Interesting opinion piece by the VC firm Andreessen-Horowitz about how AI-heavy businesses resemble both SaaS and more traditional services businesses, and how founders of such companies should plan for potentially lower margins, more custom work for customers, and poorer defensibility.
In a nutshell, the article argues that
– Running AI software requires a lot more compute power and human oversight than traditional software, reducing margins
– Customer problems are diverse and data sets are very different even for similar businesses, so there are a lot of ‘edge cases’, sometimes as much as 40-50% of total cases
– AI models, however expensive they are to run, are becoming commoditized, bringing down barriers. There are also fewer – or no – network effects.
Regardless of how challenging it is, AI is making it possible to solve many new types of problems including those involving analysing and generating media – images, audio across all sorts of domains including health and medicine. I’m looking forward to this over the next decade.