AI & Amortization

In every AI-based conference or meeting, there is this question of "What is AI? " . As a speaker in over 25 conferences, I can tell you this definition has never been settled. Every time you hear a new perspective and indeed a credible one.

One such perspective is:

AI ===> Software + Data = Software

Traditionally, a software means you create software code and publish as compiled code. So it is essentially software creating software. Data is external to the software.

AI models are made by processing tons of data + software code. This mashup produces new software ( that we call AI models) which can predict outcomes and understand unstructured data.

Organic AI research is pretty much concentrated in the hands of big boys like Nvidia, Google, Open AI and like (lets call them Category A).

The "AI companies" in fintech, healthtech, edutech etc (Category B) essentially use AI frameworks from these Category A companies and provide the data element. This data is required on a large scale to make AI really work. End-Customers, even large corporations, often do not have that volume of data available in-house.

Hence, essentially, these Category B companies train millions/billions of data inputs (images, text/video) and hope to amortize that cost by selling the data models to multiple end-customers.

I found this amortization analogy pretty interesting. And this "software+data=software" a new refreshing way of defining AI to end-users.