Waiting on the promise
How AI could change the way we shape our world
In the most recent episode of the podcast Acquired, the hosts, Ben and David, give an in depth history of Microsoft. It’s fascinating, to say the least, however the following exchange stood out as relevant to the current state of software development:
David: The way computing worked back then, it was basically still the ENIAC days. A computer meant two things. It either meant a massive room sized machine that had about the computing power of a calculator, or it meant a human. People talked about computers as humans. Have you ever seen the movie Hidden Figures about the black women who did the calculations?
Ben: Yes. Those women were the computers.
David: They were called the computers, yes.
Are our current generation of AI tools the 21st century equivalent of “massive room sized machines” lurking in the periphery, ready to commandeer the word “developer” from our warm, fleshy hands?
Or is the trend of AI — this current generative AI boom — just hyperbole? Let’s have a look at some indications of where things might go from here.
Breaking a primitive trend
Some 2.5 million years ago an early human ancestor smashed two rocks together forming a sharp stone tool. Since then, we’ve been using our hands to shape our world.
That same primitive instinct is at work in how we build software today. If you really stop to think about it, there’s something arcane about our most advanced technology still being made with our hands. Is this really how humanity will be building technology forever?
“When I was your age, we used to make software by hand.” — No one… yet.
Similarly, back when humans were the computers — doing the manual work of crunching numbers — we used our hands to do the work. Then came the new micro-computer and we steadily began opting to offload that burden to the new digital superiors. Eventually, the wonders of the computer became so great that today, you’d be hard pressed to find a company that can run without them.
We may now be witnessing a similar culmination of events:
- There is a shortage of engineers who can build technology (aka, the IT bottleneck or global talent shortage).
- There is an ever growing and veracious need for software and technology across nearly every surface of our lives, including many seemingly insurmountable societal problems.
- And there are emerging solutions that enable the autonomous development of those technologies.
According to management consulting firm Korn Ferry, the global talent shortage is expected to reach $85.2 million by 2030, with companies around the world standing to lose over $8 trillion in lost revenue due to a lack of skilled workers. — Forbes
These factors indicate the imminence of a “computer moment” similar to that of the early human computers. It’s difficult to imagine humanity sticking to the analog method of development (manual typing of keyboards) and ignoring the tantalizing possibility of a truly autonamous translation of human ideas into real solutions.
It’s certain the way we shape our world is changing in ways that our primitive ancestors couldn’t even imagine. And now we face all new challenges — global warming, mass extinctions, forever chemicals and on — that will require revolutions of technology and societal change to fix. It’s unimaginable that we’ll still be using our slow, human bodies to do it over the coming generations of problem solvers.
Engineers shift right, then disappear… sometimes
To date, the creation of software or technology has been reserved for highly skilled — or highly funded — individuals. An ability to craft a piece of technology has been highly dependent on skill (or access to skilled labor).
Dylan Field, Figma’s CEO, has the following to say about AI’s impact on creators:
Well, the way we’ve talked about it internally is that AI has this opportunity of lowering the floor but raising the ceiling. — Dylan Field, Figma CEO
That seems pretty intuitive: AI will enable more people to create and the many creators will be able to make better work. However, I believe he is describing a temporary moment in time.
If you progress Dylan’s insight to an eventual end, it means that AI will eliminate humans from the creation process. The thought exercise to demonstrate this is the following:
- If AI can create compelling, but imperfect things…
- And there’s a way to improve that imperfection…
- Then AI will eventually create perfect things.
Unlike Dylan, I believe AI will do two things 1) push creators “to the right” in the majority of creative processes to be focused on validation and confirmation of quality and 2) in some cases, eventually eliminating the creator from the process altogether:
We’re already seeing examples of generated experiences. Here’s a demo of NVIDIA using AI to generate a world in Minecraft in real time. If this technology can be improved to the point of near perfection, or at least simplifying the feedback cycles, what’s the point of human builders?
I believe we’re in phase 1 — Augment and that we’ll begin seeing more and more AI solutions in the testing space. In fact, Chrome recently launched an AI-generated explanation of errors in Debugger… because how are you going to debug code that you, yourself didn’t create?
One of the strongest headwinds to this evolution is our cultural acceptance of outsourcing generating content to a machine. However, similar to Dylan Field, I believe offloading the bulk of this activity will free us up to spend our precious time solving higher order problems, only that the eventual outcome of that path is the relegation of humans to the role of consumers.
(I use the term consumer here not just in the capitalistic sense — we will consume novel solutions to challenging problems from the treatment of cancer to predicting the weather and the decarbonization of our world).
Early signs
In my lifetime, I don’t recall another technology that has inspired as much of a sense of awe and wonder as the current generation of generative AI solutions.
Over the last year, I’ve seen many people underestimate the capabilities of ChatGPT (myself included) only to be rocked by another advancement in a very short amount of time. This happened most recently with Sora — a leap frog in video generation from OpenAI.
As of this writing, we are still at the height of the hype cycle. AI startups are riding this wave, but we’ll soon see a thinning out of the noise and the technology will find its rightful place as another piece of a macro and mature technology infrastructure.
If Gartner is right, the AI cycle will stabilize sometime between 2025–2028 and become a mainstay, table stakes offering of any mature business platform or product. If you look to the equity markets for a signal, AI announcements are still causing price pops. When that dies down, you might expect we’ve reached AI saturation and will begin approaching more of an equilibrium from which we can make more accurate analyses.
Products like Devin, GitHub Copilot Workspace, and Amazon Q are all pushing the boundaries of what it means to build software and they’re early signs that the industry may be shifting. Only time will tell if these products find product-market-fit, but some pretty big names are certainly placing their bets.
If your livelihood, or the success of your business, is predicated on the act of converting human ideas into assets for others to consume, these are all signals that should be perking your ears.
Not perfect… yet.
The technology and concepts powering this wave of AI are not new. In fact, the underlying technology powering many of the generative AI products was first conceived of in the early 20th century.
What is new is the massive amount of data that humanity has amassed to train the models, the computing power needed to deliver the services, and a culmination of events that have not only enabled society to consider these technologies, but see their practical value and application.
However, none of it is perfect. From the misuse of proprietary data or art to train the machines to the biases these models may hold, it’s not clear that we’ve refined these systems to a point that they should be widely adopted yet. But, that will change.
There are tremendous challenges ahead of humanity and without a major shift in the way we’ve shaped our world for hundreds of thousands of years, we likely won’t be able to shape it fast enough.
These opinions are my own and do not represent those of the company I work for.
