Should You Skip University?

The Future of Programming as a Profession in the Incoming Decade.

?

Recently, Mr. Elon Musk is in the news again.

About 3-4 years ago, Mr. Musk questioned the value of college education. He questioned the necessity of university education, referencing figures like Bill Gates and Steve Jobs as examples to argue that college education is not mandatory and might even be redundant.

In fact, this claim reflects classic selection bias, and it is incorrect for two main reasons:

?

?

?

?

1. Survivorship Bias - The argument ignores the countless college dropouts who did not succeed. A classic example of overcoming survivorship bias comes from WWII. Consider the hitting points on returning bombers (red points in Figure 1). At first glance, one might conclude that reinforcing the most-hit areas is the best approach. However, statistician Abraham Wald recommended adding armor to the areas with the least damage. These areas represented where bombers could NOT sustain damage and still return safely to base, highlighting the importance of considering unseen data :)

?

?

2. Pareto Principle / Outlier Bias - When examining a phenomenon, it is essential to focus on the relevant grouping. In this case, it is more accurate to evaluate the average, median, or 75th percentile individual rather than the top 1 percent who may not need college to succeed. This broader perspective provides a more realistic understanding of the value of higher education and group targeting in general. See Figure 2.

?

Evaluating the Legacy of Thiel's Fellowship: Can We?

?

In 2011, Peter Thiel offered a sensational fellowship encouraging students to drop out of college. Criteria for the 2025 cohort were published here. At the time, it generated significant publicity as well as criticism. Over a decade later, you might assume that we have sufficient data to allow a critical perspective.

Surprise, surprise—this data is not publicly available (or, if it is, I couldn’t find it). While the fellowship has highlighted some outstanding success stories among its graduates, it surprisingly does not provide comprehensive statistics. Given the highly competitive nature of the fellowship, one might assume that its recipients represent the top 0.1% who genuinely do not require a college education to succeed. Based on this, one would expect at least 30% 40% 50% success rate among its participants.

Yet, the absence of publicly shared overall success metrics is curious. If the fellowship had achieved remarkable success on a broader scale, one would expect the organization—known for its PR prowess—to readily share such data. The lack of this transparency raises questions: could it imply something less favorable?

?

?

So, Do You Need to Go to University?

?

My bold claim is that for certain types of degrees—programming, for example—it can be a risky move, exactly like learning relief printing or assembly programming three decades ago.

real creativity today (and this could evolve in the future) lies in improving video results by leveraging different algorithmic approaches

I'll try to prove this by an example, see below.

?

Improving Content Discovery on YouTube

?

YouTube has an abundance of content, but finding relevant material can be challenging due to its subpar user experience. Leveraging the capabilities of ChatGPT, I was able to automatically generate a script that:

1. Connects to YouTube using the existing API.
2. Extracts videos based on specific criteria.
3. Scores and ranks these videos using existing LLM technology with contextual relevance.
4. Saves the sorted list to a playlist on my YouTube channel under a selected name, sorted by ascending relevance.

?

This example not only highlights the power of automating workflows with AI but also emphasizes how modern tools like ChatGPT can be utilized to create functional, efficient code for everyday tasks. By automating such processes, it becomes easier to tailor content discovery to individual needs while reducing manual effort—and, in fact, by generating code automatically.

?

?

?

?

It took me only a few minutes to create the code shown in Figures 4-6. In fact, the code was created automatically according to my instructions by ChatGPT. Figure 7 shows output runs of an LLM agent that rank the retrieved videos. Figure 8 shows the created Playlist in YouTube.

This example not only highlights the power of automating workflows with AI but also emphasizes how modern tools like ChatGPT can be utilized to create functional, efficient code for everyday tasks. By automating such processes, it becomes easier to tailor content discovery to individual needs while reducing manual effort—and, in fact, by generating code automatically. However, the real creativity today (and this could evolve in the future) lies in improving video results by leveraging different algorithmic approaches. To achieve this, a solid foundation—best acquired through college and the right courses—is crucial.

So, bambino, do you still want to learn Computer Science? :)

?