Eric Siegel had been working in the machine learning world for over 30 years before the rest of the world caught up. Siegel is a machine learning (ML) consultant for Fortune 500 companies, an author, and a former Columbia University professor, and according to him, AI excitement has been out of control over the past year.
While the world has embraced artificial intelligence as our great technological future, it is often difficult to distinguish from classical machine learning, which has actually been around for decades. ML predicts what ads we’ll see online, keeps inboxes free of spam, and powers facial recognition. (Siegel’s popular Machine Learning Week conference has been held since 2009.) Artificial intelligence, on the other hand, has recently come to refer to generative AI systems, such as ChatGPT, some of which can perform human tasks.
But Siegel thinks the term “artificial intelligence” exaggerates what today’s systems can do. More importantly, in his new book, The AI Playbook: Mastering the Rare Art of Machine Learning Deployment, due out in February, Siegel makes a more radical argument: The excitement around AI is fueled by the powerful, but unsexy tasks it has now proven to be. UPS, for example, was able to reduce 185 million delivery miles and save $350 million annually, largely by creating a machine learning system to predict package destinations for hundreds of millions of addresses. While it’s not exactly society-shattering, it’s certainly effective.
The AI Playbook is an antidote to the overheated rhetoric of all-powerful AI. Whether you call it artificial intelligence or machine learning – and yes, the terms get extremely blurry – the book helpfully lays out the basic steps towards deploying the technology we’re all now obsessed with. Fast Company spoke with Siegel about why so many artificial intelligence projects fail to materialize and how to get managers and engineers on the same page. The interview has been edited for length and clarity.
Why do you think the term “artificial intelligence” is so misleading?
Everyone talks about the conference held at Dartmouth in the 1950s where they started deciding how to create artificial intelligence. [Editor’s note: In 1956, leading scientists and philosophers gathered at the “Dartmouth Summer Research Project on Artificial Intelligence.” The conference is credited with launching artificial intelligence as a discipline.] This meeting is almost always reported and reiterated respectfully.
But no, I mean, the problem is what they do with branding and the concept of artificial intelligence, it’s a problem that still exists today. It’s a mythology by which you can plausibly anthropomorphize a machine. Now I don’t mean to say that theoretically a machine could never be as capable as a human. But this is the idea that you can program a machine to do anything that the human brain or the human mind does, which is a much, much, much more cumbersome proposition than people generally consider.
And they confuse [AI’s] progress and improvements at certain tasks—however impressive they actually are—with progress toward human-level capability. In other words, the aim is to abstract the word intelligence from humanity.
Your book focuses on how companies can use this technology in the real world. Whether you call it machine learning or artificial intelligence, how can companies use this technology correctly?
By focusing on truly valuable operational improvements through machine learning. We see a focus on the concrete values and realistic uses of today’s technology. The book is partly an antidote to, or a solution to, AI advertising.
So what the book does is breaks this down into a six-step process that I call BizML, which is the end-to-end application for running a machine learning project. So not only does this make the numbers crunch, but it ultimately delivers and generates a real return to the organization.
In the book you write: “ML is the most important technology in the world. This is not just because it is so widely applicable. “This is also because it is a new support that cannot be found anywhere else, a critical advantage in process optimization, which has become the latest battleground of business.” So “process optimization” seems like the most anti-AI thing possible. In five years, AI or machine What do you think the impact of learning will be on the world?Process optimization seems to indicate that things will mostly become a little more efficient and seamless.