Data is not the new oil…and AI is definitely not the new software engineering
The rapid arrival of a new wave of AI-based products and services has driven many to declare that we’re entering a new era where “data is the new oil” and AI companies are the new digital disruption darlings of the stock market. But dig just a little below the surface, and we see that there are some quite significant shifts going on. In particular, when driven by Machine Learning (ML) approaches, there is a complex and challenging relationship between “desirability, feasibility and viability” that offer three fundamental lenses for understanding successful innovative solutions. Achieving alignment across these three views is proving to be much more difficult than expected.
So, I was delighted to read this new post from Andreesen Horowitz (a16z). It is the 2nd part of a great piece on The New Business of AI by Martin Casado and Matt Bornstein. The first part discussed how AI businesses are not like software engineering businesses, and the challenges they bring:
Just as SaaS ushered in a novel economic model compared to on-premise software, we believe AI is creating an essentially new type of business. So this post walks through some of the ways AI companies differ from traditional software companies and shares some advice on how to address those differences. Our goal is not to be prescriptive but rather help operators and others understand the economics and strategic landscape of AI so they can build enduring companies.
In this latest post, they expand on this theme to describe how strategies for addressing ML must be at the forefront of any successful business approach when building an AI solution.
So while our previous post outlined the challenges facing AI businesses, the goal of this post is to provide some guidance on how to tackle them. We share some of the lessons, best practices, and earned secrets we learned through formal and informal conversations with dozens of leading machine learning teams.
Many companies are introducing AI to their products to process large data sets they source from operational activities or products in use. But these data sets are often incomplete, inconsistent, and contain many outliers. This requires significant effort devoted to the process of “cleaning the data”. The experimental nature of AI, coupled with the long-tail characteristic of ML-based data sets, requires new strategies to be developed that are based within a new economics of AI solutions. One that recognizes the costs involved in training ML-based algorithms, and supports solutions that change through adaptive learning as they process large amounts of disparate data. Experience from those deeply involved in building these systems is critical.
Artificial intelligence and machine learning are only beginning to emerge from their formative stage – and the peak of the hype cycle – into a period of more practical, efficient development and operations. There is still a huge amount of work to do around the long tail and other issues, in some sense reinventing the familiar constructs of software development. It’s unlikely the economics of AI will ever quite match traditional software. But we hope this guide will help advance the conversation and spread some great advice from experienced AI builders.
One thing is certain. We’re only in the early stages of understanding the economics of AI.