Because I use the Tor browser when I go to duck.ai, by my count that means I’m wearing at least four condoms when I use AI. I feel pretty confident that I won’t catch the AI virus, but you can never be too sure (or too safe).
Introduction
I’m acutely aware that there are many, many people who are forced to use AI at work and others who have good reasons why they use AI. For instance, there could be wonderful advancements thanks to AI in medical sectors that are helping people live longer, happier and healthier lives. I’m definitely all for that.
This article is definitively not about that. It’s specific to the tech industry and primarily concerned with the act-ask-report designed LLMs that generate code. The AI proponents and boosters that are engaged in its advocacy are the targets of this polemic article.
There are several reasons why this is particularly grating to me.
-
It greatly angers me when there is advocacy for an adoption of a technology that will put skilled people out of work. Often, this advocacy is done with a cavalier attitude about how “change is inevitable.”
-
I am very skeptical that the code that is generated and the companies that are deploying it as important as they think they are, so the resultant unemployment can’t even be pointed to as the unfortunate consequence of a greater public good. We’re not talking about curing cancer.
-
Education and computer literacy is more important then your product.
AI advocacy and its advocates are boosting these harmful effects with every excited post and video they produce. Hyperbole? Here’s more; the people that are doing it are displaying an almost pathological level of indifference. I’d argue that whether they do so knowingly or unknowingly is both condemnable – the former shows a callousness to anything outside of self, and the latter demonstrates a disturbing incuriousness that is indefensible.
Hopefully, the uncritical embrace of AI that we’re currently experiencing would have seen more hurdles to mass adoption in the recent past. I argue that there are several reasons why the ground for that adoption has proven to be much more fertile today after years of the erosion of voices that would have objected to it.
Just to keep things simple, when AI is mentioned in this article, I am referring to the LLMs that are directly advertised to the tech community, which are the code generation models I referred to above.
We have been primed to prioritize and demand convenience over everything else
There is a surety in life that is demonstrated over and over again: given the option, humans will choose convenience over inconvenience. They do this even when shown the negative consequences of this choice. Taking the easy way out and cutting corners over hard work and dedication is even celebrated and lauded in our present culture.
The latest culmination of this has been the massive adoption of AI. Some programmers, especially those who have just started their career, may legitimately claim that it is not worth spending years, if not decades, learning a subject (or subjects) deeply when they see that instead they can just ask a model. Sadly, that behavior is then rewarded.
Given the demotivating conditions we find ourselves in, it certainly is more convenient to ask an LLM than it is to put in the time to become a subject-matter expert. A sensible counter to that would be that the answer you’re given could be wrong, but how would you know? Increasingly, it seems that this is a concern that is not seen as a priority, and as a good employee and team player, it is no longer a priority to you.
The more that companies use AI, the more they are dependent upon the whims of a few AI companies. First, they were locked into a cloud company, and now they’ve also chosen vendor lock-in to an AI company. This has already shown itself to be an incredibly short-sighted and poor decision, as demonstrated earlier this year by Claude going offline with large numbers of users affected. This is so idiotic it is beyond belief; “engineers”, now having to rely upon an agent to do even the most basic tasks, found themselves unable to get any work done because their AI company of choice shit the bed. They are now unable to work locally, thanks to the convenience of the cloud.
Would the Snowden disclosures even make an impact today?
Engineering value has been redefined to be whatever the business wants
I noticed a shift about ten years ago in engineering teams that I found disturbing. Specifically, whenever “value” was mentioned, it began to be framed in business terms, not in engineering terms. On the surface, this seems innocuous and even commonsensical. Delivering “value to the customer” is how we all get paid. And, what that really means is just get more customers. If that happens to be what they want, well that’s a nice coincidence, isn’t it?
But, this was a pretty seismic change. Naturally, engineering has its own values, and they will (and should) be different from those of sales and marketing. These values take time, care, maintainability and iterative review.
My own view is that the religious adoption of Agile across the industry has played a large part in slowly bending engineering’s values to align with those of business, to the detriment of both. Something is now only valuable if it can be done quickly and makes money.
By redefining what “value” means to be profit-motivated, young programmers can reasonably wonder if it is even in their best interests to focus on and commit to intellectual study and devotion to deeply challenging subjects. It’s just going to be more convenient to generate all the code, and there’s no other alternative now but to buy those subscriptions and lock-in to big AI.
This should be viewed as infeasible and a chance for engineering to stand up and defend its values, but the “well, what are you gonna do?” attitude that I generally see about it is anecdotal evidence that business gets what business wants. Everyone wants to be a good boy employee and a team player.
To be fair, I believe that programmers themselves and engineering departments are partially to blame for the subservient position in which we find ourselves. As mentioned, it’s a problem when the interests of engineering seem to align with those of the business interests most of the time. The friction that is a result of different values and priorities is a healthy thing and can lead to discussions that help strengthen the company overall.
We have been primed to distrust expertise
America has always had a dark underbelly of anti-intellectualism. This has been brutally accelerated by the current administration in the United States. Political candidates and appointees and jurists are extremely unqualified for the office for which they are running and are elected or appointed regardless. Things that worked and were taken for granted begin to break, and there is no Plan B. When this type of dystopian leadership exists at the top, it will have an influence on everything below it.
Given this acceptance and embrace of a distrust of and outright mockery of expertise from the top down, it is no wonder that a technology such as AI would be massively adopted. In the past, I believe there would have been more critical thinking applied and more questions asked of the self-empowered few that have unleashed AI on us. Actually, given how toothless the U.S. media is, perhaps not.
The more ready we are to distrust expertise, the more easily we will be beguiled and bamboozled by bullshit. We become easily gullible and will fall for anything. Without the ability to rely on critical thinking and common sense, we are desperate to embrace any silver bullet.
Senior programmers aren’t who they think they are
Something that I’d like to address that doesn’t seem to get challenged are the programmers who claim to be “concerned” with AI when it comes to more junior programmers but aren’t worried about their own use, ostensibly because they have arrived.
There aren’t a lot of truisms in life, but one I can get behind is that you don’t know what you don’t know. Do you really believe that you have expert-level knowledge in all the areas of the code that has been generated by a model? You may believe that “you’re all good” just reviewing the sloppy output of AI, but your learning and growth will atrophy just as surely as anyone else who relies on AI to produce code. Reviewing code is just not enough. You need to write it.
There are many anecdotal cautionary tales of programmers losing their skills through over-reliance on AI. This is a real problem fueled by over-confidence in one’s own skillset.
Moreover, there is a responsibility to mentor and train the younger generation. This is now in danger, because these overconfident senior programmers are losing their skills with every passing day that they outsource their brains to AI.
We need serious programmers. Serious programmers are people who care about the importance of understanding what they produce and its tradeoffs. People who are interested in continuing to learn and getting better at what they love. People who care about passing on that hard-earned knowledge to a younger generation. People who realize that knowledge is not a destination.
AI advocacy is a moral failure
Even if AI gets to the point of generating flawless code – code that never breaks, is security-minded and is context-aware of the rest of the codebase – I still say “so what”? Who cares? What is the benefit to you of generating something that you haven’t written? Can you feel a sense of pride in that? How can you know what it does? Maybe you now think you’re a “10x engineer”, but that questionable “productivity” is merely lining the pockets of other people who are just looking to automate your job anyway. Are they paying you a 10x salary?
Speaking of salaries, there are many people who are not receiving one, through no fault of their own. Many of these job losses have been directly attributed to AI (to be fair, many dispute this, instead blaming the job cuts as a “correction” on the over-zealous hiring practices seen during COVID). This makes this enthusiastic AI advocacy even more gross and unwholesome.
There are many more terrible reasons why AI should not be used. AI boosters know these reasons and boost anyway. This should be considered unacceptable considering the harm that is inflicted upon all of us, these short-sighted people included.
Not using AI is a public good. It supports:
- working people
- young people looking for their first jobs
- education
As well as other areas that are important but not mentioned in this article:
- the environment
- mental health
- privacy
Conclusion
You’d have to be an unserious person with a weak moral compass to choose AI advocacy as its promotion inflicts great harm on society – whether through cynical ambition and greed or incurious ignorance. Perhaps, these boosters were not using the proper protection when using AI, and, as a result, we’re now subjected to their endless, predictable and boring hot takes. Either way, they should not be seen as authorities but as shameless promoters for the ultra-wealthy whose ranks they hope to join at the expense of the rest of us.
Incidentally, I did take the time to use OpenCode a bit and even created my own agent, agent-pete. I thought it would make sense to build a project that uses at least some of the popular features of agents and to code it myself. It is, after all, important to investigate the other side of one’s argument and be open to change, and I felt the best way to do that was to get my hands dirty.
Even though I did learn more about the tools that people are using to build their million-dollar companies, I came away from it feeling like I had wasted my time and that I should have spent it doing something useful, like learning a new language or learning something new about the Linux kernel. After all, as there is so much cool stuff to learn about programming and systems, and a single lifetime just isn’t enough time to learn deeply about all of the things that interest me.