An Exceptional Parrot: Investing in an Age of AI Wizardry
At Gutter Capital, we send regular updates to our investors where we share our private point of view on the state of early stage venture and the best opportunities therein. Based on requests to share, we have decided to make a portion of our Q2 2023 quartery letter available publicly here. The following is a reproduction of what we sent to our investors:
August 24, 2023
Dear Limited Partners,
Whether you spend your days reviewing early stage pitches or following public markets, you don’t need us to tell you that AI is here. The AI narrative has captured the public imagination in a way we haven’t seen since the great metaverse rush of 2021. It is our job as early stage software investors to separate hype from reality in every cycle, and make decisions accordingly. The most common question we received from existing and prospective LPs is how our strategy has changed given recent AI developments, and we wanted to take this opportunity to share our view on the subject with you.
This AI moment is one that we at Gutter are well suited for. Over a decade ago I completed my masters in Computer Science with a focus on machine learning at Johns Hopkins. In 2010, I served as the TA for the graduate level Artificial Intelligence course. Years later, in my career as a professional gambler, I built AI tools to complement my research, and learned its limitations first hand. While I make no claim to be a leading mind on AI, understanding the technology is a good place to start.
AI is exceptionally good at solving problems where: 1– it is easy to verify that an answer is correct; 2– there is plenty of data to look to for guidance; and 3– the future is likely to resemble the past. Or, in the case of generative AI, the answers don’t need to be accurate to be interesting. If a problem doesn’t share those characteristics, AI is not the right tool to make decisions. While AI researchers know this well, OpenAI’s ChatGPT has brought conversational AI to a far wider audience of people who do not.
ChatGPT is unique in its ability to make people forget its limitations. It has dazzled the public with its human-like responses, and developers with its emergent abilities. Admittedly, it is easy to close your eyes and envision that you aren’t talking to a machine, but something else. A being that thinks and feels and dreams much like you do. Who can be plugged into the internet and with little direction cut loose to solve the most pressing of humanity’s problems. It is a beautiful vision, but it is not real.
AI is not magic. Fundamentally, it is an application doing what it was programmed to do. In ChatGPT’s case that is predicting the word that is most likely to come next. There’s a big difference between imitation and understanding. ChatGPT is not a decision maker, it’s a parrot. It repeats what it hears. It is exceptional at it, but an exceptional parrot is still a parrot. Far from general intelligence, it has the potential to create more work for humans than it will eliminate. ChatGPT is a very convincing liar.
This quarter we heard a number of pitches that asked us to forget that. We have seen founders with months of experience working in AI handwaving past the specifics, counting on large language models, like ChatGPT, to do the heavy lifting required to solve complex problems. We have seen investors with even less experience in AI rushing to fund these companies at skyhigh valuations. When there is nothing but white-space to imagine what the future might look like, some founders need only get out of the way and leave the rest to the imagination.
For us, the devil is in the details. Our criteria for investments does not change simply because AI is in the pitch. A Gutter pitch must confront an urgent problem, propose a novel solution, and identify a customer who is willing to pay a price that allows for attractive margins at scale. All of this at an entry price where we can underwrite a reasonable chance of 25x return in 5 years. To date, we haven’t seen an AI pitch yet that clears this bar, and we’re not sure we need to in order to deliver top decile returns to our LPs.
That is not because we are skeptical of an AI-centric vision for the future. In fact, we believe that we’re in the early stages of a multi-decade transition to AI that will transform the way we live and work. The question is not “is AI interesting?”, but rather “who will capture the most value?” in this generational platform shift.
We have no reason to believe that AI native startups will be the biggest winners in the new paradigm. History tells a different story. The Nasdaq-100 is full of businesses who had no mention of cloud in their initial pitch, but went on to build market leading cloud-based software applications. Netflix, today a cloud darling, spent a decade shipping DVDs across the country before even launching a streaming service. As with Netflix, the great businesses of every cycle are built in response to customer needs, using the best available tools of the time.
This cycle is no different. As every company transitioned to the cloud, so too will every company use AI. The biggest winners will be the companies that put the customer first, and the technology second. It is plausible that the market leading AI companies of the next generation may already be in the Fund I portfolio today. Their stories are barely written.
Across the portfolio, we are witnessing the transition to AI take place. Opus is auto-generating training content to take friction out of customer on-boarding. The Climate Choice is aggregating publicly available information to auto-generate carbon emissions reports for customer acquisition. Forerunner has been using AI for years to automate the upload and quality assurance of building elevation certificates. While these examples pale in comparison to pitches of AI wizardry, they are very much real. In our experience, it is the small advantages compounding into big ones that transform startups into market leaders.
In just two years as a firm, we have resisted the siren’s song of crypto, web3, and the metaverse. We are not becoming an AI fund either. We believe that AI will change all of our lives in the long run, but it’s not going to happen overnight. While business as usual is never a very satisfying answer, it is often the right one. We believe that our current strategy — engaging at the earliest stages of company building to back mission-driven founders hell bent on solving the world’s biggest problems continues to be the right one. As always, we are grateful for your support as we keep at it.
Best,
James Gettinger and Dan Teran
