Altman just released his year-end review.

picture

A retrospective of past entrepreneurial experience:

These years have been the most rewarding, the most interesting, the best, the most interesting, the most exhausting, the most stressful years of my life so far, and especially the last two years, the most unpleasant.

He recalls being fired without warning more than a year ago, when he was told he was being let go by video call from his hotel room. "It was like a dream came crashing down." The whole affair, in his view, was a failure of governance on the part of his team, including himself.

But even so, in those two years, they built OpenAI from scratch, the equivalent of 10 years for an average company.

There are also prospects for the New Year: 2025, the first batch of AI agents will join the workforce; Extending the time dimension, their goal has shifted to the true sense of superintelligence.

For a time, it attracted the attention of countless netizens.

picture

Altman's year-end review

ChatGPT's second birthday is just over a month old, and now we've transitioned to the next model paradigm that allows complex reasoning. The New Year is full of reflective emotions, and I wanted to share a few personal thoughts about my progress so far and some things I've learned along the way.

As AGI gets closer, now is an important time to review the company's growth. There's a lot we need to know, a lot we don't know, and it's still early days. But we know a lot more than we did when we started.

We started OpenAI nearly nine years ago because we believed that AGI was possible and that it could be the most impactful technology in human history. We want to know how to build it and make it broadly beneficial; We are happy to make our mark on history. Our ambition is extraordinary, and we also believe that this work may benefit society in an equally extraordinary way.

Few people cared about us then, and if they did, it was mostly because they didn't think we could succeed.

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

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

We always knew, abstractly, that at some point we would hit a tipping point and the AI revolution would kick in. But we don't know when that moment will be. To our surprise, this is what happened.

The launch of ChatGPT kicked off an unprecedented growth curve for our company, our industry, and the world. We're finally seeing the big gains we've been hoping for from AI, and we can expect more to come in the near future.

Over the past two years, I've rebuilt OpenAI

It won't be easy. The path is not easy, and the right choice is not obvious.

Over the last two years, we've had to build an entire company around this new technology, almost from scratch. There's no way to train people on this other than hands-on, and when the technology category is completely new, no one can tell you exactly what to do.

Building a company at such a fast pace with little to no training is a messy process. It's usually two steps forward, one back (and sometimes one step forward, two steps back). Mistakes are corrected as you go along, but there are virtually no manuals or road signs when doing original work. Traveling at high speed through uncharted waters is an incredible experience, but it is also hugely stressful for all involved. Conflicts and misunderstandings abound.

These years have been by far the most rewarding, the most interesting, the best, the most interesting, the most tiring, the most stressful, and especially the last two years the most unpleasant of my life. The overwhelming feeling is gratitude; I know that the day I retired and watched the plants grow on our ranch, it was a little boring, and looking back, how cool it was that I got to do the job I'd dreamed of doing since I was a kid. On any given Friday, when seven things go wrong before 1 p.m., I try to remember that.

A little over a year ago, on a particular Friday, the main thing that happened that day was that I was suddenly fired during a video call, and then just after we hung up, the board published a blog post about it. I was in a hotel room in Las Vegas. In a way, the feeling is almost unexplainable, like a dream gone wrong.

Being publicly fired without warning started a crazy few hours, and a pretty crazy few days. None of us can get satisfactory answers about what happened or why it happened.

In my opinion, the whole episode was a major failure of governance by well-meaning people (myself included). Looking back, I certainly wish I had done things differently, and I'd like to believe I'm a better and more thoughtful leader today than I was a year ago.

I also recognize the importance of having a diverse and experienced board of directors when managing a complex set of challenges. Good governance requires a lot of trust and credibility. I am grateful that so many people have come together to build a stronger governance system for OpenAI that will enable us to fulfill our mission to ensure AGI benefits all of humanity."

My biggest feeling is that I have so much to be thankful for, and so many people to be thankful for: Thank you to everyone who works at OpenAI and chooses to spend their time and energy chasing this dream, to our friends who help us through our moments of crisis, to our partners and customers who support us and entrust us to help them succeed, and to the people in my life who let me see 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. We've done the best research we've ever done. Our weekly active users have grown from about 100 million to more than 300 million. Most importantly, we continue to put technology into the world that people really love, solving real problems.

In 2025, the first AI agents will join the workforce

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 twists and turns in the future.

Some of these twists and turns are pleasing, some of them are hard. 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, a lot of people attack it for all sorts of reasons, especially if they're trying to compete with it.

Our vision will not change, but our strategy will continue to evolve. For example, when we started out, we had no idea we needed to build a product company; We thought we just had to do the 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, while learning lessons 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 according to our traditional understanding. 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 repeatedly putting great tools in people's hands leads to great, widely distributed 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're willing to go through 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.

How fortunate and humbled to be able to play a role in this work.