Has AI Become Corporations’ New Y2K Problem?

Even though #AI #artificialintelligence has been rapidly and constantly evolving, the one issue I’ve targeted or heard frequently mentioned is reliability or the lack thereof. There’s a variety of causes associated to this problem but it should be a much more widely recognized issue that force companies to become more cautious on their adoption of AI. I do see some parallels from the infamous Y2K period that are perhaps are worth discussing.

First, I think it’s important to revisit a little bit of history in reviewing the original Y2K problem. The gist of the Y2K problem was over the format of dates for Gregorian calendars. To conserve on space (i.e. cost/money), the year format would be truncated to a two digit notation (e.g. 1985 -> 85). With the impending 2000 year, 00 would be considered indistinguishable between 1900 and 2000. Depending on the type of system implemented, the effect could be minor to catastrophic where some people predicted billions of dollars in potential losses if it wasn’t addressed.

As a result, consultants were brought in to remediate the issue. Once considered obsolete programming languages like COBOL were in vogue in a perverse manner for a small period at least on the software side (I don’t know exactly how a fix could’ve worked if it were at a lower, hardware level). Either way, a few people managed to make a reasonable living because of this situation.

But there’s a variety of causes we can enumerate here that seem to be typical of corporate behavior (maybe some even being of the anti-pattern nature):

• Build and bury – Some developers probably recognized the issue including one person who brought the problem up. But companies probably either forgot that this issue existed or the present management only cared about the immediate moment.
• Systemic rot – You can look at storage especially for something fundamental like dates as a core architectural design issue. In essence, your foundation is basic built with match sticks next to a termite hole.
• Managerial apathy – Even if the issue was known, until it was nearly too late, it feels as though management was not made aware or just how big of an issue this could have been. And if they were made aware (not in the apocalyptic tone that eventually came upon us), for however long management seemed to do nothing until the very last minute.
• Lack of maintenance/upkeep – I once heard in software that actually development is 20% of the cost while maintenance can go up to 80%. Unfortunately, software is one of those areas where maintenance tends to be quite low on the priority list for companies because I believe upper management tend to view software as a form of sunk cost. Feature development is fine for them because that’s where the money is made supposedly but maintenance doesn’t resonate.
• Lack of education – Part of managerial apathy is the symptom of a larger problem where a general lack of education on software becomes more fundamental.
• Prioritization – Issues like these aren’t huge priorities because of the prior point. As a result, they get buried because management tend to be selfish and near sighted as most times they’ll simply ask whether or not the problem directly affects them in the moment.
• Overall laziness – This isn’t just a managerial issue but one from the developers’ side in agreeing with the rollout. The compliance by developers for knowingly accepting the solution and effectively not giving a shit until things mattered is an issue of ethics on their behalf where they themselves did little to nothing to push back on what could have been a bigger disaster. Instead, as with most people, they simply “went with the crowd” apparently.

With these ideas established here, I think it’s worth comparing this new AI “revolution” in examining whether or not some of this behavior has changed and if we’re looking at potentially similar problems that will infect numerous corporations with their AI mandates.

• Build and bury – A lot of these Agentic tools generate a huge amount of code. A common tale I hear about is that the role of the developer has changed from someone who writes code to someone who reviews code as has replaced the QA person and occasionally the product person. While it might seem from an ivory glass tower that this proliferation and volume of code may appear in favor of companies, what this merely does is create a different version of burying a problem. The developer loses control over the intimate architecture and nuances by handing over the reigns almost completely to the AI. In a worse case scenario, a developer becomes apathetic from reviewing massive PRs where a clear problem may exist but the AI skips over because it only tries to solve the immediate problem its given.
• Systemic rot – Pushing out features and code become more important as companies focus more on KPIs as opposed to valuing quality. QA has been a contentious role in the past because their main function is to push back when something goes wrong. However, their purpose also was to counterbalance developers who try to cover their tracks by generating a sense of accountability. However, in recent years, I’ve seen the role of QA becoming more diminished over time where that responsibility becomes part of the developer’s job (on top of other things like DevOps and occasionally the PM) But this is a conflicting issue as there’s no checks and balances to say whether or not a feature truly behaves as intended. In turn, with the coming of AI, existing developer apathy will only increase systemic rot as less experienced people or those with nefarious agendas will incorporate AI as a replacement for themselves rather than as a side tool to augment their capabilities. We have already seen some of this showing up in various recent systemic failures. So how much more will show up as this trend continues?
• Management apathy – I don’t think this scenario changes much because people on top and the layers in between will mostly only care about KPIs and quarterly revenue to appease shareholders. Only when crucial errors start to appear (meaning that managements’ livelihoods are at stake) will this mental state switch but briefly.
• Lack of maintenance/upkeep – One trend I don’t really hear about for AI that should be one is the improvement on software maintenance. This is something very surprising to me considering that it’s a major problem that’s highly costly in the industry. Usually what I’ve seen from AI in generating code is how the AI loses context as the context window fills up. Over time, AI experiences a kind of “AI Alzheimer’s” as I describe it where its memory becomes distorted because it cannot sort out what the state of a conversation session is while the window evolves. And there have been attempts to fix this problem but it’s not a standard at this stage. The result though that I’ve seen is a constant re-solving of a problem or that the AI tries to build from scratch as opposed to using existing tools and solutions to help resolve a problem. So I think software maintenance is going to be an even bigger headache moving forward for companies.
• Lack of education – If management were clueless before on software (unless they directly wrote it themselves), then I think the lack of education especially on what AI can and cannot do is going to become an exponential issue. The one common problem about management I’ve seen (and this is really a human quality) is that they want the silver bullet/golden hammer solution, which is an anti-pattern that I commonly cite regarding management. Generally, what I see is that there’s this ethereal, imagined hopefulness in management (i.e. those incapable of accomplishing a real feat themselves outside of talking out their assholes) where they might choose a solution that they think will solve all their problems. Equally bad are the sales people that try to pitch to management rather than the developers that they have this ultimate holy grail that will make them millionaires (or more). One meme I read was hearing how some goofball would ask AI something along the lines of, “Tell me how to become rich.” Like WTF? AI is not a genie in a bottle. But this line of thought demonstrates the clear dissonance between reality and what management think AI is and it’s a very state of mind since these people are what controls the direction of these companies.
• Prioritization – In a weird state of irony, AI itself has become a priority and even a type of lifeblood that management believes needs to be elevated as a first class citizen (many times above the employees themselves). But as with the silver bullet idea, this prioritization effort is often misplaced because of the lack of education on what AI can do, should do vs what it isn’t doing. What I’ve seen happen in companies is that this AI mandate comes from above and somehow everyone has to scramble suddenly to learn how to incorporate it even though there probably aren’t actual solutions that might match the problems these people face. You could also call this another square peg round hole issue.
• Overall laziness – There’s no doubt that the acceptance of AI has in part been the result of an inherent human flaw: laziness. However, I’ve stated in the past that there are both good and bad versions of laziness (especially if we embrace the three virtues of a GREAT programmer philosophy). Most times, the laziness is more in the spirit of “I don’t want to think/do the work” as opposed to “let me build things to do the work for me”. I know there’s a split in how veteran developers view AI especially in how each individual developer may embrace or pushback on it. Some people want the old school, “In my day, I used to mine copper for the wiring!” attitude while younger developers might see AI as doing all the stupid stuff without having a solid foundation to understand what the AI is doing.
  However, the true dangerous aspect in all of this is the near blind acceptance and conformity of AI where I can squarely blame on developers. What I mean by this is that I perceive a kind of lack of ethical standing for many people here. I think AI is like another one of these fancy tools that developers lie about on their resume and simply insert it to try and bolster their own value because of selfishness, greed and any moral character as the expense of pushing humanity down the wrong path.
  Again, I don’t see AI as necessarily being a bad thing but developers as a general have a nasty tendency to shoot themselves in the foot when it comes to these things. Usually, down the line they’ll pay the price because of shortsightedness which ends up causing problems for everyone.

I think AI has this foundational issue that something like Y2K had where it’s now becoming part of the fabric for numerous companies. What I’m afraid of in part is when a disaster happens and your house of cards built by AI and/or is built upon AI crumbles kinda like that glacier in Nepal. Companies may have heard about the mass amounts of code and features being built up without understanding the actual nature where it counts. And like the build and bury problem, these issues won’t be seen because people have delegated the solutions to AI and in many cases trusted the output.

Another scary issue to think about is how cheap companies can be. I often think of the old The Towering Inferno (1974) movie when it comes to how cost effective minded people on top can be and where that can lead. In the case of the past few years, we’ve seen companies do massive layoffs and a pivot to AI where it seems on the peripherals that AI will somehow be able to replace various jobs. In the case of software, I think one thing I’ve learned is that AI is still quite immature and that you need experienced people guiding the AI every step of the way as well as requiring a core group who understands the problems and the engineering solutions rather than blindly entrusting the AI to handle all the dirty work.

However, as we’re seeing, AI is not cheap in any manner. Even a company like Microsoft has cracked down on token spending after one person used up $28k or so within 28 days. And at this stage, it’s really hard to judge the concrete returns AI can bring compared to how much is being spent. Or you have cases where someone is in the middle of a project and their usage halts because of their allotted tokens gets used up. If you decided to replace your senior developers with a bunch of trash level, offshore juniors because you’re being a Scrooge, who’s going to solve the problems that the AI creates?

I have read that some companies are starting to do an about face on the situation. Facebook/Meta is one that has admitted some regret and have been rehiring. However, to me it’s a poor look when you’ve basically admitted that you only value “quarterly revenues” as opposed to your employees in this fashion (although I think part of this was done to get more H1Bs but that’s a different story entirely)..

Nonetheless, what happens when you start having this massive dearth of domain knowledge that the AI does not know nor the junior overseas H1Bs that can’t solve your problems? Do you suddenly make a massive apology and hire back all the people you just canned for double the price because these people have managed to find a better spot? I mean isn’t this a case where these reliable codger types were cast aside and you just insulted them, thinking you had all the answers but ended up holding a bomb you have no idea how to diffuse? That’s how I see this problem. It’s more about history repeating itself even though the names and context have changed slightly.