202501031435 Status: #idea Tags: #ai #business #strategy # Algorithmic advances do not represent a true business moat It has frequently been said that part of Silicon Valley's success is the lack of non-compete clauses for employees. This allowed trade secrets to proliferate rapidly in the Bay Area, creating more efficient competition dynamics and allowing many engineers to learn from each other, rather than restricting learnings and competitive advantages to a single firm. However, I rarely see this same line of argument applied to business moats. If this holds true, then it implies that algorithms alone *cannot* provide a durable moat to a business. Employees can easily leave one company and take all of its hard-won knowledge to a competitor, allowing the competitor to catch up. Hence, algorithms only provide moats insofar as they facilitate the construction of another type of moat. As a concrete example, Facebook's website code and "People You May Know" algorithms are trivially replicable, even when they represented the state of the art. As such, the algorithms and code themselves did *not* provide Facebook with a moat. However, they did enable Facebook to construct a real moat, which were the network effects generated by getting many people onto the service and sending friend requests as quickly as possible. This strong network became insurmountable for its competitors. We are now seeing a similar dynamic at play in the race to Artificial General Intelligence (AGI). OpenAI raced out to an early lead because, realizing the implications of the transformer architecture and its scaling laws, it was the first lab to put serious effort into scaling up the compute and data applied to the auto-regressive transformer. This work did indeed require significant engineering work to ensure training stability, scalable data ingestion pipelines, and more. However, this algorithmic work was easily replicated by other AI labs (such as DeepMind, Anthropic, XAI, and more) once it became clear that scaling transformers on natural language was an effective path forward. The real moat will thus come from somewhere else - OpenAI is betting on it being RLHF from interactions with users in a chat interface, while Google is betting on it being data and compute scale: [[If data & compute are the main drivers of AI progress, then Google will win the AI race]] --- # References