AI might not replace you. But it’s probably already replaced your replacement

If you’ve played around with ChatGPT at all, there’s a good chance you’ve had a few moments of existential crisis over the past few weeks. Throughout history, every technological breakthrough has usually been accompanied by prophecies of doom and mass job losses, but something about ChatGPT feels different. When you sit with it for any amount of time, your mind can’t help but wander to all the jobs in all the industries that this thing can potentially replace right now, let alone in 5 or 10 years time. It’s so many it’s almost impossible to count, and I can’t think of another tool in my lifetime that has had such obvious potential to disrupt so many different areas of life, all at the same time.

However, having used ChatGPT for some time, I am now willing to stick my neck out and say I think it’s a long way off being able to independently handle most of the tasks it’s capable of tackling. It’s still a tool that needs an expert at the wheel. It is frequently completely wrong, but wrong with such confidence that the only way to know this is to have the expertise yourself. If you asked me a legal question and I didn’t know the answer, I would tell you that I didn’t know. Even if I tried to pretend I did know, you’d see through it almost immediately, because I’m obviously not a lawyer. But ChatGPT makes stuff up and delivers answers with such confidence that it’s impossible to distinguish lie from the truth unless you know what the truth actually is already. This capacity for confident error doesn’t just make it a useless tool when placed in the hands of an unknowledgeable person, it actually makes it dangerous.

However, what is also undeniably true is that it enables an expert to become orders of magnitude more productive. And this isn’t the good news it perhaps seems to be. Far from it, I think it’s a potential disaster. This tool creates a working world where experienced people no longer need the support of inexperienced people. Think about what that means for a moment.

I make software, and conservatively, I think my output has doubled since I began using ChatGPT. Every time I leave my desk now, for whatever reason, before I go I will ask ChatGPT to create some code I need, check some I’ve already written, or create some unit tests. Then when I get back, it has usually completed a bunch of work that would previously have taken me hours. I’m getting more work done while I go and get coffee than I used to get done in a whole day. I spend a bit of time cleaning it up, then I repeat the process. I’m far more productive, but I write far less code myself.

There are two problems with this, and neither of them relate directly to me, or other experienced people like me. For me, this tool is fantastic, and I’d be a fool not to use it. But giving this tool to someone that hasn’t ever had to actually write the kind of things it’s producing for me themselves seems incredibly short-sighted. The very reason I am able to cleanup the code ChatGPT produces is that I’ve had to make the kind of things it produces myself more times than I can count. I’ve had to solve the problems. I’ve had to learn. How will a junior developer gain this experience when they don’t actually ever have to write the code? I really struggle to see a path for growth for junior developers in the AI age.

But the bigger problem is that it probably doesn’t matter that their learning may be stunted by dependence on this tool, because I worry they might not have a job to begin with. To be brutally honest, if my senior developers each have a ChatGPT account, I can see absolutely no need to ever hire a junior software developer again. ChatGPT can do the job better than a junior. All I need is senior developers who can tell it what to write, and make sure what it does write is technically sound. I think a world where ChatGPT could replace those senior software developers may be some distance away. But the world where ChatGPT can replace juniors is already here.

It strikes me that this has to be the same in so many other professions. Just off the top of my head; Law, HR, and Copy Writing will presumably have almost no need to hire anyone junior again. ChatGPT does things in minutes that take a person hours, and it requires no training. It does make mistakes, but juniors make mistakes too, so there’s no sacrifice being made by switching to AI. It’s faster, it’s cheaper, and at the very least it doesn’t make more mistakes than a human-being. It’s all upside. All we need is experts to instruct it and validate its output.

The implications of this seem incredible to me. If companies around the world no longer need to actively refill the hopper - so to speak - with new staff, what happens when today’s experts grow old and don’t want to work anymore? Who will replace them? Unless AI has reached a level where it no longer needs any supervision at all, we’ll be left with nobody who knows enough to responsibly use it. AI will have allowed us to move incredibly fast at a much lower cost, but eventually the bill for that will come due, one way or another.

What’s troubling to me is that even if a business anticipates this problem, there will be little to no incentive for them to train junior staff anyway. The prisoner’s dilemma will come into play. Leaders will think “If we spend the money on hiring and training these juniors that we don’t really need just so they’re ready to step up when we do, someone else who didn’t bother with this kind of foresight will just come along and poach them”. I think there’s a good chance that every business will simply stop investing anywhere near as much in nurturing young talent, because they’ll struggle to financially justify it.

I hope I’m wrong about this. I would very much like to look back in a few years time from a world very different to the one I’m picturing here and feel foolish for my paranoia. But the signs are already there. I believe one consequence of this is going to be a huge increase in the competition for people with the expertise and skills that are needed to efficiently supervise (for want of a better word) AI tools, in whatever profession that may be. If the talent pool isn’t being refilled as it once was, then the people who have the knowledge needed to get the best out of AI are going to be in demand at a scale we probably can’t even imagine. If we’re not going to help build new talent, then we better be sure that our talent assessment processes allow us to make the most of the talent at our disposal already. I think the time is fast approaching where a person with skills that AI can’t replace is going to be the most valuable commodity that exists in the business world. And if your business isn’t doing everything it possibly can to identify and retain those people, there’s probably a bumpy road ahead.