Personal Blog

Digital Divide

As technology progresses, there is a noticeable gap between those on the forefront of development and those who are stuck with little access to technology. Around the world, billions of people have little to no experience with technology, leaving them unable to connect to the rest of the modern world. This phenomenon is known as the Digital Divide, and it is a big problem in the world today.

The most obvious example of the Digital Divide is within third-world countries, such as those in Africa. Many people in countries such as Nigeria and Ethiopia struggle with access to technology such as cell phones, as well as a consistent access to electricity. This leads to a lack of both access and knowledge about technology. This can make it hard for some people to get jobs in fields related to technology due to a lack of exposure to it.

Additionally, people with little access to technology can be cut off from some of the benefits of the internet, such as instant communication. There have been some individuals who have worked to help this problem by educating members of their communities about technology and the internet, but the actions of single people can only help somewhat. To fully solve this problem, governments and countries will have to work together to construct new infrastructure and invest in educational programs to help their citizens more knowledgeable about technology. Although this problem may not be solved for some time, hopefully it can be greatly reduced within our lifetimes.

The CS Job Market

One interesting phenomenon of computer science is how the job market has deteriorated in the past 10 years or so. During the mid 2010s, computer science was seen as a gold mine for those who wanted an easy career that would make a lot of money. At the time, C.S. was a new field with high demand, as technology was advancing quickly. This perceived golden opportunity led to a large increase in the number of computer science students from 2015-2020, eager to fill the supposed “easy jobs”.

Obviously, however, there weren’t any unlimited number of high-paying jobs in the market. Inevitably, the amount of new cs graduates looking for jobs met and then surpassed the demand for workers. Suddenly, the market went from being an open frontier to a competitive nightmare, with internships with only one opening getting hundreds or thousands of applicants. Today, the market is now seen as the worst part of working in C.S. If you look online, you’ll see hundreds of horror stories about cs grads who graduated college with perfect grades, and still were unable to get jobs.

This effect is similar to a speculative frenzy, where an asset’s predicted value is inflated far beyond its real value due to public attention. This is exactly what happened in computer science: public opinion was shifted to believe that there was more potential in the job market than there actually was, leading to oversaturation. Potentially, in the coming decade or two, the reverse will happen, where there is very little growth in the job market due to perceived oversaturation, leading to a lack of jobs. To me, It is very interesting how one of the most attractive parts of this field has become one of its largest problems in such a small amount of time.

AI in Computer Science

AI models like ChatGPT are LLMS (Large Language Models). An LLM functions by first being trained on a large amount of data (VERY large, with some models today essentially being trained on all of the data in the entire internet), then making predictions about what the next word in a sequence of words could be, based on that training data. By repeating this process >over and over, the AI can generate an entire paragraph one word at a time while not even really understanding what it’s writing. In this way, it can produce a string of text that seems like it had lots of thought into it, but is really just an approximation of what high level text sounds like.

While this process of text generation is well and good if you’re asking an AI to write an essay or answer an easy question, things start getting shaky when you ask it to manage your bank account. The issue that LLMS face is that you can’t rely on what the average response to something is when you want a very precise, specific result. If the AI model has even a 10% chance of just making an illogical choice about something important because it’s unable to actually think about what choice makes sense, that makes it almost impossible to use it for anything important. This is why AI is currently getting stuck in the “helpful assistant” phase: It’s just consistent enough to reliably automate some of the more menial parts of office jobs (scheduling meetings, rephrasing emails), but it’s nowhere near competent enough to effectively do complicated and risky work. Until AI technology advances to the point where LLMs can be trained to reason about what it’s doing, AI will likely not replace any jobs that require complex thinking.