In the wake of recent publicity around machine learning models, there’s a lot of discussion about AI replacing humans in both creative and technical fields. From Lalaland.ai as a replacement for photographers, ChatGPT edging out writers and coders, and Midjourney supplanting artists the hype and the fear are both off the charts.
Social media platforms and news headlines are full of questions about whether people will be replaced by AI, with arguments on both sides. Squabbles are underway between lobbying groups (or, if you prefer, think tanks) about whether research should be paused, and why it should or shouldn’t.
When ideologically opposed lobbying groups/think tanks fight, there’s rarely any thought for those caught in the middle. All collateral damage is acceptable in pursuit of their goals.
Some groups talk about pausing research because of imagined outlandish apocalyptic scenarios (something I know a little about). Others are more reasonably concerned about well-established and proven issues of bias such as encoding racism or sexisminto the models in opaque ways. Still, more people are asking questions about copyright which, if we’re being realistic, are unlikely to be resolved until decades too late to be useful.
But I want to focus on one very narrow piece of the debate. There’s broad confidence that AI tools will not replace experts in their field, as the deep learning and understanding required are beyond the capabilities of the tools today and are likely to be in the future. This intuitive mastery may be beyond the glass ceiling of AI, as I discussed with FibreTigre in a recent interview. Even if it isn’t, the big innovation with ChatGPT was more about opening access to the tools than a revolutionary step. We have time before machines catch up to those who have developed expertise in their trade.
No One Starts As An Expert
And here lies the problem, and for me one of the biggest concerns (aside from the horrendous examples of algorithmic bias). While there’s a way to go before AI starts replacing those at the top of their craft, what about those entering it?
There is definitely an argument that these tools are increasing accessibility to creation. I guarantee the graphics for this article would not be in my non-existent budget without the help of AI. For now, I’m going to put the intellectual property issues aside, as they are a huge and important debate in themselves (though not one I think will be resolved in any useful length of time). Instead, I want to look at the problem of raising the barriers to building skills as AI competes more with beginners.
Many small and budget-conscious organisations are already using ChatGPT to replace copywriters, where before they’d be going to freelancers just getting started. Levi’s has recently announced a partnership with Lalaland.ai. While they have walked back and clarified the initial announcement to say they will not be reducing their use of live models, it’s important to note that it means they will not be increasing them as they might otherwise.
Entrants to the industry, newcomers, amateurs, and people trying to build a career from the start, are the ones who are vulnerable to replacement by AI as it stands now. While the experts may be confident in their uniqueness (for now), it’s those who are trying to build that expertise who are likely to see the biggest impact on their opportunities. And this is a problem for everyone.
Are more and more lower-cost beginner freelancer jobs going to be replaced by AI, leaving freelancers trying to build expertise desperately scrabbling to snatch crumbs from the corporate bot’s tables?
There’s a lifecycle to building a skillset, from art to writing, from coding to photography. No one starts as an expert. The best education in the world doesn’t prepare someone to step in at the top of their game immediately. When people trying to build this expertise are competing against free AI, the issues we have of companies wanting to pay ‘in exposure’ are only going to increase. The barrier to making a living through a craft is going to rise as a result, without positive action to make sure these opportunities still exist.
I’m not worried in the least about AI taking my job, as I’ll be long retired (and probably deceased) before that point. I am much more worried about those trying to enter fields now, or tomorrow, and whether they will ever have the opportunity to reach the same point.
The Skills Gap
The field I work in, security, already has a well-recognised problem of a skills gap. This gap isn’t at the entry level, it’s a little higher up and it means that opportunities to enter the field are not nearly as plentiful as they should be. While it’s a great field to work in, and we have a desperate need for people, we don’t have sufficient numbers to develop the people we need.
My fear is that AI is going to lead the other way. With those trying to learn being easily and cheaply replaced by AI, the supply of human talent able to develop could dry up. As with the issue of unpaid internships in the US (and elsewhere), anyone wanting to develop along a path where a machine replacement is available may only have the opportunity if they come from a place of privilege where payment in exposure will not cause them to starve.
A mass of machine-manufactured mediocrity.
If that does happen we will be looking at a different skills gap, as the supply of masters at their craft dwindles away. Instead of unique human-generated works we could end up, as we have done in so many other areas, with a slow slide into machine-manufactured mediocrity.
This article was originally published by James Bore on Hackernoon.