Sam Altman is starting to reflect.

Less than a week into 2025, and not long after an interview with Bloomberg, the famous and controversial CEO published a blog post titled "Reflections." It reviews the evolution of ChatGPT and highlights the various "conflicts and misunderstandings" they have faced since OpenAI was founded.

He said the previous dismissals were a big failure of governance by well-meaning people. He also said that now they know how to build AGI.

reflection

ChatGPT celebrated its second birthday a little over a month ago, and now we're on to the next paradigm of models that enable complex reasoning. The arrival of a New Year always invites reflection, and I'd like to share a few personal thoughts about how things have evolved so far and what I've learned along the way.

As we got closer to AGI, it felt like a good time to take a look at where our company was going. There is still a lot to understand, a lot not yet known, and it is still early days. But we know much more now than we did when we started.

We started OpenAI almost nine years ago because we believed that AGI was possible and that it would be the most impactful technology in human history. We want to figure out how to build it and make it broadly beneficial to the world; We are excited to make our mark on history. We are ambitious, and we also believe that this work may benefit society in an equally extraordinary way.

At that time, few people paid attention to us, and if they did, it was mainly because they didn't think we had a chance to succeed.

In 2022, OpenAI is still a quiet research lab working on something tentatively called "Chatting with GPT-3.5." (We're better at research than naming.) We've been watching people use our API's playground features and know that developers really enjoy talking to models. We thought we could build a demo around this experience that would show people some important information about the future and help us make our model better and safer.

In the end, we named it ChatGPT and released it on November 30, 2022.

We've always vaguely understood that at some point we would reach a tipping point and the AI revolution would begin. But we don't know when that will be. To our surprise, ChatGPT was that tipping point.

ChatGPT set in motion a growth curve we had never seen before - in our company, in the industry, and in the world. We're finally seeing the huge benefits we've been hoping for from AI, and we can see far more benefits coming soon.

It won't be easy. The road is not easy, and there is no obvious right choice.

Over the last two years, we've had to build a company almost from scratch around this new technology. There's no way to train people to be ready for this, you just have to do it yourself, and it's such a new category of technology that no one can tell you what to do.

With so little training and having to build a company very quickly, everything becomes chaotic. We often take two steps forward and one step back (and sometimes even two steps back). We'll correct mistakes along the way, but we don't really have any manuals or guidelines for this kind of unprecedented work. Sailing fast through uncharted waters is an incredible experience, but also a huge stress for all players. Conflicts and misunderstandings go hand in hand.

These years have been the most rewarding, interesting, wonderful, and intense of my life so far, and especially the last two years have been the most unpleasant. My heart is filled with gratitude. I know I'll retire one day and be a little bored "watching the plants grow on our ranch," and lament how cool it is to be doing the job I've dreamed of doing since I was a kid. I try to remember that on any given Friday, even if seven things are going badly by 1 p.m.

A little over a year ago on a Friday (November 18, 2023), the most frustrating thing happened to me when I was abruptly fired during a video call. Then, just after hanging up the phone, the board issued an official announcement on the matter. I was in a hotel room in Las Vegas, and the feeling was almost inexplicable, like a dream gone wrong.

Being fired in public without warning kicked off a few hours of madness, and a pretty crazy few days thereafter. The strangest is the "fog of war" (who is in charge of the event). None of us will ever get a satisfactory answer as to "what happened" or "why."

In my opinion, the whole affair was a colossal failure of governance by good people, including me. Looking back, I certainly wish I had done things differently, and I believe I am a better and more thoughtful leader today than I was a year ago.

I also learned the importance of a board with diverse perspectives and experience when managing a complex set of challenges. Good governance requires a lot of trust and credibility. I appreciate that so many people are working together to build a stronger governance system for OpenAI, which allows us to pursue our mission of "ensuring AGI benefits all of humanity."

My biggest takeaway: I have a lot of people to thank. Thank you to everyone who works at OpenAI and spends their time and energy pursuing their dreams, to our friends who help us through times of crisis, to our partners and customers who support and trust us to help them succeed, and to the people in my life who show me how much they care about me.

We all got back to work in a more united and positive way and I'm very proud of our focus since then. The research we've done is undoubtedly the best research ever done. OpenAI has grown from about 100 million weekly active users to more than 300 million. Most importantly, we continue to bring technology that people really like to the world and solve real problems.

Nine years ago, we really didn't know where we were going to end up; Even now, we're only scratching the surface. The development of artificial intelligence has gone through many twists and turns, and we expect more to come.

Some of the twists are exhilarating, some are poignant. It's interesting to watch the endless stream of research miracles that have turned so many naysayers into true believers. We've also seen some of our colleagues go their separate ways and become competitors. Teams tend to evolve as they scale, and OpenAI scales very quickly. I think some of this is inevitable: startups typically see a lot of turnover at each new major scale level, and at OpenAI, there are orders of magnitude turnover every few months. The past two years have been like a decade for the average company. When any company grows so big so quickly, interests naturally diverge. And when any company is a leader in an important industry, many people attack it for various reasons, especially when trying to compete with it.

Our vision will not change, but our strategy will evolve. For example, when we started out, we had no idea we had to build a product company; We thought we were just going to do great research. We didn't know we'd need such a crazy amount of money. Now we have to build something new that we didn't understand a few years ago, and there will be new things in the future that we can barely imagine now.

We are proud of our research and deployment track record to date and are committed to continuing to advance our thinking on security and benefit-sharing." We have always believed that the best way to ensure the safety of AI systems is to roll them out into the world iteratively and gradually, giving society time to adapt to the technology and evolve with it, learning from experience and continuously improving the security of the technology. We believe it is important to be a world leader in safety and adjustment research and to guide research efforts with feedback from practical applications.

We are now confident that we know how to build AGI as we have traditionally understood it. We believe that in 2025, we may see the first AI agents "join the workforce" and materially change the output of companies. We still believe that consistently putting great tools in people's hands leads to great, far-reaching results.

We're starting to look further afield, towards superintelligence in the true sense of the word. We love our existing products, but we are here for a brilliant future. With superintelligence, we can do anything else. Superintelligent tools can greatly accelerate the rate of scientific discovery and innovation far beyond our own capabilities, in turn greatly increasing wealth and prosperity.

It sounds like science fiction, and it's crazy to even say it. It's okay. We've been there before, and we'll do it again. We firmly believe that in the coming years, everyone will see what we see, and it is important that we proceed carefully while maximizing the reach and empowerment. Given the possibilities we work with, OpenAI can't be an ordinary company.

I feel very fortunate and humbled to be able to play a role in this work.

In those few days, there were a lot of people who did an incredible amount of work to help OpenAI and me personally, but two people stood out.

Ron Conway and Brian Chesky are so far beyond my job description, I don't even know how to describe it. Of course, I've heard of Ron's ability and tenacity for a long time, and I've spent a lot of time with Brian over the past few years, getting plenty of help and advice.

But there's nothing like being in a foxhole with people to see what they can actually do. I have reason to believe that without their help, OpenAI would have fallen apart. They worked day and night for several days until the job was done.

While their work is incredibly hard, they remain calm, with clear strategic thinking and good advice. They prevented me from making a few mistakes, but they didn't make any themselves. They used their extensive network of contacts to provide me with all the services I needed and were able to navigate many complex situations. I'm sure they did a lot more than I knew about.

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

I thought I knew what it was like to support a founder and a company, and in a sense, I did. But I had never seen or even heard of anything these men did before, and now I understood more fully why they had legendary status. They are different and fully deserving of their truly unique reputations, but they are similar in their extraordinary ability to move mountains, help others, and their unwavering commitment when needed. The tech industry is better off with both of them in it.

There are many others like them. It's a wonderful special thing about our industry, and they have contributed far more to the growth of the industry than people realize. I look forward to carrying that spirit forward.

On a personal note, I want to especially thank Ollie for his support over the weekend and throughout. He was outstanding in every way and no one could have asked for a better partner.

Blog link: https://blog.samaltman.com/reflections