Academic inflation vs. artificial intelligence (AI vs. AI)

December 2024

Over my 25-year career, I have witnessed and observed the phenomenon of academic inflation, which I define as the value of a degree only ever decreasing from the moment it is conferred. There are three main mechanisms how this happens.

First, the democratization of higher education. Prima facie, this is a desirable trend, as it gives more people an opportunity to earn a higher income. However, it also leads to tertiary (post high school) degrees being less scarce, and therefore less valuable for everyone. Case in point: enrolment has more than doubled at academic institutions worldwide between 2000 and 2014 (faster than population), and in the UK alone, universities have increased the overall number of diplomas by five times since 1990 (Morgan, 2021). Anecdotally, when I hire at my firm, I receive tons of CVs that have become difficult to tell apart, because everybody seems to have a Masters degree — and that was not the case when I joined the workforce.

Second is the half-life of knowledge (by analogy with the radioactive decay process), by which I mean that knowledge acquired through a degree becomes obsolete over time. I earned my MBA in 2000, and much of what I learned back then is either stale or missing much of what constitutes a modern curriculum. This is, obviously, especially true in fast-moving fields related to technology. Back in 2000, the internet was still somewhat novel; you had to sit down at a wired desktop computer to access it, because there were no Wi-Fi nor smart phones. The concept of ownership of digital assets was completely foreign to everyone, and blockchains were still a decade away from being ideated. The SaaS model was yet to transform the software industry. Etc.

Third, degree holders forget the knowledge they acquired due to natural memory attrition. I probably forgot most of what I learned at university 25 years ago; at least the many parts that I do not use actively. From an employer's perspective, this makes a recent degree more attractive than an older one, or at least erodes the advantage that an experienced candidate has over a greener one. This is, in part, what motivated me to pursue a PhD later in life — I felt the need to update my "mental software" to remain competitive. I've also pursued a bunch of professional certifications along the way to maintain currency in my field.

These trends have deep ramifications for the job market. Employers respond to the surplus of qualified graduates by increasing their demand for academic qualifications for a given level of employment over time (Fuller & Raman, 2017; Morgan, 2021). College graduates start taking high-school jobs, and overall economic efficiency decreases as students spend extra years pursuing degrees that they will not fully utilize (Yi & McMurtrey, 2013). Of course, this becomes a self-reinforcing feedback loop, as people seek even more and higher degrees to differentiate themselves further, even as the disconnect with the actual demands from the job grows wider.

Today's OpenAI announcement of o3, its next-generation reasoning model that is leaps and bounds more performant than o1, marks another step toward AI being more capable than graduates. o3 scored 96.7% on AIME 2024, a test of 15 challenging math questions that was previously unpublished (meaning the model could not have been trained on it). It also beat human PhD-level experts with a score of 87.7% at the Graduate-Level Google-Proof Q&A Benchmark (GPQA Diamond) MCQ test written by domain experts in biology, physics, and chemistry. And reasoning models are just warming up.

So what does it mean for the future of education? An unstoppable force (AI trending toward superhuman performance in virtually all intellectual tasks) is about to meet an immovable object (people seeking more degrees to stand out). This will lead to an even greater opportunity cost for people who've invested time and money into tertiary degrees, only to find that their expected academic moat is being leaped over by reasoning agents. This is not just yet another "AI will replace workers" doom post — what I mean to convey is that new entrants to the workforce are throwing more money than ever at what appears to be an ineffective job security insurance policy. Knowledge is being commoditized faster than the secular trend of academic inflation is changing. I predict that know-how will soon be more important than knowledge, and that those workers who are less educated but proficient in using reasoning models to augment their capabilities will win over those who still try to outsmart their peers through sheer educational attainment.

References