Welcome the new AI World Order
Welcome AI Overlords
It’s official (tm) AI will delete every job that you like, and also those that you don’t like. It’s totally not hype this time.
Ok the rage title game is still on, but it’s been months since ChatGPT has left its mark on the programming zeitgeist. Beyond all the huff and puff of AI and GPT models taking our jobs is some introspection on what it means to make code.
I guess I’ll join the discordant crowd
RIP my Job?
There’s been a lot of talk about replacing the coder. The highly technical person who interfaces with the machine where the bosses or the creatives cannot. These highly trained (and highly expensive) programmers have been a target for automation ever since they became a thing, and I don’t think that’s changing. In fact, a lot of coders try to code themselves out of a job by making templates and boilerplates that handle the boring work that no one wants to do.
COBOL was designed to be “highly readable” with a English-like syntax, but in reality it’s an over-designed and jumbled mess of a technology, and the result was even more highly specialized technicians who could create COBOL programs that did what the suits wanted. This pattern has been repeated a few times, and the most recent blip of this “trend” was the No-Code movement. Tools like Power BI, Zapier, and Celigo promise power without having to pay coders. The big issue is that anything that’s more difficult than the creators of these tools can think (or code & support) is impossible. As long as you don’t need customization or more complex tools you don’t need coders… but if you do need coders you need them.
The threat of automation taking away jobs is real*. AI will probably change the way we do work in the same way the internet, computers, and mechanization has changed work in the past. The historical lesson we’ve seen with disruptive technology is less dire than world-ending madness. Most don’t seem to hold this opinion, but I’ve seen people worried that “art is dead,” or that “coding is dead,” but we’ve seem this same thing many times before.
I’ll explain why I don’t think my job is actually at risk in a series of examples.
Machine Power
The advent and improvement of power transfer has “destroyed” jobs. Wind and water mills remove the need for animal power, steam engines (with heavy infrastructure) remove the need for travel by foot, and electricity compacts and further enhances traditional technology.
In the span of 300 years, humans (maybe 1% of our total time on the earth 300,000 years) have developed power into such a scale that it’s essentially unrecognizable as the same advancement from animal power. It’s true that the skill of driving animals, packing a coach, and organizing animal power into work are dead skills, but nowadays it’s not seen as a large problem. If I’m entirely mistaken, feel free to email me yocky@yockyard.com, but I’ve not seen a large movement of traditional haulers hitting the roads with coaches. If that is a thing, please let me know ASAP yocky@yockyard.com.
Automatic Fabrication
Famously, the assembly line was instrumental in Ford’s automobile and revolutionized production. We’re living in the aftermath of an industrial revolution (for better or worse), most if not all of the products and services we use rely on this mass automation. Are the jobs that Ford’s assembly line destroyed missed? Not usually, but even when they are there are artisans who keep and investigate the older methods of fabrication.
Woodworkers, iron workers, and weaponsmiths all still produce their craft but the goal is different. For most individual creators the act of creation is its reward. For most that purchase the results of these crafts do so on the merit that a person created it, or there’s no mass market for the thing, so an artisan has to make it. Automation didn’t kill these pursuits, it really on ly made them unnecessary for the majority of people.
Automatic Code Creation
Code and software run on silicon chips (for now). These chips need to send 1’s and 0’s to their many complex and interconnected parts. This process has iterated and built successively higher until we reach the point we’re at today where saying that all software is 0’s and 1’s drastically oversimplifies the leaps and bounds we have today.
Within the span of 30 years the first computer programs were created to run on specific machines with specific architectures. Each time a programmer had to make something work, they often needed comprehensive knowledge of the machine that would perform the task. The same discipline has changed so much that it’s often an accepted practice to simulate an ideal machine and run code on the virtual machine.
The practice of programming used to require advanced knowledge of intricate machines. The practice of programming used to require a programmer to know and work around the distinct limitations of a machine. The practice of programming used to ask that the programmer know what boolean algebra was and how to exploit it. The practice of programming used to require knowledge of the processor you were building to, and the commands it could use. Programming used to be a discipline of electrical engineering and mathematics - now most programmers cannot build a 557 with instructions.
These are not bad things. Programming and development should not be held back so that we can solder chips, plug wires, and punch cards ‘like real coders’ just to preserve the old ways. But things have changed in drastic ways in a single lifetime. They will change again.
The practice of programming today requires writing code in JavaScript, C++, Python, Go etc. Future programming may not require this. The practice of programming today requires the ability to diagnose complex problems based off of limited information. Future programming may drastically simplify this. Programming requires critical thinking and problem solving - I think this is the core quality that will change, but can not be removed.
ChatGPT and more directly GitHub Copilot are the first ventures into this new world. The results haven’t been groundbreaking, but I intend to try this out myself. It’s a new world and it’s more accessible than ever, now is the time that the new way of software development is being established.
Automation and Technology killing Jobs?
Until the divide of the technically-minded individual and the business-minded individual is the same, there will always be a need for developers - even if that definition changes. COBOL was designed to make it easy for business people to write programs. This did not happen.
Will this gap be bridged? In our life times? It’s hard to say; Elon certainly thinks (or thought) so in 2020 but the reality of the AI uprising hasn’t yet come to pass, and we have 2 years for that prediction to come through. Unless he meant that we have algorithms and models that perform better than humans in some tasks… then sure he’s right?
Color me skeptical, but I think AI as a tool will fundamentally change a lot of how we work. I also think that AI isn’t a silver bullet, and I’m not alone: we’re going to come to grips with the real limitations of AI quickly. I think instead of having robots which are smarter than us in every way by 2025, we’ll have new classes of AI that are pretty bad at several things, and we’ll discover the new ways that intelligence and AI have gaps and voids that are difficult to bridge.
RIP Truth
AI lies.
People lie too, but AI also doesn’t have any emotional or ego-related checks that could deter lying (yet).
I think that this is one of the worst weaknesses of the models that are intended to produce text.
The most unsettling thing about AI and GPT enabled “content” is the possibility of untruth sneaking in and masquerading as the real thing. The problem (for us) is that humans aren’t great at telling truth from fiction. Venture Beat has journalists and writers that go into this more eloquently than me, but my thoughts are that the tool is problematic unless your goal is to make even more noise and bad information.
My thought is that GPT models will enable bad actors to get more content out. Helpful (and also true) communication and content will not be helped.
So AI post-docs, AI professors and AI researchers aren’t near in our future; unless we get a formula for truth the best we’ve done is speed up the time to write papers - even then I don’t think that will be the case.
And the core problem of producing truth and the fact that discerning truth from fiction seems to be genuinely hard may be the one thing that keeps humans in the driver seat, as strange as that might seem.
Here’s my stop
The hype train might be here, but I’m not staying on. I’ll try it out. I’ll give it a go. I’ll learn about it, because it’s a great tool. AI isn’t here to delete jobs, and usher in the new order. AI is the tool, and people are still at the wheel (for now).