resources,” OpenAI once wrote in an article. “In the next few years, we will invest in data Billions of dollars to acquire cloud computing resources and attract and retain talented people." Paradoxically, every time the company successfully raises more resources, it triggers a wave of contributors leaving. Dario Amodei, former OpenAI chief security researcher, has spoken at the heart of this conflict, concerned about “commercial interests interfering with security priorities.”

The first batch of contributors left in early 2020. At that time, Altman introduced a $1 billion investment from Microsoft to promote OpenAI to form a for-profit entity. Amoudi and a group of colleagues left to start the AI ​​company Anthropic. That time, OpenAI lost 8 of the 30 GPT-3 paper authors.

Altman believes in the concept that "economic growth can make democracy more universal." The pursuit of growth is the instinct of this former YC entrepreneurial mentor in running a company. ChatGPT was launched on November 29, 2022, OpenAI raised tens of billions of dollars from Microsoft, and held a press conference in November 2023 that looked like early Apple. Altman seemed to have become the star of a new generation of technology companies.

However, less than a month later, the contradiction between the pursuit of growth and business scale and the AGI belief in benefiting mankind completely broke out within the company, and shocked the world in a super dramatic way - the then chief scientist's announcement The board of directors of the Sutskvi Union suddenly announced the expulsion of Altmann.

"He has repeatedly provided false information about security measures. The board of directors is basically unable to understand the operating effectiveness of security measures and cannot judge which aspects may need improvement." Participated in the actionsaid Helen Toner, Director of OpenAI.

In just five days, this revolution that shocked the world came to an end, and Altman won. He returned to the company and a majority of the board resigned. Six months later, Sutskvi left OpenAI, once again triggering the departure of a group of contributors who were concerned about security.

Altman remains determined to move forward with the next step: turning OpenAI into a for-profit company.

The new OpenAI comes from big manufacturers

The old and new alternate , OpenAI’s way of attracting talents has also changed: using generous salaries to compete with Google, Meta, etc. for talents. At the beginning of 2023, OpenAI has offered an annual salary of US$800,000 to fresh doctoral graduates. The annual salary of senior researchers can be several times higher, comparable to Google and Meta, which have tens of billions of dollars in annual profits.

Money has become the driving force for controlling talents, and it is also a time bomb that cannot retain people. As soon as the company stops growing, compensation becomes a burden and options lose their appeal. The cost of choosing the business route became clear. Now, the pursuit of growth is not only Altman’s pursuit, but also the foundation for sustaining OpenAI’s operations.

The newcomers will not only reshape the company's team, but also reshape the ChatGPT product.

Only Sutskevi was fully involved in the development of GPT-1 to GPT-o1. His departure leaves only Alec Radford with similar experience. Radford has participated in the research and development of GPT-1 to GPT-4o. He joined OpenAI in 2016 and was one of the first researchers to apply the Transformer architecture proposed by Google to large-scale language training. He also established the current large language model. One of the people in training mode.

“Only Alec Radford’s departure will determine the collapse of OpenAI.” Emad Mostaque, founder of the large modeling company Stability AI, once said on social media Said above.

The rapidly accelerating flow of talent in the past two years has pushed OpenAI to quickly adapt to the stage that all startups must go through - a company that rises because important people create products will inevitably let individuals no longer have the power to make decisions in its development. The role of product life and death.

OpenAI’s new models are still being steadily introduced to the market. There are 390 core contributors to the two generations of GPT-4o and o1 models, and more than 70% joined the company after the release of ChatGPT. They have shaped the new OpenAI. In fact, high salary does solve the problemThe problem of the loss of elders has been solved. The three months after the struggle in November 2023 will be the peak period for contributors to the two generations of GPT-4o and o1 models.

OpenAI’s recruitment strategy has not completely continued from when it was first established The proposal given by Sk: Focus on recruiting students who have not yet graduated but are very smart. “It is best to let them join before they achieve breakthrough results.” The contributors who joined ChatGPT after its release were only more than 10 fresh graduates.

The new contributors mainly come from large companies, and 54% have worked in the seven largest technology companies in the United States.

In its early days, OpenAI attracted a group of employees from major companies to join it based on their faith. But between 2017 and 2020, OpenAI could hardly maintain operations, and not many employees from major companies were willing to join. These years have also been a low period for OpenAI’s recruitment.

By 2023, the probability of OpenAI becoming the next major technology company looks higher, and its attractiveness to employees of major companies has greatly increased. More than 40% of contributors with working experience in large factories joined this year, far exceeding previous levels.

Google and Meta, two companies that have deep accumulation in the field of AI but are slow to move in the field of large models, have sent the largest number of contributors to OpenAI, with 119 and 103 people respectively. Only these two companies Family accounts for nearly 40% of the total.

They also contributed the technical backbone to OpenAI. There are a total of 118 leads in the three generations of GPT-4, GPT-4o, and o1 models, and 45% of them have worked at Google and Meta.

Following this are the YC companies, with Stripe ranking third and Dropbox ranking fifth. Together with Airbnb, Instacart, Twitch, Cruise and other companies, a total of 59 people were transported. These companies share values ​​similar to Altman's: growth above all else.

Altmann once said on social media that he is very interested in a university alternative: "Find the smartest and most enterprising 18-year-olds in the world and give them 10+ years of Salary and resources, allowing them to work on any project they want, paired with smart peers - in exchange for a few percentage points of their future income."

At OpenAI, his preferences are mapped to employees'. on educational background. Among the contributors to OpenAI, we found the schools from which 543 of them graduated. Most of them came from top universities, with graduates from Stanford University, California Berkeley, and MIT accounting for 30%. Graduates from Chinese universities also have a strong sense of presence. Two universities in northern Qing DynastyThere are more than 20 graduates from the school.

High-level purge, Altman’s team

2017 , OpenAI, which was just founded two years ago, has discussed whether to transform into a for-profit organization. Musk and Altman are vying to be CEO. Suzkowe and Brockman, who are responsible for daily work, resist Musk, but they also doubt Altman’s motives. .

In an email, they threw the question directly to Altman: "We cannot fully trust your judgment... We do not understand why the title of CEO is so important to you... ....Is AGI really your main goal? What is its connection with your political ambitions?"

Musk quit in anger, but Altman relied on his "confrontation with Africa" For-profit organizations are still enthusiastic." His statement won trust and expanded his influence in OpenAI.

The subsequent development of OpenAI proved that Suzkowe and his colleagues were not wrong. What Altman calls "nonprofit enthusiasm" lasted only two years before beginning to fade.

Since becoming the CEO of OpenAI as he wished, Altman has been free to transform OpenAI according to his own vision.

In November 2020, half a year after the GPT-3 model paper was released, Altman appointed OpenAI’s first vice president of product, Peter Welinder, to be responsible for making the GPT-3 model a API products are sold externally. Soon after, Altman hired Anna Makanju, a former senior policy adviser to the Vice President of the United States, from Facebook to be responsible for policy and public relations. Later, Macanjiu used his experience in foreign affairs to arrange a trip for Altmann to travel around the world and meet with dignitaries from various countries.

This round of high-level changes is relatively mild. The bigger knock-on effect is that Dario Amodei, the vice president of research and responsible for security research, left to found Anthropic, but the core team is still stable.

After the release of ChatGPT, OpenAI received a US$10 billion investment from Microsoft. Altman once again elevated the status of product growth and commercialization, and hired three vice presidents to take charge of related matters within half a year:

Srinivas Narayanan, who joined in April 2023, is responsible for the engineering work of ChatGPT and API construction of various generations of models. He previously worked on similar tasks at Facebook. Peter X. Deng, who joined in May 2023, is responsible for the products of ChatGPT and ChatGPT Enterprise Edition. He was previously on Facebookk worked as a product manager for more than 10 years, and later became the product leader at the database company Airtable. Krithika Muthukumar, who joined in August 2023, is responsible for OpenAI product GTM work. Previously, she worked in marketing for more than 10 years at YC companies such as Dropbox and Stripe.

The original vice president of product Peter Verinder was transferred to be responsible for new product development, and the number of OpenAI's product executives caught up with the technology executives.

The reason why it has not been exceeded is not because of restraint, but because of external constraints. The research team headed by chief scientist Sutskvi has a religious fervor for “safe AGI.” A report in "The Atlantic Monthly" wrote: Suzkevi will lead employees in offline gatherings to shout "Feel the AGI! Feel the AGI!". He is also the organizer of OpenAI's "Super Alignment Project" and pursues Safety.

Sutskvi’s subsequent expulsion failed, giving Altman the opportunity to expand his influence and take full control of OpenAI. In 2024, OpenAI's senior management entered a period of large-scale purges:

In February, founding team member Andrei Kapasi returned to OpenAI and left again. In April, Vice President of Human Resources Diane Yoon resigned. In May, Sutskvi resigned. In August, Vice President of Product Deng Xiuping, who had just joined a year ago, left. In September, co-founder John Schulman, CTO Mira Murati, VP of Research Bob McGrew, VP of Research Barrett Zo Husband (Barret Zoph) resigned. In November, Vice President of Research and Safety Lilian Weng resigned; President Brockman returned to the company after a three-month leave and will be transferred to a new position.

At the same time, Altman hired a chief product officer, chief financial officer, and chief human resources officer from outside, and promoted loyal subordinates to key positions. For example, Mark Chen, a key figure who called on employees to express their stance during the expulsion crisis, "If Altman is not allowed to return, he will resign and join Microsoft." Mark Chen was promoted two levels in a row, from director of research to senior vice president of research.

Looking back at the comings and goings in the past two years, as Altman said in an interview in November, "OpenAI has quickly gone through the process. It usually takes ten or even more years for medium and large technology companies to go away." complete development process.” Today, it is an unquestionably commercial company, still expanding its product, operations, and design teams to plan the company's next star product.

OpenAI regards Agent as the most important issue next year. Currently, at leastThree related products are under development, and one is expected to be launched as early as 2 months next year.

However, the competitive pressure of Agent is greater than that of large models. Its logic is to let the large model act as a scheduler, understand user needs, and automatically schedule databases and tools to complete complex tasks. Agent can give full play to the capabilities of large models, but it relies on the software ecosystem controlled by large companies such as Google, Meta, and Microsoft. In this direction, OpenAI is already half a beat behind Anthropic, which demonstrated related tools in October and is promoting standards for large model calling and data transmission.

The weirdness of fate appears here: Anthropic is an industry newcomer created by contributors who left OpenAI.

AI company's "Whampoa Military Academy", the former OpenAI cultivated opponents for today's self

< /p>

Two years after ChatGPT was released, OpenAI’s valuation jumped from less than 20 billion US dollars to 157 billion US dollars, becoming the world’s third largest super unicorn. It is preceded by 12-year-old ByteDance and established 22 years of SpaceX.

“It’s easy to criticize Sam and OpenAI, but remember, ChatGPT is only a little over 20 months old.” Brad Guest, a well-known Silicon Valley venture capitalist who participated in OpenAI’s latest round of financing Brad Gerstner said, "This is an unprecedented rocket."

In 2022, the Federal Reserve ended its interest rate cutting strategy during the epidemic era, and the enthusiasm for venture capital cooled. According to market research firm CB Insight, global AI investment fell to its lowest level in two years in the third quarter of this year.

The emergence of ChatGPT near the end of the year has sustained global AI investment. In two years, $132 billion has poured into AI companies, creating an average of six unicorns every quarter.

Although Altman left the startup incubator YC for OpenAI. But looking back on the past two years, he has turned OpenAI into an entrepreneurial incubator in an unexpected way. It’s just that the more successful the incubator becomes, the more difficult Altman’s situation becomes—almost all those incubated by OpenAI become his opponents.

69 contributors who have resigned from OpenAI and have a clear destination have built an AI entrepreneurial ecosystem, even an ecosystem that competes with OpenAI - whether it is a big technology company or an emerging technology wave, this phenomenon is Extremely rare.

More than half of people have started a new business or joined an AI startup, and Anthropic is the main gathering place.In addition, there are large model companies such as Safe Superintelligence, Cohere, and AI21 Lab; large model application companies such as Character AI and Adept AI; and AI robot companies such as The Bot Company and Light Robotics.

Some of the remaining contributors have gone to the AI ​​departments of major companies such as Google, xAI, and ByteDance, and have gone to universities to continue studying AI. There is also a group of people who, after OpenAI’s valuation skyrocketed, went directly to VC funds such as Sequoia and a16z to invest in AI and support OpenAI’s opponents in batches.

Most of these opponents are strong. A considerable number of core people in startup companies are former executives of OpenAI, and their personal appeal has accelerated the brain drain of OpenAI. For example, recently departed CTO Meera Murati has been inviting former colleagues to join her new company.

The departure of these contributors can be said to have led to the blooming of the AI ​​market and the proliferation of large model technology. There are 24 contributors with papers with more than 50,000 citations, and 10 of them have left OpenAI.

ChatGPT and GPT-4 shocked the world. The important reason is that most people don’t know how OpenAI does it. OpenAI at that time was equivalent to a black box. But the departing contributors know that the foundations of these products were developed with their participation.

Competitive pressure has never been more prominent: the non-profit organization MERT (also founded by former OpenAI employees) released a report in November. When evaluating the performance of large models in solving 7 AI research problems, it was found that 5 of them had problems. On the same day, Anthropic's large model Claude Sonnet 3.5, founded by another group of former colleagues, outperformed OpenAI's most advanced model o1-preview.

“Powerful AI will appear sooner or later, and it is best to be realized by people outside of Google.” In May 2015, Altman sent an email to Musk to discuss starting a project comparable to the Manhattan Project. To break Google’s monopoly on AI. Their worries are understandable. In a few months, Google's AI will defeat the world Go champion, rekindling mankind's imagination of machines with super intelligence.

9 and a half years later, OpenAI has indeed achieved its goals that year, in a way that the two main sponsors did not expect.

Keywords: Bitcoin
Share to: