The AI Apocalypse is Here

I looked when He opened the sixth seal, and behold, there was a great earthquake; and the sun became black as sackcloth of hair, and the moon became like blood. And the stars of heaven fell to the earth, as a fig tree drops its late figs when it is shaken by a mighty wind. Then the sky receded as a scroll when it is rolled up, and every mountain and island was moved out of its place.

Revelation 6:12-14 (NKJV)

Since ChatGPT was released, there has been no shortage of predictions, warnings, and fears about a coming AI apocalypse. In these prophecies, the outcomes are often described in ways as vivid as the quote from Revelation above. Yet it's not that apocalypse I'm concerned about; it's the one that is already here.

The word apocalypse has come to mean a cataclysm or the end of the world, but its original meaning referred to an unveiling or revealing. Hence, the "Apocalypse" of John in the Bible gets translated to "Revelation." In that book, what is being revealed is what will happen in future times, but its meaning is broader than that. This larger definition of apocalypse was brought to mind by a collection of thoughts and observations from Alberto Romero. One of these observations was:

ChatGPT is the largest social experiment in human history and it's revealing the uglier sides of both technology and humanity.¹

This framing of AI as a revealer of ugly truths helped clarify my own thinking and unveils a common thread that ties several disparate AI narratives together.

The real AI apocalypse is the revealing of the true state of reality once you remove the constraint of human beings. Humans are embodied, emotional, and relational, and our world has been built around those assumptions. AI is a digital, emotionless machine, and when it starts to replace human beings, those assumptions begin to collapse. We are feeling the effects of this earthquake, and the temptation is to blame the AI companies for causing it. While they do bear responsibility, they did not cause the underlying problems. They cannot and will not fix them; that is up to us. We must embrace the reality of what is being revealed and work to change it.

Let's start our tour of the rubble with higher education. Students were one of the earliest adopters of ChatGPT because of its remarkable ability to produce the outputs of the education process. Papers, the answer to problem sets, code examples: LLMs could do it all. While some students do use these tools to assist them in an effort to actually learn, the usage of it to cheat or avoid hard work quickly came to dominate the discourse.

Though there was a buffet of articles to read on the subject, one that really broke through was “Everyone Is Cheating Their Way Through College” by James D. Walsh in New York Magazine. Its most striking portions are the quotes from students, who understand that having ChatGPT do the work for them is a problem but continue to use it. This one is from a student Walsh calls “Wendy”:

I use AI a lot. Like, every day … and I do believe it could take away that critical-thinking part. But it’s just — now that we rely on it, we can’t really imagine living without it.²

The professors Walsh spoke to are mostly in despair or looking toward retirement. There have been many suggestions on how professors can “fix” the problem, and some are trying. In-class essays in “blue books” are making a return, as are oral exams. One “solution” that's already been abandoned by most is AI detection software, similar to previous software that checked for plagiarism. The software that has been released so far is not reliable enough to avoid false accusations, and even if it was, professors aren't sure administrators would have their backs if they started failing large numbers of students for using LLMs. Given all that, some professors have taken to an “if you can't beat them, then join them” approach. According to research from Anthropic, a meaningful number have started using AI for grading. If students are going to use AI to write a paper, why not have AI grade it?

This is the point where we have reached absurdity. Is schooling only about outputs of an educational process being evaluated and assigned a grade? Sadly, the answer is “yes” in many cases, and LLMs are revealing that sad truth. This is the real AI apocalypse in higher education, the great unveiling of the shift from learning to schooling³. The goal of learning is to acquire knowledge and develop skills to apply that knowledge. The goal of schooling is to earn grades and a degree, credentials you can use for financial or social benefit. Using LLMs for learning may be possible, but that is not what most people are doing. Using LLMs for schooling makes perfect sense. If the point is the degree, why not use LLMs to make it easier?

This is not a new problem introduced by ChatGPT. As Walsh says in the article:

The ideal of college as a place of intellectual growth, where students engage with deep, profound ideas, was gone long before ChatGPT. The combination of high costs and a winner-takes-all economy had already made it feel transactional, a means to an end … In a way, the speed and ease with which AI proved itself able to do college-level work simply exposed the rot at the core.⁴

If we could somehow remove AI from universities, this underlying problem would still be present. It would only treat a symptom, not the disease.

The next stop on our tour isn't far from the first: finding a job. The advent of the internet brought exciting possibilities to the job search. Until then it could be difficult to determine what jobs were available. Companies would post jobs in newspapers, physical job boards, or with agencies, but as a seeker it was a lot of work to sort through. It was also geographically limiting, as companies hiring in New York City generally didn't post their jobs in Pittsburgh newspapers. The promise of the internet was that the job search process would become easier and more fair. You could quickly find more open positions that matched your skill set and apply online. If you were the best candidate, you would get the job.

To some degree, the internet did live up to its promise. If you compare finding and applying for jobs on LinkedIn today to finding and applying for jobs before the internet, things have significantly improved. In my most recent job search, I could easily apply for jobs across the country at companies I'd never heard of. I even used ChatGPT to help with my résumé, cover letter, and application questions. Since the goal of the process was to maximize my odds of being the “best candidate” at a single company, why not use ChatGPT to help apply at a higher number of companies?

Apparently, I'm not the only one using this strategy. As I have seen firsthand as a hiring manager, and as has been reported in several stories, the number of applications to an individual job has increased significantly in the last few years. According to The New York Times:

The number of applications submitted on LinkedIn has surged more than 45 percent in the past year. The platform is clocking an average of 11,000 applications per minute, and generative artificial intelligence tools are contributing to the deluge.⁵

This wave of applications has made things much more difficult for recruiters and hiring managers, who are forced to sift through them to try and find qualified candidates. Some employers have even turned to AI to dig through this pile of résumés. Derek Thompson summarizes it well in his piece “Young People Face a Hiring Crisis. AI Is Making It Worse.”:

AI has turned the job application process into a bizarre high-tech blitzkrieg, where an unprecedented barrage of AI-written student applications slams into HR departments, that turn to AI-enabled recruitment technology to handle the onslaught, leaving both sides feeling overwhelmed, angry, and anxious.⁶

This is where the original promise of the internet breaks down. The real job search apocalypse is that this promise was broken from the start. Getting a job is and was about who you know, in addition to who you are. Those of you who already know this no doubt noticed my omission of networking and referrals in my earlier description of job hunting. The internet did help to some degree on this front—it did become easier to get a job without having a person at the company to advocate for you. However, many people still had to rely on personal connections. This new tsunami of AI-driven applications makes referrals even more important than before. When there are more applicants, it's much harder to get picked out of that pile without someone working on your behalf.

So what is the thread that ties these two examples together? In both cases, the underlying truths were already known by some but remained hidden to many. The constraints imposed by the embodied nature of humans helped to cover up the truth, making it appear on the surface that college was about learning or that jobs go to the most qualified applicant. But the combination of the fast adoption of LLMs and their ability to remove these constraints has torn the cover off, unveiling the truth underneath.

In higher education, students have been taking shortcuts for many years in an attempt to avoid the hard work of learning. In my college days, I never used SparkNotes, but I did use 2.2 line spacing to get to the required number of pages for a paper and size 2 font to put the entirety of my notes onto the 4x6 index card we were allowed to use on an exam. If you knew who to talk to, you could find last year’s exams to study, or previous students’ papers to draw heavy inspiration from. But these methods had both limitations and friction that ChatGPT does not. Once ChatGPT was launched, removing that friction for millions of students, the underlying desire to avoid the hard work of learning was unconstrained. It became hard to deny that learning was never the goal for a significant number of students.

It's a similar story for the job search. “Spray and pray,” submitting as many applications as you can, has been a strategy for some time. I too thought it was just a matter of finding the one company where I was clearly the best candidate, so I applied over and over again. Without LLMs, this was a process that could take hours per application, a time constraint that helped sustain the narrative that jobs went to the best candidate. But that time constraint is gone, and for many people so is that belief. It was always about a different kind of human constraint: personal connections.

Now that these truths are being revealed, what is to be done about it? The temptation is to blame the AI companies. While they do bear responsibility, they did not cause the underlying problems. We should not repeat the mistakes of social media and expect Mark Zuckerberg or Sam Altman to change the technology or fix things. Another proposed solution could be to ban AI, say, in higher education. Setting aside the practical impossibilities, the focus on degrees and grades over learning would remain. I worked for a leader at my last job who would often use the phrase “embrace reality” to encourage us to accept our current challenges and work to change them. We must do the same in this moment of revelation. Instead of focusing on the technology, focus on the ugly truths about humanity that are being revealed, and embrace them. It is only by accepting them that we can begin to address them, and we must address them as embodied, emotional, and relational human beings.

Another common thread that connects higher education and the job search is the shift away from human relationships. If we want college to be about “engaging with deep, profound ideas,” then that will require rebuilding and strengthening relationships between students and professors, relationships that have been weakened by the focus on grades and degrees. I am confident there are students and professors on every campus who want this; they must find each other, come together, and begin to restore what has been lost. Job searches cannot be individual exercises, despite what the internet and ChatGPT may make us think. If you are searching, find others who are and support each other. Strengthen your relationships, both personal and professional: get coffee with that person you haven’t talked to in a while.

These acts may seem small in the midst of the apocalypse. Surely we need a prophet who will lead us through this time of tribulation—one who can fight against the powers of this world. Yet in this case, I think it’s best to trust the wisdom of one who refused to be called a prophet, Ivan Illich:

The time of prophecy lies behind us. The only chance now lies in our taking this vocation as that of the friend. This is the way in which hope for a new society can spread. And the practice of it is not really through words but through little acts of foolish renunciation.⁷

Renouncing ChatGPT may be an example of such a foolish act, but we need to be more radical than that. Renounce schooling by gathering together to learn; renounce the isolation of the job search with connection. To endure the apocalypse, we must embrace the truths that AI is revealing and find small ways to renounce them together.