Bradford Cross is back. Following up on his projections for the AI startup market, he now has a four factor approach to building a business around AI. The factors from the top level? 1. Full stack products 2. Subject matter expertise 3. Proprietary data 4. AI delivers core value.
Read Cross’ in-depth post to get the details. And there’s plenty of detail. Apparently there’s also a YouTube recording of a parallel talk he gave that has some further pertinent details.
The one thing I have in my mind is how much “productionization” (sic) factors in. Cross would probably just roll it into point 1, but it’s pretty clear that at this point, actually building stable production systems around AI is ridiculously difficult. You begin in a place, model design and construction, that has extreme asymmetries in regards to expertise. Then layer on that the basic practices of software engineering aren’t particularly applicable to AI products which in addition consume resources in vastly different fashions than typical applications. There’s no Rails or Django for this stuff. Agile and DevOps aren’t straightforwardly applicable to AI unless you’re Google, MS, Amazon, FB, etc.
Could production AI capabilities be commoditized? Soon? Maybe, maybe not. I wouldn’t hold my breath though.