Let's begin with a reiteration of a point made prominently in this book’s predecessor: that point being, This time it’s different.
That’s the answer to the challenge thrown down in response to the emergence of real-world AI in 2023, a broad assertion that new technology is always disruptive, but things will settle back down... this is just another burst of hype, and real change will take years...
Nope.
The specific argument addressed in The Pod Bay Doors was about what’s going to happen in the job market, now that AI can automate not just physical labor but brain work. ChatGPT and a cupboardful of new and innovative generative AI tools, to say nothing of the growing predictive power of machine learning in general, seem to be marching into office buildings, rather than factories; taking up cognitive tasks, rather than picking things up and putting them down.
When the industrial revolution moved millions of farmers out of the fields and onto the factory floors, that was a very lateral move. Yes, the workers needed to be retrained, but it was manual labor-for-manual labor. No great net displacement. This new breed of AI is not the same thing; tens of millions of service industry employees, truck drivers, clerks, and other low-threshold workers will be getting pink slips over the next decade, but there aren’t tens of millions of new, more-or-less equivalent jobs standing by. And those new roles that the advent of more powerful AI is creating won’t, in general, be accessible to displaced fast food workers.
This time it’s different.
And so we open our ongoing tour of AI here with another example of just how it’s different.
We take our cue from Martin Casado, a general partner at Andreessen Horowitz, who wrote in a Wall Street Journal essay that artificial intelligence has finally become transformative.
AI has been around for a good long while, of course; apart from factory robotics, which were limited in their behavioral range during their first quarter-century of deployment, there were expert systems - inference engines and knowledge bases that captured human expertise and leveraged it for decision support. These had limited success, but didn’t change much of anything.
And even when AI got really good, after the deep learning breakthroughs a little over a decade ago, its development was hampered by a discouraging development cycle. Casado described that cycle in his essay:
An AI company comes up with a new AI application;
Humans must perform the function the AI will take over, until the AI is sophisticated enough;
This usually means hiring quite a few people;
Upon launch, the AI can only handle the most common cases, and the humans have to be retained to handle the long tail of the uncommon ones;
There is high initial growth for the company, but in the end, the new app won’t scale.
That’s not transformative, either.
Generative AI, on the other hand, is another matter altogether. Its applications are highly accessible, already scaling easily, and can be plugged into an endless array of tasks in every imaginable domain. In less than a year, it has permeated the business universe, has made its way into homes and schools, and is growing exponentially.
It is changing the way we work in the office. It is changing the way we learn in school. It is making creative work easier (and faster). It is recapturing hundreds of millions of work-hours a week, a benefit distributed across tens of thousands of organizations.
That’s transformative.
“While still very early,” wrote Casado, “we’re already seeing use cases in large existing markets with orders-of-magnitude improvement in time, cost, and performance. This has led to some of the fastest-growing technology and product adoption in the history of the software industry.
“We may be experiencing what is likely the start of a new supercycle on par with the advent of the microchip or the Internet.”
That’s transformative.
He goes on to point out that even the core purpose of computing technology – accuracy and precision exceeding our own – has been displaced. ChatGPT, after all, is often wrong; it experiences ‘hallucinations,’ actually providing false information. But because of how we use it, that doesn’t matter; the Internet has already provided us with copious fact-checking capacity.
What generative AI provides, in place of accuracy, is inspiration; it can serve up suggestions, possibilities, options, all kinds of kick-starter input that can get us moving and keep us moving in the course of producing whatever it is we produce. It can write the first draft of that email or report; all we have to do is edit them. It can give us five subtopics for a blog entry or article we need to create; all we have to do is pick one (and then let it write the first draft).
Generative AI is our new incentive, our new spark, across ten thousand day-to-day office tasks – no matter our station, no matter our role
That’s transformative.
AI’s adoption rate – lackluster for decades now – has suddenly skyrocketed. And it’s no longer just being used in isolated pockets of our offices, wielded by nerds; it’s on everybody’s desk.
There will be great benefits, of course, but also great consequences. Pondering these carefully and entering into on-going dialog is going to be an essential next step.
But let’s not make the mistake of insisting that this is just business as usual, and that the furor is hyperbolic. That’s not true, and we really don’t have time to waste. Let’s get past it, and accept...
This time it’s different.
Comments