Right now, there's a mixture of excitement and unease among tech workers about the potential of AI bot ChatGPT and what its evolution could mean for work now and in the future. While some are impressed by its features, others are worried about it possibly taking over their jobs in the not-so-distant future.
I'm one of those people who tend to always imagine the worst-case scenario of any situation, so I can understand the anxiety and apprehension some people are currently feeling. The job insecurity sparked by ChatGPT should challenge us to take a closer look at it and reflect on its possible disruption of the tech labor market as we know it.
But before we do that, let’s take a step back into history and briefly explore how industrial and technological revolutions in the past have affected the job market. After all, they say the best way to predict the future is by studying the past.
A look back in time
According to Britannica, the industrialization and technology boom started in the 1800s specifically in Britain and spread to other parts of the world. The horse and carriage business gave way to the automobile revolution that produced cars powered by steam engines. These were gradually replaced by vehicles built with internal combustion engines(ICEs). Even now, electric motors are supplanting the internal combustion engine with the invention of electric cars that aim to decrease the effect of ICE's use of fossil fuels on the climate. The early 1960s also marked the beginning of significant technological advancements in automation and robotics in factory operations. These inventions steadily took over manual labor in manufacturing, agriculture, and other industries.
Recently, some major players in the hospitality industry, like hotel franchises, have begun integrating robots into their operations. This move has resulted in portions of their staff being displaced by robots that specialize in performing repetitive tasks like cleaning, deliveries, and even front desk assistance. Ride-hailing apps have replaced traditional taxis and Netflix, Hulu, etc, have eliminated DVD and video rental services.
So history proves that this trend has always existed. Human beings are ambitious and will constantly innovate, and these innovations will always cost some people their jobs and businesses. It is inescapable and inevitable.
Given this knowledge, what attitude/mindset should we, as tech professionals, have toward an invention like ChatGPT? I believe our answers lie in understanding ChatGPT, its capabilities, and why it was invented in the first place.
ChatGPT-What it is, What it does and What it’s not
ChatGPT is an AI chatbot model recently introduced by the OpenAI organization on November 30 and is available on their website.
It is a large language model trained by two machine learning techniques on a vast amount of text data and code(yes, that’s correct, actual programming source code), curated from different sources from before Q4 2021. It specializes in human dialogue and can respond to the questions asked by users in a natural, human-like manner.
It is specifically designed to handle conversational text and can understand the natural ways human beings communicate. Once it takes in your text input, it’s able to generate remarkably detailed human-like answers in the form of text or code. It also has a moderation API that can block certain types of unsafe input.
At first glance, it looks like a really advanced search engine because of its ability to answer questions and provide information about a wide variety of topics. But a closer look reveals it’s so much more than that.
What it does/Things it can do
I started my ChatGPT experiment with some relatively simple input. I asked it to tell me who the past Ghanaian presidents have been and it was able to do that easily.
I read on Twitter that it is pretty helpful with planning and scheduling, so I asked it to help me come up with an efficient exercise schedule for a busy software engineer, and it did that pretty well. Here is a snippet of that.
Sure, here are some suggestions for an efficient exercise schedule for a busy software engineer:
Break up your workouts into shorter, more frequent sessions. If you don't have time for a long workout, try breaking it up into two or three shorter sessions throughout the day.
Exercise in the morning. Many people find it easier to fit in their workouts first thing in the morning, before the demands of the day take over.
Make use of your lunch break. If you have a longer lunch break, use that time to go for a hike or hit the gym.
Don't be afraid to ask for help. If you're struggling to fit in exercise, consider hiring a personal trainer or joining a group fitness class to help motivate and structure your workouts.
Remember, it's important to be consistent with your workouts and to listen to your body. Don't push yourself too hard and make sure to take rest days when you need them.
Then I attempted to test its coding prowess. I asked it to solve the popular Two Sum Leetcode problem in Java for me and the solution it gave me was surprisingly correct. It used an efficient well-known technique and was also able to give me the variant of the solution in Python and Javascript. The image below is a partial view of the solution in Java.
I finally asked it to suggest some fun activities for the upcoming Christmas holidays and it came up with some cool ideas.
Pretty cool, right? These examples are just a scratch on the surface of what ChatGPT can do. People on social media are presently using it for various tasks ranging from planning their schedules to writing love poetry 😅.
In fact, it is being used so heavily that OpenAI has had to constantly scale its servers to handle the burgeoning volume of requests. Every time I visit the site, I see this banner message. "We're experiencing exceptionally high demand. Please hang tight as we work on scaling our systems".
Its founders say their aim for inventing ChatGPT and its sister bots is to create AI systems that are more natural and safe to interact with, and I think they are onto something here. ChatGPT is not the first AI model/project created by OpenAI. It’s an improvement of another AI bot called Instruct GPT. Unlike ChatGPT, it is not trained for human dialogue but for instruction processing.
What it is not
ChatGPT is not a calculator. Some users say it calculates incorrectly when asked to compute complex calculus problems. But it is not WolframAlpha. It was not built to perform math computations.
It is not an AI art generator. Others have tried to feed it text to generate AI art and its response was “I'm sorry, I am a text-based AI and am not able to draw pictures. Is there anything else I may help you with?”. You can try DALL-E if you are interested in AI-generated art. It's also an AI model from Open AI.
It is not a prediction or fortune-telling application. It cannot answer open-ended questions like where do you see yourself in 5 years or when is the world going to end.
In short, ChatGPTs' ability to perform particular tasks or offer answers to certain categories of questions is constrained by the dataset on which it was trained.
ChatGPT's Limitations
Like any new AI prototype, ChatGPT has its own imperfections ranging from phrase overuse to exhibiting biased behavior. One of its limitations consistently harped on by most users is factual inaccuracy. Even though it has been trained on a dataset that is quite wide and diverse, its responses to certain queries are not exactly correct, but even this issue has been worked on to an extent.
Currently, when you ask it about topics it has no data for or questions it does not understand, rather than fabricating a seemingly correct answer for you, it tells you it does not have data set to give you an answer. I consider that a slight improvement on its previous behavior, but the fact is ChatGPT is still largely restricted by its data set. It does not have access to information about current events and cannot tell if the information it is serving you with is actually accurate.
There are also some problems with how it interacts with human beings. It lacks the intelligence to prompt users to clarify themselves when they ask ambiguous questions. Its inability to convey human emotions in its answers results in some of the writing it produces sounding flat and monotone. It can also be overly verbose and is prone to overusing the same phrase multiple times in the same paragraph.
But some of this behavior is expected because the ChatGPT bot itself says it's useful for simple queries and not built for excessively complex or sensitive tasks, which is a reasonable explanation.
Eventually, OpenAI will resolve the hurdle of factual inaccuracy and the other issues with the new algorithms, cheaper compute, and greater data availability. It is just a matter of time.
So should tech workers feel threatened by AI like ChatGPT?
To be honest, it looks like ChatGPT is going to be a real game-changer. While it's not the first AI bot to be invented, it's undoubtedly one of the most intelligent and has shown companies they can conveniently harness AI to make work like coding and writing extremely efficient.
Some users have even started to say they are willing to pay if ChatGPT is monetized to provide more advanced services. Interestingly, this AI as a service concept has already been started with some AI art generation apps and websites like Midjourney 4 and Lensa AI that charge customers to provide premium, high-res art, and photos. It looks like Open AI has struck a goldmine with ChatGPT, and many companies will soon begin adapting it for their business needs.
Now, as if the job insecurity being caused by ChatGPT is not enough, I found out Google's DeepMind AI program AlphaCode has started competing with human programmers in coding contests. Things are about to get really interesting in tech because, fortunately, or unfortunately, inventions like ChatGPT and AlphaCode can be further developed to easily outperform human beings in technical work.
Over time, a lesser number of employees may be needed at work since AI may be doing the majority of the work anyway. Tech organizations will likely hire fewer programmers if advanced coding AI tools or low-code AI programs are invented. In the publishing world also, AI is likely to replace certain lower-level jobs. The possibilities are limitless, especially with companies downsizing to save money because of the unstable financial markets.
But!, even though AI programs have the potential to do some really impressive work, they are still not capable of the same level of creativity, intuition, and problem-solving abilities that humans possess. They currently lack the ability to understand the context or the subtleties of human communication and can, therefore, not completely replace humans.
I don't think AI can totally beat the software craftsmanship exhibited by experienced and exceptional software engineers and the unique writing styles that distinguish the voice of different writers. That being said, we still need to keep an eye out for it. It is a rapidly evolving space, and if history is anything to go by, it is clear what it can do to tech jobs. From what I know, it seems to have begun competing with some artists in the art community with the rise of AI-generated art programs.
For now, we should view it as a powerful tool that can make certain aspects of our work considerably easier, but my advice to fellow tech workers and myself will be this. Let's constantly upskill and reskill to keep ourselves relevant and sharp so we can easily surf the waves of the impending AI storm.