Author: Sam Altman, CEO of OpenAI; Compiled by: Wu Zhu, Golden Finance

ChatGPT was born just over a month ago, and now we have transitioned to the next model paradigm that can perform complex reasoning. The new year has prompted reflection, and I want to share some personal thoughts on how it has developed so far, as well as some things I've learned along the way.

As we draw closer to AGI, it seems like a crucial moment to reflect on the company's progress. There is still much to understand, many things we do not know, and it is still too early. But we know much more than we did at the beginning.

We founded OpenAI about nine years ago because we believed AGI was possible, and it could be the most influential technology in human history. We wanted to figure out how to build it and ensure it benefits broadly; we were excited to try to leave our mark on history. Our ambitions are very high, and we also believe this work could benefit society in equally extraordinary ways.

At that time, very few cared, and if they did, it was mainly because they thought we had no chance of success.

In 2022, OpenAI was a quiet research lab working on something temporarily called 'Chat with GPT-3.5.' (We were much better at research than at naming.) We had been observing how people were using the playground features of our API and knew developers really enjoyed talking to the model. We thought that building a demo around this experience would convey some important information about the future to people and help us make our models better and safer.

We ultimately graciously named it ChatGPT and released it on November 30, 2022.

We have always abstractly known that at some point we would reach a tipping point where the AI revolution would begin. But we did not know when that moment would be. What surprised us was that the outcome was as it is.

The launch of ChatGPT opened up an unprecedented growth curve—within our company, the industry, and the world. We finally saw the huge benefits we had always hoped to get from artificial intelligence, and we can see that even more benefits are coming soon.

It hasn't been easy. The road has not been smooth, and the right choices have not been obvious.

Over the past two years, we have had to build a company around this new technology almost from scratch. There was no other way to train employees except through hands-on practice, and when the technology category is completely new, there is no one who can tell you what to do.

Building a company at such high speed with so little training is a chaotic process. It’s often two steps forward, one step back (sometimes one step forward, two steps back). Mistakes get corrected as you move forward, but there are actually no manuals or guides when you are doing original work. Moving quickly in uncharted waters is an incredible experience but also very stressful for all involved. Conflicts and misunderstandings abound.

These years have been the most rewarding, interesting, awesome, fascinating, exhausting, and stressful years of my life, especially the last two years, which were the most unpleasant. I feel an overwhelming sense of gratitude in my heart; I know that one day I will retire, watch plants grow on a farm, be a bit bored, and reflect on how cool it was to do the work I dreamed of since I was a child. I try to remember this on any given Friday when, by 1 PM, seven things have gone wrong.

On a Friday over a year ago, the worst thing that happened was my sudden firing during a video call, and just after we hung up, the board published a blog post about it. I was in a hotel room in Las Vegas at the time. The feeling was almost indescribable, like a dream gone wrong.

Being fired in public without warning triggered several hours of madness, followed by days of craziness. The 'fog of war' was the strangest part. None of us could get a satisfactory answer about what had happened or why it happened.

In my view, the entire incident was a massive failure of governance by well-intentioned people (including myself). In hindsight, I certainly wish I had done some things differently, and I want to believe that I am a better, more thoughtful leader now compared to a year ago.

I have also learned about the importance of having a board with diverse perspectives and rich experience when tackling a range of complex challenges. Good governance requires a lot of trust and credibility. I am grateful to so many people who have come together to build a stronger governance framework for OpenAI, enabling us to fulfill our mission of ensuring AGI benefits all of humanity.

My biggest takeaway is that I have too many people to thank: thanks to everyone who has worked at OpenAI and chosen to spend their time and energy pursuing this dream, thanks to the friends who helped us through crisis moments, thanks to the partners and clients who supported us and trusted us to help them succeed, and thanks to the people in my life who have shown me how much they care about me. [1]

We all returned to work in a more united and positive way, and I am very proud of our focus since then. The research we have done is undoubtedly the best research ever. Our weekly active users grew from about 100 million to over 300 million. Most importantly, we continue to push technologies that people genuinely enjoy into the world and solve real problems.

Nine years ago, we really had no idea what we would ultimately become; even now, we only have a rough idea. The development of artificial intelligence has experienced many twists and turns, and we expect more twists in the future.

Some twists were joyful; some were difficult. It was interesting to watch the continuous stream of research miracles occur, and many skeptics became true believers. We also saw some colleagues split off and become competitors. Teams often change as they scale, and OpenAI's expansion has been very rapid. I believe some of this is inevitable—startups typically see significant staff turnover at each new major scaling level, and at OpenAI, the numbers increase by several orders of magnitude every few months. The past two years have felt like a decade for an ordinary company. When any company grows and develops so rapidly, interests naturally diverge. When a company in any important industry is in a leading position, many people will attack it for various reasons, especially when they are trying to compete.

Our vision will not change; our strategy will continue to evolve. For example, when we started, we did not know we had to build a product company; we thought we were just going to do great research. We also did not know we would need such a large amount of funding. Now we must build things that were unfamiliar to us a few years ago, and in the future, there will be new things that we can hardly imagine now.

We are proud of our record in research and deployment to date, and we are committed to continuing to advance our thinking on safety and benefit sharing. We still believe that the best way to make AI systems safe is to iteratively and gradually release them to the world, allowing society time to adapt and evolve alongside the technology, learn from experience, and continue to make the technology safer. We believe it is important to be a world leader in safe and coordinated research and to guide this research through feedback from real-world applications.

We are now confident that we know how to build AGI according to our traditional understanding. We believe that by 2025, we might see the first AI agents 'join the workforce' and significantly change corporate output. We still believe that repeatedly putting excellent tools in people's hands will lead to widely distributed great outcomes.

We began to turn our sights toward true superintelligence. We like our current products, but we are here for a brilliant future. With superintelligence, we can do anything. Superintelligent tools could greatly accelerate scientific discovery and innovation, far beyond what we could achieve ourselves, thus greatly increasing abundance and prosperity.

It sounds like science fiction, and even discussing it is a bit crazy. That's okay—we've been there before, and we can go there again. We are very confident that in the coming years, everyone will see what we see, and it is crucial to proceed with great caution while still maximizing broad benefits and empowerment. Given the possibilities of our work, OpenAI cannot be an ordinary company.

How lucky and honored I am to be part of this work.

(Thanks to Josh Tyrangiel for the prompt. I wish we had more time.)

Note:

[1] During those days, many people did incredible work to help OpenAI and me personally, but two individuals stood out.

What Ron Conway and Brian Chesky have done goes far beyond their responsibilities, and I don't even know how to describe it. Of course, I have heard about Ron's capabilities and tenacity for years, and I've spent a lot of time with Brian over the past few years and have received tremendous help and advice.

But nothing is better than seeing what people can really do together in the trenches. I have reason to believe that without their help, OpenAI would have collapsed; they worked day and night until things were done.

While they worked very hard, they always remained calm, had clear strategic thinking, and offered great advice. They prevented me from making several mistakes, and they themselves made none. They leveraged their vast network to meet every need and were able to navigate many complex situations. I am sure they did a lot that I am unaware of.

However, what I remember most is their care, compassion, and support.

I thought I knew what it was like to support founders and companies, and to some extent, I did know. But I had never seen, nor had I ever heard of, the things that these people do, and now I understand more fully why they hold legendary status. They are each different and fully deserving of their truly unique reputations, but they have an extraordinary ability to move mountains and offer help, making unwavering commitments when needed. The tech industry is better because of the two of them.

There are others like them; this is what makes our industry very special, and their contributions to making everything work are far greater than people realize. I look forward to repaying them.

From a personal perspective, I especially thank Oli for his support that weekend and throughout; he has been incredible in every way, and no one could ask for a better partner.