The Mirage of Progress

Introduction to The Return

Hi everyone,

I have been chomping at the bit to write Chapter 1 of The Return: AI, Progress, and the Illusion of the Future but realized this morning that I need to put in an introduction. So, I’ve modified the book proposal overview here to make a full Intro for The Return. It’s called “The Mirage of Progress,” and it provides a good overview and states the thesis of The Return. I hope that it excites the reader as well. Have a look!

The Mirage of Progress

When Machines Meet Reality

In 2018, Uber assigned a self-driving Volvo to one of their gig workers in Scottsdale, AZ. The route was routine, but the Volvo was pure AI: rigged with cameras and LIDAR sensors, all fused by a giant “brain” using the latest machine learning algorithm known as “deep learning.”

Around 10 pm, the AI-controlled Volvo exited a bridge above Tempe Town Lake and proceeded onto a long stretch of blacktop between a median and the baked dark desert. The sensors registered a vehicle 5.6 seconds up ahead, but the deep learning computer dismissed it. Three seconds later, the computer identified the vehicle as a bicycle. After another 1.1 seconds the computer waffled again, switching its constantly updated assessment to “other.” Nine-tenths of a second after that, the computer beeped a warning sound that it should slow down. As the driver grabbed the wheel, the two-ton vehicle struck a bicyclist, killing him instantly.

In 2021, after other such disasters and failed expectations—including a self-driving car ramming a school bus after mistaking it for an overpass—Elon Musk tweeted a comment that could apply to any number of elusive or failed technologies. “Generalized self-driving is a hard problem,” he said, soberly. He extended his understatement with, “Didn’t expect it to be so hard, but the difficulty is obvious in retrospect.” Then, with a philosophical twist, he added: “Nothing has more degrees of freedom than reality.”

Indeed. As technologists have come to learn the hard way, reality can bite.

That’s why Rosie the Robot remains fiction, decades after those Jetson cartoons beamed into our living rooms. ChatGPT and LLMs are impressive, but they too are a fiction: foundational models masquerade as artificial intelligence (AI) that can think, while mindlessly crunching trillions of words and other (multi-modal) data culled from the internet using gobs of computing power and statistics. AI also messes up, sometimes to the point of seeming to hallucinate—and when it does, as in the case of self-driving cars, it reveals that the emperor has no clothes.

AI has made a leap forward with the release of ChatGPT in November of 2022 and its core innovation, the “attention mechanism,” an architecture for processing a sequence of tokens that sits atop a deep neural network. No doubt. But in a sense it’s even more obvious now that we’ve been chasing a dream. Generative AI—the broader field of which LLMs like ChatGPT are a part—isn’t leading to “Artificial General Intelligence” (AGI) or to synthetic minds as we once thought. It’s leading to a world of increasingly clever and sophisticated tools. This isn’t a magical future but, to put it bluntly, one with just more digital technology. As shocking and counter-intuitive as it may be, a close look at the past and present reveals that AI is not leading us forward but reinforcing a cycle, one we fail to recognize because of our flawed view of progress.

It turns out that in a world where human experience has been reduced largely to staring at screens and simulations, thought and reality still matter. And while the question of progress in the twenty-first century might seem obvious and answered, it turns out that we’ve gotten it all wrong. We’re not leaping ahead. We’re returning to places we’ve already been.

The Myth of Technological Destiny

One factor is that by and large we’ve bought into a flawed definition of humanity and, consequently, of human progress.