My previous blog posts on AI safety and effectiveness have tended to take the perspective of those who build AI systems, since until recently users traditionally held very little responsibility in the AI space. I wasn’t writing only for these folks, but also for a general audience of people keen to take their places as informed citizens in an increasingly data-fueled world.
In the context of the last decade, both groups needed to understand the same thing: what AI is useful for and how smart organizations make decisions about the AI systems they build.
I’d intended for business leaders and AI professionals to use my musings to sanity check their own thought processes, while members of the general public could use them to learn enough about the decision-making that goes into these systems to insist on accountability and better technology leadership.
Seeing generative AI as a raw material might be the perspective shift that’ll make everything click.
In a previous blog post, I explained that the key difference between today’s most hyped AI products and the traditional enterprise-grade AI of the last decade boils down to UX:
Last decade’s AI user experience (UX) emphasized seamlessness, so users didn’t need to know about the AI components they were interacting with. That’s changing with a new philosophy that encourages a user’s direct interaction with AI outputs as useful raw materials rather than finished products.
In other words, let’s give the users an easy interface for tinkering with AI and let them use it in whatever creative ways they like. Today’s AI buzz is the result of all those suddenly-empowered voices. The revolution is a more of a user experience (UX) revolution, much more so than an AI revolution per se. And as a UX revolution, it deserves its hype: what an exciting new way to interact with computers!
But wait, what’s this about raw materials? That’s the crux of the whole matter… and the key to understanding this brave new world. This article is Part 1 of a series, so if you stop reading now and you take only one insight away with you, let it be this:
Seeing generative AI as a raw material might be the perspective shift you’ve been looking for to get your thinking unstuck and help you make sense of the rhetoric around you.
How is this a departure from the usual way of thinking of AI?
Well, speaking of departures, if you’re running an airline, you’ll want to build separate traditional enterprise-grade AI systems that each solve one monolithic task really well at scale, such as pricing airline tickets or optimizing your equipment maintenance. You probably don’t want the same system handling both. Before you deploy one of these monsters throughout your business, you’re going to put plenty of care into making sure that it works safely. If you’re interested in how enterprises think about AI productionization, here’s 2 hours of me telling you all about it:
When we’re dealing with typical enterprise-scale AI systems, trust and safety is easy.
Before you all rush at me with your pitchforks, give me a moment to explain myself in Part 2 of this miniseries on AI regulation before we continue our journey to understanding why the raw material rhetoric makes generative AI reliability and regulation such a tricky topic.