Why I Won’t Invest In Your AI Company

There will be lots of companies built from the ground up leveraging AI. The vast majority will not be AI companies.

The investment universe is continually evolving and reshaping itself, pivoting around the next 'big thing.' For the past couple of years, Artificial Intelligence (AI), especially large language models (LLMs), has been the lone brightspot amidst carnage in the startup world. It's cutting-edge, it's cool, and it's a game-changer. Yet, in the words of Sam Lessin, I'm inclined to believe that "AI is a shitty place to invest." This statement isn't an indictment of AI's potential or its technological prowess but an analysis of its economic viability.

Sure, AI is cool, but being 'cool' isn't synonymous with 'profitable' or 'sustainable'. The extraordinary pace at which AI, and particularly LLMs, have evolved is awe-inspiring. The advent of entities such as OpenAI, Anthropic, and Cohere, not to mention models built by tech giants Google, Amazon, Microsoft, and Facebook, have all been made possible due to the tremendous volumes of human writing available across open protocols like the web. 

However, the fact that LLMs had readily accessible training data is the very reason why competitive moats are absent. At best you’re going to train your data on the same datasets OpenAI and others already have. There's no scarcity, no proprietary edge. As companies like Reddit and Stackoverflow clamp down on low cost use of their content for AI training the ability to start from scratch gets harder. The ability to build a defensible, sustainable business gets tougher.

If you're considering venturing into AI investing now, you might be too late to the party. The field is already crowded, and the big players have an unbeatable advantage. The vast budgets and closed data sets that these tech behemoths possess are virtually unmatchable. Even for unicorn AI startups, keeping pace is a formidable challenge.

This doesn't mean that there aren't opportunities within the AI landscape. OpenAI, for instance, has fostered a vibrant ecosystem. There will be countless sub-venture scale businesses, perhaps consisting of small teams who will fine-tune LLMs for specific use-cases. These ventures may not scale to a billion-dollar valuation, but they hold the potential for targeted, niche impact.

Consider the way AI will permeate across sectors and startups. In much the same way that businesses have leveraged smartphones for various use-cases in the last decade, AI will become integral to the operational blueprint of future startups. What once seemed revolutionary, such as check scanning, eventually becomes table stakes. This trajectory will likely be true for many AI applications.

One area where AI falls short, however, is creativity. Take this blog post, for instance. While an AI like ChatGPT might assist in crafting the text, it can't write the post entirely on its own. Human creativity draws upon a rich tapestry of senses, experiences, and contextual knowledge that AI simply doesn't possess.

There will be lots of companies built from the ground up leveraging AI. Most of them will not be AI companies. I invest primarily in healthcare, financial services, insurance and other unsexy sectors. I won’t invest in your AI company.