Disclaimer: These thoughts are my own and do not constitute investment or legal advice. They are purely educational. Please refer to the terms and conditions for further information.
OpenAI has been at the forefront of the latest hypercycle in Artificial Intelligence. Driven by the visibility of ChatGPT, OpenAI has managed to captivate the news cycle and pushed AI into many different businesses. With the presence in the news cycle and the associated fear of founders and investors missing out on the AI boom, many overlooked that OpenAI is itself a startup. Worse still, it is a startup that is losing money. As of mid-August 2023, OpenAI is moving more toward bankruptcy than profitability.
As an investor and board member of multiple startups, including AI-aligned ones, reliance on OpenAI becomes a strategic risk question. Thus, let us look at the impact of OpenAI and what steps are possible to protect against an overly strong dependency.
The strategic importance of OpenAi
Everyone has heard about the ChatGPT chatbots, which allow everyone to interact with Large Language Models (LLM). These have driven AI to the forefront of the technology and general news cycle and dominated the discussion about the opportunities and dangers of the technology.
Yet, for the technology sector and OpenAI, GPT-3 and GPT-4, accessible via APIs, are far more critical. They allow companies to integrate the capabilities of AI into their software without going through the trouble of training an in-house LLM.
Consequently, they enable companies to rapidly and cost-efficiently prototype AI-based features. Yet, the dependency means that any risk to OpenAI becomes a risk to the companies utilizing GPT-3/4. Bankruptcy or reorganization after acquisition are two of these risks.
Current worries about OpenAI
Currently, OpenAI is spending significant funds to drive the development of AI. At the same time, the revenue streams by selling access to its AI model aren’t yet well developed. Many applications that utilize AI need to be developed and sold. Thus, high development costs and low revenue make OpenAI currently unsustainable.
An unclear vision is compounding this problem. While OpenAI should focus on strengthening its balance sheet, its founder is more focused on playing politics and turning AI closer into the Science Fiction version, as we know from the media. Thus, new investors that might come in to profit from the current state of OpenAI are skittish about the prospects in the future.
Compounding the issue is the current status of ChatGPT and how it might translate into OpenAI’s other products. Thus, the loss of users and the reduction in accuracy that ChatGPT has experienced over the past couple of weeks is impacting the expectation of the API customers. So far, there are no reports that the GPT-3 and 4 versions experience similar problems. However, the perception of the issues and loss of momentum might be enough to have business customers reconsider.
While none of the issues in itself might bring OpenAI down, the combination of problems and viable competitors might be enough to push the company over the edge.
Strategic Risk Management to Protect Startups
Consequently, the management, risk committees, and boards of AI-utilizing companies should look closely at potential mitigating and risk management strategies. The IT management should take the lead and look at the technical aspects, such as modularization and separating in-house code and business logic from integration. The accompanying board question would be how fast IT and product development can switch and replacement the integration and whether strategic partners are needed to replace OpenAI.
Given the meteoric rise of open-source AI, talent management should be another area to examine closely. Given that a team at Stanford was able to adapt Llama to their needs, companies with a core focus on AI might want to explore having the capabilities in-house.
Lastly, with the ebbing of the hype cycle, it is worthwhile to question the strategic need for AI. While, in many cases, AI goes beyond glorified chatbots, the current capabilities and advances in AI aren’t always worth the substantial costs and risks involved.
Even if all these steps remain pure thought experiments, going through the motions will help prepare a company for a future where it might outgrow OpenAI.
AI Strategies cannot depend on one player
It is unquestioned that OpenAI significantly boosted the visibility and kickstarted the next level of digital transformation. However, the leadership of OpenAI shouldn’t translate into a monopoly.
Given that startups and companies alike survive if they can solve customer problems. We shouldn’t make our companies dependent on a single partner.
Even if we never need to change our AI partner, it is only prudent to consider the failure of OpenAI. The strategic risk analysis and possible mitigation strategies are logical next steps.