Well, it’s vise grips for pliers, and pliers for a wrench A wrench for a hammer, hammers everything else It just don’t seem to make much difference I sure do like him but he’s hard on equipment
- Corb Lund, Hard on Equipment (Tool for the Job)
Corb Lund’s music is self-described as “agricultural tragic”, waxing poetic a lot on things that are more recognizable to a rancher than a technogeek. However, there are common themes – stubbornness of men, being neighbourly, the weather – that relate to nearly everyone. Hard on Equipment talks about That Guy – we all know one – who just forges ahead without thinking very hard.
With all the push that’s going on with AI these days, this song rings louder more and more.
Artificial Intelligence (really, Augmented Intelligence) is everywhere, being wedged into everything, because Product Owners out there are convinced that everyone wants it, that’s some panacea that’s going to solve all the world’s problems. We’ve already seen far too many examples of AI slop that strongly suggests that a lot of these uses of AI are solutions aching for a problem.
AI is a tool, a wrench, if you will. And there are many wrenches, in different colours and sizes, all attempting to solve a problem. Some try to be an adjustable wrench, trying to address many different uses. But in the end, they’re still a wrench, and sometimes, you just don’t have a nut or bolt that needs tightening.
Worse still is the person wielding said wrench – directly or indirectly – and demanding it do more than it’s capable of. Too often, we’re seeing uninformed and ignorant perspectives of what AI is actually capable of, and (deliberately) avoiding the risks of using it appropriately.
Several companies have announced layoffs due to AI. All in the last couple of days. And that’s not counting all the layoffs in previous months.
Why? Because apparently people are expensive. Congratulations folks, capitalism has made you a commodity. You’re a widget, an appliance, a tool.
There’s only one difference: AI can’t create. Not truly. And it certainly can’t adequately assess risk or identify potential problems for proceeding on a decision.
Back in February, AWS suffered a serious outage of its services in US-EAST-1, it’s most important data center. Although Amazon tried very hard to say otherwise, the news suggested that the outage was due to using agentic AI as part of the management of infrastructure. Github, one of the major code management and CI/CD platforms, has been pressed by owner Microsoft to force more AI (specifically Copilot) into near every corner of its operations. The result has been greater reliability issues, which is starting to drive long-time users, especially heavy-demand users, to other services. But even those options are limited as one of Github’s biggest competitors – GitLab – is now laying off staff in favour of AI. And Cloudflare, a darling of network services for many years, has unapologetically favoured AI over its own people.
Business thinks this is being efficient. AI is “good enough” to manage the needs and can compensate for fewer developer hands and/or very expensive senior developers. What happens next, however, is not cost savings or even efficiency gains. And in technology fields, business also loses the single most important part of its future: innovation.
AI is incapable of innovation. AI is incapable of pure creation. Even a three-year old has more creativity with a crayon and a blank piece of paper than any AI in use. Business is not forwarded by AI on its own – it requires human insight and direction.
But businesses are pressed by investors who are inspired by news articles expounding the unbridled and unlimited capabilities of AI. All of them miss that, just a like a car, AI can’t go anywhere without being driven by a human. So when a business decides to sacrifice humans for a machine, there is a very high risk of poor returns.
Here’s why:
- Company decides that AI investment is critical to success
- Cost analysis determines that a percentage of staff can be dropped, as AI can handle the gap
- Layoffs include a variety of roles deemed redundant
- Morale suffers
- Company forces employees to adopt AI, often without direction, structure, or training
- Senior staff (particularly engineers) become unhappy because AI removes them from the parts of the job they loved doing, and start to leave
- Morale suffers more
- Junior staff are pressed to make up the difference with AI, but lack the experience and knowledge to guide the AI appropriately, producing sub-standard output
And that’s just the start of the downhill slope.
One of the biggest challenges that companies face is adoption. It’s no different than forcing someone to use Microsoft Teams instead of Zoom – its a different tool and people are resistant to change if they see no particular value to them in their day-to-day activities. “Change for Change’s Sake” is a concern for many – if it ain’t broke, don’t fix it, right?
AI can be supportive, it can help with simple problems and improve one’s efficiency. But for some roles, AI isn’t just a detriment, it’s an insult, if not an outright execution of that role’s purpose. Take programmers, for example. Most people who aren’t writing software just think of these people are troglodytes who sit in dark rooms. (Movies tend to portray us as more Dennis Nedry than Neo.) But at their heart, all programmers are out to solve problems.
(This, incidentally, is one reason where there are so many different programming languages. It’s not that any of them are particular awful – there are outdated ones, to be sure – but programmers often land in situations where to solve one problem, they have to solve others. Thus different languages, libraries, operating systems, interfaces, networking protocols, and so forth.)
Now give that problem solving to an AI. Senior engineers will have difficulty working with an AI because it’s not intelligent enough to be considered a “pair programmer”, and progress will be slower. Junior engineers simply don’t have the knowledge to ask for the right thing. The expectation is that using AI will accelerate programmers’ abilities.
Vibe coding is a thing, make no mistake. Who exactly is doing vibe coding is a bigger question, which is immediately followed with: and what is their skill level. Can a senior engineer with 15 years of software development experience vibe code something good? Most likely, yes. (And they will probably hate themselves afterwards.) A junior person? Or someone who has never programmed before? They might successfully code something, but then we end up with the Vibe Code Repairman.
An AI – any AI – is a tool, nothing more. It should be considered as an advantage, not a replacement, for those who understand how the job was done before there was AI. It also helps old farts like me who have lost some of their skill due to having to focus on other needs – I can still read code, it just takes me longer to write it. But it is a tool, no different than the right-sized 10mm wrench for building a simple extension to ensure the correct serialized data is exposed for a form export (literally what I’ve used an AI to do, by the way).
I worry for our future, where rely on technology that cannot deliver on the expectations. We are facing a world where the unskilled are directing the unenlightened to ask the unwise to solve the unknowable. I shudder to think of what the first truly man-made disaster will be as a result.
