After the dramatic changes What is the current situation of OpenAI? What is the main driver of their revenue?

OpenAI Assessing the Current State and Main Revenue Generator Following Significant Changes

This interview combines the interview with COO Brian Lightcap by CNBC after the Dev Day event, as well as the discussions following the leadership changes at OpenAI last week. The article also addresses several questions: What is the current state of the company? What is the biggest revenue driver at the moment? What is currently the most overrated and underrated aspect of AI? What is OpenAI’s biggest revenue driver at the moment? What does this technology really mean for businesses?

Why launch ChatGPT?

In the weeks leading up to the launch of ChatGPT in November 2022, the OpenAI executive team dedicated specific time to discuss one question: Should they release this tool?

• Celestia ($TIA) and the Cryptocurrency You Can’t Miss Stack Up Now!
• Chimpzee Turning Sustainability into Crypto Passive Income – Investing in a Greener Future with Profits
• Robinhood cryptocurrency trading in November grew by 75%, annual revenue may reach a figure of nine digits.

“If you know Sam Altman, you’ll know he likes to discuss various topics efficiently and quickly, so the fact that we spent so much time on this topic meant it was important. No one can be 100% sure if this would be the right thing to do or if it’s worth our time.”

Brian Lightcap recalled that at the time, OpenAI’s GPU resources were very limited, and the company positioned itself as a developer and enterprise tool company. Even in such circumstances, CEO Altman was very supportive of “trying it out,” based on the notion that interacting with models based on text is crucial and has personalized elements.

This decision ultimately paid off. Data shows that ChatGPT broke records and became the fastest-growing consumer application, with approximately 100 million weekly active users and over 92% of the Fortune 500 companies using ChatGPT. According to PitchBook, earlier this year, Microsoft added an additional $10 billion investment into the company, making it the largest AI-focused investment this year.

However, the recent ups and downs the company has experienced in the past few weeks seem to overshadow those impressive milestones. Last month, the OpenAI board suddenly decided to remove Sam Altman from his position, triggering opposition and resignation threats from nearly all OpenAI employees, including a signed open letter and shocking the investors, including Microsoft.

In less than a week, Altman returned to the company. Last Wednesday, OpenAI announced the formation of a new board of directors, including former Co-CEO Bret Taylor, Larry Summers, and Quora CEO Adam D’Angelo. Microsoft has an observer seat on the OpenAI board.

This interview combines the interview with COO Brian Lightcap by CNBC after the Dev Day event, as well as the discussions following the leadership changes.

Q: We are about to celebrate the one-year anniversary of ChatGPT. At this time last year, during the first few weeks of ChatGPT’s debut, DALL-E was still in its beta phase, Stable Diffusion was getting a lot of attention, and ChatGPT didn’t exist yet. What was your team like at that time?

Answer: At that time, we all thought of ourselves as a company specialized in developing tools for developers. So for OpenAI at that time, considering making something that “ordinary people can pick up and use” was a bit novel.

We had a similar feeling with DALL-E, a product we released before. After people used DALL-E, we found that many users were very excited about it. But we always believed that DALL-E essentially had visual attributes (which are easier to attract attention), so it became a high point of interest for consumers in these tools. Therefore, when we were researching ChatGPT, we used DALL-E as a benchmark to assess how many people would use it, who would be interested in it, and whether ChatGPT would make people feel like it’s not a real tool but more like a toy after playing with it for a while.

At that time, the OpenAI team even made a bet on the scale of ChatGPT. My bet was that at its peak, there would be one million concurrent users at any given point. We even made plans based on this goal, and as a financial person, I did a lot of modeling calculations for this prediction. That was the situation back then, but now looking back, that estimate was not accurate.

As for business opportunities, what were OpenAI’s expectations for ChatGPT at the time?

Answer: We couldn’t know all its uses at the time, and that’s the paradox of this technology. ChatGPT has too many possible applications, and as a tool, it has permeated every aspect of the world and people’s lives, but users didn’t even know they needed such a tool.

Therefore, it’s necessary to conduct business analysis in advance and try to think, “Alright, what would people use ChatGPT for? What drives continuous payment for it?” And then start trying to make it practical. Try to think of ChatGPT as, “People might use ChatGPT for creative writing, or for this or that.” In a way, there are many things that, looking back now, we know people use it for. But at that time, we couldn’t imagine it at all, and we couldn’t prove why ChatGPT would be so successful.

This may be an interesting lesson that business analysis doesn’t always explain everything, but being able to bet on and truly discover the wide practicality and value of something, as well as its resonance as a new thing with people, is even more powerful than business analysis.

Question: In August of this year, 80% of the Fortune 500 companies used ChatGPT. Now, as of November, it has reached 92%. Are there any trends for the remaining 8% of companies who have not adopted this tool?

Answer: In a sense, the 8% of companies are mainly focused on heavy industry. … Industries like oil and gas, which are capital-intensive, or industries with a lot of heavy machinery, these industries are more focused on manufacturing products rather than information or service businesses.

Question: In your opinion, what are the most exaggerated and underestimated aspects of artificial intelligence today?

Answer: I believe the most exaggerated aspect of artificial intelligence is the notion that it can bring about substantial business transformation overnight. We have spoken to many companies who want to associate their long-standing goals with OpenAI, such as saying “we want to increase revenue by 15% year over year” or “we want to cut costs by X amount from this project.” But there is no magic pill, and you can never solve a problem entirely with artificial intelligence. I think this actually proves that the world is vast, complex, and these systems are still evolving, they are really still in their early stages.

On the other hand, what I think is underestimated about this technology is the level of personal empowerment and authorization it provides to end-users. We hear stories from users or clients that thanks to tools like ChatGPT, people now have superpowers they didn’t have before.

Question: Let’s talk about the business of generative AI. Opponents mention the abundance of consumer-level applications, but will there be an oversaturation? What does this technology ultimately mean for businesses?

Answer: We are currently in the very early stages of AI, and it is crucial to maintain a high rate of experimentation and trial-and-error in the world. If you look at the history of technological paradigm shifts, there is always this important experimental phase. It is challenging to get the right technology from scratch. We will eventually reach our destination, the final state of the technology, and we will eventually converge on that point – but only after truly trying out many things, seeing what works and what doesn’t, can people create the next best thing based on an effective foundation.

My perspective is that the most significant things built on this technology are the ones that haven’t been created yet. To truly understand the capabilities of these tools and how they can be combined with other technologies to create something even more powerful than the sum of its parts, it takes a cycle of using these tools. Therefore, it is expected, and I think it’s very healthy.

Question: A few years ago, people were surprised by the level of AI applications in the trucking industry, as it was considered a rather traditional field. Now, AI has become prevalent in almost every industry. Based on your recent observations of application trends, is there a similar thread where an industry is using AI in a completely new or different way that particularly surprised you?

Answer: The technology industry is definitely very attractive. What we see is that ChatGPT is a great technology assistant, whether you’re a software engineer, mechanical engineer, chemist, or biologist, there is a vast knowledge base on the other side of the discipline, and your mastery of it determines your work efficiency.

People strive to master this discipline in their careers and absorb as much knowledge in the field as possible. Especially in certain fields – whether it’s biology, chemistry, or artificial intelligence, the literature in these fields is constantly evolving, constantly discovering new things, and constantly completing new research. So, I don’t know if this itself is surprising, but one of the coolest things we see is that ChatGPT is almost like an assistant in this regard, almost like a research assistant. …and we feel the attractiveness from these industries, which is something I didn’t expect in November 2022.

Question: ChatGPT Enterprise has been online for a few months now. I remember you launched ChatGPT in less than a year of development time and had over 20 companies (such as Block and Canva) participating in the testing. Specifically, how has the usage grown? Who has been your biggest customer since the launch?

Answer: Overwhelming enthusiasm, but we are still a small team, and much of the focus in the past two months has been on making sure that the first batch of customers we implemented and onboarded see the value of the product….we are still dealing with thousands of waitlisted customers, and we hope to reach everyone, which will be the goal for 2024.

Question: Now that there is ChatGPT Enterprise, what is currently the biggest revenue driver for OpenAI? How do you think this will develop?

Answer: OpenAI almost never takes a revenue-driven approach to building and launching products. We always take an application-centric approach, which means we highly value the products we build, which need to meet one of two criteria in both fields: 1. They need to be genuinely useful tools for developers to build things; or 2. They need to be genuinely useful abstractions for users to find more value in the product.

For example, if we look at GPT, it actually maps quite nicely, hoping it checks the second part of the box: “This is a way of abstracting the intelligence of ChatGPT and pointing it to something very specific, and give it the right context, the right tools, the right connections, to really excel at solving specific things? Something that might be useful for your work, or it might be useful for your life, or it might just be something fun.

Question: Multimodal ChatGPT, which offers image generation and other tools within the same service – this is a big priority outlined by the company at Dev Day. Please tell me why it is so important.

Answer: The world is multimodal. If we think about how we as humans perceive and interact with the world, there’s more to it than just reading and writing text. The world is bigger than words. So, for us, relying solely on text and code as a single modality, a single interface, to understand these models and what they can do, always feels incomplete.

That’s why we’re starting to layer in visual capabilities. The fact that computers can now see what’s happening in the world, describe it, engage with it, and reason about it is mind-blowing to me, and it’s probably the most shocking thing I’ve seen in my five years at OpenAI. I still don’t fully internalize what it means, but when you stop and think about it, things that were previously unthinkable are now becoming possible.

Think about simple things like helping the visually impaired better understand the world around them in low-latency, high-quality ways. Or companies being able to better understand their devices and create experiences for consumers simply by pointing a camera at something and getting insights into how it works. Or in educational settings, being able to help people understand and analyze things – a lot of people are visual learners – being able to see something and engage with it in a way that suits their style of learning is a completely different ability that we’ve unlocked.

So, what excites me is that it provides us with a way to use technology in a way that aligns with how humans engage with the world – and ultimately makes technology more human-friendly.

Question: We know that OpenAI’s GPT-4 language model may be more reliable than GPT-3.5, but also more susceptible to potential biases. Can you tell me about the new Turbo model that was announced at Dev Day and what, if any, plans you have to address that?

Answer: We’ll be releasing a Turbo model card [a transparency tool for AI models], which might be a better place to reference some of the technical benchmarks.

Question: What is your biggest hope for the next year? What can future versions of GPT achieve that the current versions cannot?

Answer: I think the curve of progress here is going to be around the quality of reasoning. At a fundamental level, what humans are good at is taking a bunch of disparate concepts, putting them together in a creative way to produce an outcome that addresses the thing we want to do or what someone has tasked us to do. We do it every day, and we do it artistically every day, and that’s how we create the world.

That’s the direction we hope to see the technology go – enhanced reasoning capabilities; being able to handle increasingly complex tasks and figure out how to break those tasks down into their constituent parts in order to perform them at a high level of proficiency; and then, adjacent to all of these tasks, being able to perform them safely, and we emphasize the safety aspect of the technology from a research perspective. As systems become more capable, we need to simultaneously raise the standard of safety, because over time, these systems will become increasingly autonomous. Without taking care of the safety aspect at the same time, it won’t work.

Question: Which day in the past year impressed you the most at the company?

Answer: The day we launched GPT-4. People didn’t realize how long we had been researching before releasing it. So, there was a lot of excitement internally at OpenAI because we knew it would be a true transformation in the capabilities of these models and how people perceive high-quality language models. Once you have such a model, you want to share it with the world. Our team gained a lot of energy from the world’s response to these innovations, and seeing the excitement from our customers, developers, and users.

In the seven to eight months prior, we knew that moment was coming, so we were always excited…

At that time, we didn’t have a big release event like Dev Day. But right after we released it, there was a moment, I think we were in the all-hands space in the cafeteria, everyone looking around, a mix of excitement, relief, and exhaustion, but everyone was smiling. It was a very special moment… moments like these are precious.

Question: What did you personally do to celebrate when you got home?

Answer: I celebrated by working late into the night.

Question: In less than 10 years, OpenAI has gone from a non-profit company to a “research and model deployment” company. People ask what that means, what your structure is like, and how much ownership Microsoft has. Can you explain this journey?

Answer: At a high level, we wanted to keep the core structure of the company aligned with the original OpenAI (the non-profit organization). When we formed the company, the question was how to achieve that. That was the work I did when I first joined OpenAI: figuring out if there was a way to put OpenAI’s mission – and the non-profit organization as an embodiment of that mission – at the heart of our new structure.

So, I think the first thing to understand about OpenAI is that, in that sense, it’s not a regular company. In a literal sense, structurally, and spiritually, it is an extension of the non-profit mission. Its primary responsibility is to carry out the non-profit’s mission of building safe and beneficial artificial general intelligence for humanity. Perhaps it sounds crazy, and certainly, there are simpler structures and technical approaches to building a company, and there would be lower legal costs and less hassle, but for us, getting this right is really important. So I don’t know if we have done it, I think time will tell. One good thing is that this structure is highly adaptable. So over time, we’ll learn more and we’ll have to adapt to the world, and we can make sure that this structure is set up for success, but at its core, what we want is to retain the core mission of OpenAI as the reason for the company’s existence.

Question: Let’s talk about Microsoft’s ownership?

Answer: I won’t comment on specific details of the ownership structure, but it is a structure designed to collaborate with the world, and Microsoft happens to be a great partner in that. But we have always been thinking about how to extend this structure to the world and interact with the world in a way that aligns with the mission of non-profit organizations. So, I believe this is also part of the foundation of the profit-capping model.

Question: When OpenAI was first founded, you worked together with Sam Altman. What are the main differences between you in your work? And what advantages and disadvantages do you complement each other on?

Answer: Sam works at an incredible speed. I think one common thing between him and me is that we like to maintain high-speed in everything we do.

I think where we balance each other is that Sam is definitely future-oriented, he is someone who lives in the future. And I think he should indeed live in the future, he is excellent in that aspect. My job is to ensure that the way we build the company, the way we conduct our business, the way we establish collaboration models with customers and partners, not only reflects where we think the world will go based on a five-year horizon, but also accomplishes the goals we want to achieve today.

The challenge we face is the rapidly changing technology. Therefore, we place great importance on communicating to the world how to use technology, the types of work we do (from security to capability), our thinking about products, and the constantly evolving nature of our products. When the ground beneath you is changing rapidly, you need to coordinate well to do these things while operating at high speed. So, I hope my added value is here, focusing on doing these things well – building an excellent team to help us do these things. If we can do that and stay grounded, we will ultimately find the right path.

Question: OpenAI has undergone significant changes in just a week. Now Sam has returned to the company, and the new board structure has also been announced. What are your thoughts on how this will affect daily work? Are there expected to be other structural changes in the coming months?

Answer: There will be no changes to daily work, OpenAI’s mission remains unchanged, and our focus is still on conducting excellent research and development for our customers, users, and partners, and serving them. We have already shared that we now have a preliminary board and hope to add more board members.

Question: What is the overall atmosphere in the company now?

Answer: In the past few weeks, the company has come together in an indescribable way. I am grateful for the team and deeply appreciate my customers and partners who have supported us throughout. This support energizes us and makes us work harder to fulfill our mission. Personally, I’m very focused.

(Lightcap and OpenAI decline to comment further on the specifics of Altman’s resignation and reappointment)

References:

1. https://www.cnbc.com/2023/12/04/openai-coo-brad-lightcap-interview-with-cnbc.html

• Modal GPT-4

We will continue to update Blocking; if you have any questions or suggestions, please contact us!