This article is part of our series that explores the business of artificial intelligence
OpenAI is in discussions to sell shares in a tender offer that would value the company at $29 billion, according to a report by the Wall Street Journal. The news comes a little over a month after the artificial intelligence lab released ChatGPT, its latest and most impressive large language model (LLM).
ChatGPT has been the center of much excitement since its release. It amassed more than one million users in its first few days. There is already discussion of the LLM disrupting several industries, including online search. And Microsoft, OpenAI’s main financial backer, is considering integrating ChatGPT in its Bing search engine.
The hype around ChatGPT comes on the back of other successful releases by OpenAI, including DALL-E 2, a text-to-image generative AI model, and Codex, an LLM that specializes in generating software code.
The latest capital raise (probably an easy sell) will put OpenAI on the path to fast growth and accelerated profitability. But how will it affect the scientific mission that the San Francisco–based AI lab embarked on in 2015? That remains to be seen.
The market for generative AI
OpenAI began its work in late 2015 with $1 billion in funding from tech luminaries such as Elon Musk, Sam Altman, Peter Thiel, and Reid Hoffman. The lab’s mission was the safe pursuit of artificial general intelligence.
Since its founding, OpenAI has engaged in various types of projects, including virtual simulations, game-playing bots, robotic hands, large language models, code-generating AI, and image-generation models.
More recently, however, the company’s efforts seem to be focused on generative AI. While there is much debate on whether advanced deep learning models such as DALL-E, GPT-3, and Codex are bringing us closer to creating real AI, there’s no denying that generative AI is extremely useful.
Many successful applications are already built on top of OpenAI’s generative models. GitHub Copilot, based on Codex, has become a very successful tool in increasing the productivity of programmers. Microsoft has integrated DALL-E into several of its products. Several startups have built successful business models on top of GPT-3. And the early success of ChatGPT hints at more interesting discoveries and applications to come.
The short but fruitful history of generative AI is one of scientific interest turning into a fast-growing market—one in which OpenAI will play a big role.
OpenAI’s path to profitability
Since its founding, OpenAI has struggled with funding its extremely expensive research. In 2019, it changed its structure from nonprofit to capped profit to attract fresh funding. This led to its partnership with Microsoft, which provided OpenAI with $1 billion in funding.
Microsoft gained exclusive access to OpenAI’s state-of-the-art models, helping it catch up with Google in AI research. On the other hand, OpenAI gained subsidized access to Microsoft’s advanced cloud infrastructure to train and run its AI models. Additionally, OpenAI was able to monetize its technology through Microsoft’s expansive market reach, shortcutting its way to profitability.
According to a Reuters report, in a pitch to investors, OpenAI has said it expects $200 million in revenue in 2023 and $1 billion in 2024.
But OpenAI faces challenges to meet those targets. First, OpenAI’s experiments are extremely costly, and not all its AI models are marketable.
Moreover, OpenAI is not the only player in generative AI. It is facing big-tech competitors (e.g., Google and Meta), as well as smaller companies (e.g., Hugging Face) and startups that are providing open-source equivalents of its paid services (e.g., Stability AI and EleutherAI).
The new funding—which comes at a time when the capital market for AI is drying up—will enable OpenAI to invest in fast growth and the rollout of new products. This is especially important as incumbents and new players try to use first-mover advantage to establish their position in this emerging market.
With its track record of success in generative AI and fresh capital, OpenAI will increase its chances of remaining a prominent player in generative AI.
What is the tradeoff?
As I have discussed here before, in many cases, doing scientific research and developing marketable products are conflicting goals. The former requires testing various ideas and exploring paths that might (or might not) deliver results in the long term. The latter focuses on techniques that are more likely to deliver profitable business models in the short term.
This partly explains OpenAI’s growing focus on generative AI. That’s where the money is. But will LLMs, diffusion models, and CLIP models solve the safe AI problem? We’ll never find out if companies like OpenAI stop exploring new ideas and start exploiting marketable technologies.
Unfortunately, by selling more of its stock, OpenAI will become more beholden to the interests of its shareholders. And with the company’s promises to turn in $1 billion in 2024, I fear that the company will become stuck in doing more of the same and less innovation in the coming years.