Google has presented an interesting problem for those building LLMs. However, what about those intending to build platforms? How can someone develop moat-building end-products on LLMs when these are essentially whipped together on a weekend for a few thousand bucks with open source or potentially disrupted with AGI?
How would you create moat for linear regression? AI is more complex than liner regression but this helps you focus on the business model and not be distracted by AGI and other such bull.<p>Some companies (matlab and SAS) created proprietary software that made it easy for businesses to do data science using linear regression. Thats one approach.<p>Other companies are effectively consultants. They hold your hand to apply linear regression to your data. Thats another approach.<p>In general its not easy to build a business model around algorithms.
Making the LLM an interchangeable core and build the moat with other technology. If the LLM itself is the core technology - there won't be a moat.
You probably can't. Hardware is the biggest moat, and you need billion-dollar investments just to step up to the plate against Nvidia and AMD right now. You could try implementing something in software, but if it's not a sufficiently complex product then someone might just ask ChatGPT to write a FOSS version.<p>If there's one nice thing about the age of AI, it's going to be a lot harder to profit from "$1 fart button" apps and "DogTV" products now.
You can build a moat by connecting various groups of people and empowering them on a mission. Technology itself rarely provides the moat for end-products.<p>Your brand is your moat for end-products. Make some cool videos and start a movement