and hear about AI. “AI is going to solve the hardest problems and change our world”, “AI can make you 10x more productive”, “AI is going to end the world,” are common headlines these days.

There is increasing discussion that Artificial General Intelligence (AGI), an AI that can match or surpass human capabilities in a wide variety of cognitive tasks, is just around the corner. Responses are varied: from being skeptical about these claims, optimistic that this may solve many challenging problems, to worrying that this may take jobs or even dominate the human race. 

Although AGI is a genuine problem to think about, I think that it is more important for the general population to think about more immediate but worrying trends that we are currently seeing as a consequence of AI. 

In the last year or so, there have been a number of disturbing reports of the results of over-reliance or misuse of AI. This is prominently seen in school-college going kids and young professionals, who are most likely to use AI extensively. There are numerous reports of students using AI to solve homework or exams, cheat at interviews, or professionals using it to produce unverified code or reports. In several cases, teachers have immediately identified the cheating in homeworks, and sloppy/error-filled code or documents submitted at the workplace without quality control have led to professional disasters.

A number of news articles and videos have highlighted a decline in reading, writing, comprehension and critical thinking skills in students as a result of over-dependence on AI. 

Several anecdotal reports by professional programmers have noted a drop in programming skills after using AI, a recent paper showed decline in creativity and cognition among users, and a recent MIT article highlighted the hidden costs of using AI for coding.

We also see increased AI-generated content on social media, termed “AI slop“. This, combined with the addictive nature of social media leads to people spending more time on these platforms than ever before, which also leads to cognitive decline. 

The pressing matter at hand is therefore not the consequences of Artificial General Intelligence but what I will call Reduced General Intelligence (RGI).

Why is this a problem

To be relevant in today’s world, you need to have critical thinking skills and be able to solve problems under a deadline. To be valuable, you need to produce work of value. If you are heavily reliant on AI, then the quality of your output is dependent on how good the AI is. Let’s be honest, AI is not always reliable, and if you can be replaced, you will be replaced. This atrophy of core skills is why RGI is so dangerous.

Critical thinking skills and problem-solving abilities are like muscles. You cannot develop these overnight, and it requires training. Think of it this way — if you want to be fit, you need to exercise. Using AI to complete your assignments and expecting yourself to improve is like trying to improve your stamina by driving your car for an hour everyday. Just as physical exercise is important to build strength and stamina, solving those problems and personally attempting those writing and coding assignments are crucial to improve yourself intellectually. 

Over-reliance on AI also leads to moral problems. You are expected to think deeply about the problem and turn in the assignment given to you. Instead, you offload it to the AI. The writing produced by most AI tools can be easily distinguished from human writing, and you lose credibility in the eyes of the reader once they know that it is not your original work. The reader is less likely to take you seriously because they now know that at least part of your work is not your own. 

Repeated use of AI also leads to people overestimating their abilities, and we increasingly see this in many areas. Users have poor understanding of the basics but wrongly assume that they know “advanced stuff”. This is a consequence of the Dunning-Kruger effect, where people with low abilities in a specific area (which is true when you are beginning to learn a specific field) overestimate what they know about it. Developing expertise in a field requires you to put in a lot of practice and think independently about the problems. You cannot ask an AI to think for you and assume that you now have mastery.

Over-reliance on AI also makes you lazy and generally results in sloppy work that may take a lot of effort to correct. AI tools hallucinate: create references that don’t exist, call functions or variables (in programming) that are undefined, put in meaningless code, make statements that physically or logically don’t make sense, and so on. You then have to put in a lot of effort to identify and correct those mistakes. 

Programmers generally spend significantly more time debugging and improving code compared to writing code. However, we are seeing a sharp surge in “vibe coding“: using AI to produce large blocks of code. This code is much harder to debug or improve, since you are not sure why the AI chose to write those lines. It is natural human tendency to assume that the code is correct, or simply give up frustrated when you discover that it does not work the way it is intended. This leads to more unstable code that we don’t really understand.

What we can do about it

In today’s age, we cannot afford to ignore AI. We are expected to know how to use these tools, and must learn to use them judiciously.

We must therefore make conscious and deliberate choices when we use AI. If you are a student, then don’t use AI to solve your homework assignments or exams, as you are losing the long-term benefits of developing critical thinking skills to get short-term benefits of good grades. In the long term, a minor bump in grades doesn’t matter so much compared to the skills you have developed.

AI can be very useful in clarifying concepts that you don’t understand, or getting more references for something that you are deeply interested in. However, be sure to think about the concept deeply before asking AI for the answer, and cross-verify the answer with solid/reputed sources.

There are many situations where AI can be used to automate certain tedious tasks, or simply save time. This is particularly helpful if you use this to complement your skills.

The most important thing is to critically consider the output of any AI tool before using it. Generally assume that the output may be wrong, understand what it says, and carefully check for mistakes. This ensures that you don’t face trouble later on, and develops the habit of critically evaluating things.

If there is a particular field that you should/want to get better at, then use AI sparingly. Don’t use AI simply to avoid mundane or repetitive tasks. You develop mastery only through repetition.

Those algebra and geometry problems that you studied in school may feel boring and you may wonder if you ever use them in real life, but by solving those you learn to think about problems logically and develop abstract reasoning skills. While one can debate the actual proportion and quality of content that we see in the school syllabus, there is no doubt that solving a variety of math problems and writing a lot of essays are the mental equivalent of regular gym workout. 

While using AI to solve all your problems is a bad idea, it can be a useful tool to evaluate your work and get feedback. For example, suppose that you have to write a thousand-word essay or a program to solve a problem. An AI tool could very easily do this for you. However, if you want to get better at writing, critical thinking, or programming, then you must do this on your own. Once you write the essay or code independently, you can ask the AI for feedback, and use this feedback to improve the next time.   

AI is expanding in scope and ability at a rate that we cannot imagine. It can be a great boon if we use it to improve ourselves, or a bane that takes our jobs if we rely too much on it and make ourselves replaceable.