AI on Trial: What's at Stake for API-calling AI Startups?

February 16, 2023
The issue of many startups calling themselves AI startups when the major part of their value proposition is calling APIs.

In recent months, the buzz surrounding Artificial Intelligence has reached a fever pitch, with the general public becoming increasingly aware of its potential. While the AI market is growing rapidly, it is important to differentiate between hype and real value. At Merantix, we evaluate the AI landscape to identify areas where AI can provide sustainable value. Herewith, I want to provide our perspective on the array of newly emerging “AI Startups” that leverage foundation models at the core of their business models.

Let’s start by looking at what kicked off this concurrent hype cycle.

The main driver of the momentary AI hype has been the open-sourcing or API availability of powerful foundation models, such as Open AI’s text-to-text GPT3 model and Stability AI’s text-to-image Stable Diffusion model. This accessibility has resulted in a surge of generative AI startups, with over $2.6B in equity funding invested across 110 deals in 2022. While some of this funding has gone towards startups developing the foundation models in the spotlight, a large portion has gone to companies relying on APIs to access the latest AI models created by others.

While the API-reliant business model of these startups has advantages, such as quick launch times and minimal R&D investment, it comes with significant risks. These startups have limited proprietary IP, and their value proposition is largely in the hands of the companies controlling the APIs. It is possible that these companies may start charging more for access to these models or revoke access entirely. Venture capitalist, Michael Jackson, recently posted this meme that went viral, critiquing the lack of unique value proposition and differentiability of these “AI Startups”.

No alt text provided for this image

This risk plays out right in front of our eyes as the legal landscape for AI startups is becoming increasingly complex. Leading foundation model creators, including OpenAI, Stability AI and Midjourney, are facing lawsuits for allegedly violating open-source licenses and using data to train their models without proper authorization. In addition, Microsoft and GitHub are also being sued for the use of the Open AI-powered Copilot tool in GitHub repositories.

The outcome of the lawsuits is uncertain and arbitrary, leaving the future of these startups in the hands of the court. If the foundation model creators are found to have used data improperly or infringed on open-source licenses, they may face penalties or lose access to the data, which in turn would impact the startups relying on their APIs.

In conclusion, the growing number of lawsuits against AI startups highlights the need for responsible data usage and ownership. Despite the convenience of utilizing APIs, startups must be aware of the potential risks involved. To safeguard their business from potential setbacks, companies that rely on APIs and open source models should thoroughly examine the usage practices of the foundation models' creators and have an eye on their unique value proposition when offering AI tooling, to avoid being left with just a pretty interface when seeking future funding.

Read more