Just now, Altman posted a long blog post called "Reflections."

This blog post is very timely. Not long ago, Altman just accepted an interview with Bloomberg, exposing a lot of fierce information, including the attitude of the former board of OpenAI, the feud with Musk, and the view of the direction of AI in the era of Trump.

Of the former board members, he described it this way: "All the people who I thought seriously harmed my personal and corporate interests are now gone, and I'm left to pick up the pieces they left behind."

In today's "reflection," Altman looks back at the various "conflicts and misunderstandings" he experienced at OpenAI.

He called the incident "a major failure of governance by people of good will."

However, the good news is that "OpenAI already knows how to build AGI".

picture

"Rethinking" the full text

Just over a month after ChatGPT's second birthday, we have now entered a new paradigm of models capable of complex reasoning.

The New Year, always make people deep in thought. Here, I'd like to share some of my personal thoughts on the current state of the industry and the inspiration that this journey has given me.

As we get closer to general artificial intelligence, now seems like an important time to review the progress of OpenAI. To be sure, there are still many unknowns and puzzles, and we are still in the early days. But we know a lot more than we did when we started.

We started OpenAI nearly nine years ago with a vision for AGI. Because we believe this technology could be the most impactful technology in human history.

We want to figure out how to build it and make it broadly beneficial; We're excited to try and make our mark on history. Our ambitions are extremely ambitious, but we also believe that this work has the potential to benefit humanity in an equally extraordinary way.

At the time, few people cared, and if they did, it was mainly because they didn't think we could succeed.

In 2022, OpenAI is a quiet research lab working on a project tentatively called "Chat With GPT-3.5." (We're much better at research than we are at naming.)

We've been watching people use our API's playground features and found that developers really enjoy talking to the models.

We thought building a demo around this experience would show people important features of the future and help us make the model better and safer.

Eventually, the project was named ChatGPT and released on November 30, 2022.

In theory, we always knew that at some point we would reach a tipping point and the AI revolution would be detonated.

But we don't know what that moment will be. To our surprise, this was the moment.

The introduction of ChatGPT led to an unprecedented growth curve - in OpenAI, in the industry, and around the world.

We're finally seeing some of the great benefits we've been hoping for from AI, and we can expect more breakthroughs in the near future.

Starting at zero, it's tortuous, it's annoying

It hasn't been an easy journey. The road is not easy, and there are no obvious right choices.

Over the past two years, we've had to build an entire company around this new technology, almost from scratch.

The only way to develop talent is to learn by doing, and no one can tell you exactly what to do when faced with a completely new technology field.

Growing a company at such a high rate, with little ready-made experience to follow, is a tortuous process.

We often take two steps forward, then have to take one back (and sometimes even take one step forward, but have to take two steps back).

While mistakes can be corrected along the way, there really isn't any manual or guide to refer to for this kind of pioneering work.

Moving at high speed through uncharted territory is an incredible experience, but it also brings tremendous stress for all involved. Conflicts and misunderstandings ensued.

These years have been the most rewarding, the happiest, the most exciting, the most interesting, the most exhausting, the most stressful for me, and the last two years in particular have been the most stressful and the least relaxing for me.

But in the end what haunts me is gratitude; I know that one day, when I retire and watch the plants grow in the pasture, I will remember how precious it was to realize my childhood dreams.

I try to remind myself of this on Fridays when seven bad things have happened before 1 p.m.

That day, it was like a nightmare

A little over a year ago, on a Friday, the worst happened: I was suddenly fired during a video call.

Immediately after the call, the board published a blog post on the matter.

I was in a hotel room in Las Vegas. It felt like a nightmare, unreal almost beyond words.

Without warning, I was fired in public, hours of chaos followed, and days of turmoil.

What is most confusing is the "fog" - no one can get a satisfactory answer as to how and why an event happened.

This whole episode, in my view, reflects a major management failure that was not handled properly by all involved, including me, despite a well-intentioned act.

Looking back, I do wish I had done things differently, and I believe I have become a more mature and thoughtful leader over the past year.

It made me realize how important boards with diverse perspectives and management experience are in tackling complex challenges.

To achieve good corporate governance, it is necessary to establish sufficient trust and credibility.

I am truly grateful to the many people who have worked together to build a stronger governance system for OpenAI so that we can continue to pursue our mission of making AGI good for all."

My biggest lesson is gratitude, gratitude is beyond words.

Thank you to every OpenAI colleague who has chosen to invest their time and energy in pursuing this dream; Thanks to the friends who lend a helping hand in times of crisis; Thank you to our partners and customers who have always supported us and trusted us to help them succeed. I want to thank the people in my life who show their love through their actions.

We returned to work with greater unity and positivity, and I am proud of the focus we have shown since.

We've done some of the best research ever done. Our weekly active users have grown from about 100 million to more than 300 million.

Most importantly, we continue to bring to the world technology that users really love and can actually solve real problems.

Soon, the first AI agents were on the job

Nine years ago, we had no idea where we were going to end up. Even now, we are still in the process of finding out.

There have been many twists and turns in the development of AI, and we expect more to come.

Some twists are exhilarating, others challenging. It has been heartening to watch research breakthroughs emerge, and many former skeptics have turned into staunch supporters.

We have also seen some of our colleagues leave and become competitors. As we scale up, team changes are inevitable, and OpenAI is growing at an astonishing rate.

I think this is somewhat inevitable - startups typically experience a lot of turnover at each significant stage of scale growth, and at OpenAI, metrics are growing exponentially every few months.

The development of the past two years is comparable to the journey of the average company in ten years.

When any company grows and evolves at such a rapid pace, diverging interests naturally arise.

And when a company is a leader in an important industry, it is subject to many attacks for a variety of reasons, especially from competitors.

Our vision remains the same, but our strategy will continue to evolve. For example, when we started, we had no idea that we were going to build a product-oriented company, just focus on research excellence. We did not expect to need such a large investment.

Now we have to start building new things that we could not have imagined just a few years ago, and there will be many innovations in the future that we cannot imagine now.

We are proud of what we have achieved so far in research and technology deployment and will continue to advance our thinking on security and benefit-sharing.

We have always believed that the best way to build a secure AI system is to introduce it to the world iteratively and incrementally, giving society time to adapt and evolve with the technology, learning from practice, and continuously improving the security of the technology.

We believe strongly in maintaining global leadership in safety and alignment research, guided by feedback from practical applications.

We are confident that we now know how to build AGI in the traditional sense. We believe that by 2025, we may see the first "official employment" of AI agents, fundamentally changing the productivity of enterprises.

We continue to believe that iteratively putting great tools into people's hands will ultimately have a broad and positive societal impact.

We're starting to move in a much grander direction, which is truly Superintelligence.

We love what we have, but our ultimate mission is to create that glorious future.

With superintelligence, anything is possible. Superintelligent tools can dramatically accelerate scientific discovery and innovation far beyond current human capabilities, thereby greatly enhancing social wealth and human well-being.

Talking about this now may sound far-fetched, even unrealistic. But that's okay - we're used to this kind of questioning, and we don't mind continuing to face it.

We are confident that over the next few years, everyone will see what we have seen and realize how important it is to be prudent while pursuing maximum value for all.

Given the significance of our research, OpenAI is destined to be no ordinary company.

It is both a great honor and humbling to be part of this work.

Acknowledgments

There were a lot of people who went to unimaginable lengths to help OpenAI and me personally during the previous turmoil, but there were two people in particular.

Ron Conway and Brian Chesky went above and beyond the call of duty, beyond words.

Over the years, I have heard of Ron's outstanding abilities and tenacity, and I have been in close contact with Brian over the years, receiving invaluable help and advice from him.

But it is only by standing side by side in a time of crisis that one can truly witness one's strength. I can say that without their help, OpenAI would probably fall apart; They worked day and night until the problem was solved.

Despite the incredible intensity of the work, they remained calm and demonstrated clear strategic vision and insight throughout the process.

Not only have they saved me from a few major mistakes, they have never made a mistake themselves. They mobilize huge human resources to deal with all kinds of affairs and skillfully navigate various complex situations. I'm sure they're doing a lot of things in secret that I don't know about.

However, what moved me most was the care, empathy and support they showed.

I thought I knew what it was to support a founder and a company, and in a sense I did. But I have never seen, or even heard, a story like theirs. Now, I really understand why they have such legendary status.

While they are all different and deserve their own unique reputation, they have something in common: an extraordinary ability to help others and an unwavering commitment to support them when they need it most.

The tech industry thrives on their presence.

Of course, there are many others like them. This is a valuable trait in our industry, and it plays a bigger role in the development of the entire industry than people think. I look forward to one day being able to repay others in the same way.

On a personal level, I want to especially thank Ollie for her support over the weekend and throughout; He was perfect in every way and could not have asked for a better partner.

Reference materials:

https://x.com/sama/status/1876081442775793984