With the growing use of artificial intelligence we have to start to question if it is really helping us, or if it is harming us in the long run. While some may be inclined it is the latter, I’m going to be the devil’s advocate and explain that it is both a good and bad thing, and what my thoughts on everything is.

This post will be split up into three sections detailing everything to the best of my abilities, those being:

• The Good — The good things about the growing use and development of AI.
• The Bad — The bad things that are resulting from the growing use of AI.
• The Ugly — The worst of the worst that is coming from growing use.

It should be noted these are my opinions these are not facts, and are heavily subjective.

This all being said, let’s get started.

The Good — Why I Am Supportive Of AI

Just like your terminal, or text editor, AI is a tool. It can provide helpful information and help with repetitive tasks such as writing loops or boilerplate, and can sometimes act as a rubber duck when needing to debug or think something through. It has helped me to figure out things that I didn’t understand before. For example, I have been studying bytecode and for the longest time didn’t understand how function calls, recursion and loops worked when it comes to something like bytecode which is more linear. Claude was able to help me understand how function calls and such work when they are compiled to bytecode:

Function calls in bytecode typically work through a stack-based mechanism where the calling code first pushes the function arguments onto a runtime stack, then executes a call instruction that pushes the current execution context (like the return address and local variables) onto the stack, jumps to the target function’s bytecode address, and begins executing the function’s instructions. The called function operates on its own stack frame, potentially pushing and popping values as it runs, and when it finishes, it places any return value on the stack and executes a return instruction that pops the saved context to restore the previous execution state and jump back to the instruction following the original call. This stack-based approach allows for nested function calls, recursion, and clean separation between different execution contexts, with the virtual machine or interpreter managing the stack operations automatically as it processes each bytecode instruction.

This is a good explanation, and it makes sense to someone like me who is just getting into low-level systems such as compilers and interpreters. But like I said at the start of this segment, it is a tool, you don’t rely on an editor by itself to help you produce correct code for example, you rely on compiler errors, an LSP maybe, and documentation as well. This should be the case with using AI as well, don’t soley rely on it, use it along with the other tools that you would normally use.

That is the important thing here, it is a tool, and I feel like I need to say that repeatedly for it to get across to some people. It is not omnipotent, it is not “all-knowing”, it is a tool which is trained on books, videos sometimes, but most notably, human generated data, humans are wrong a lot of the time, and because of this, the AIs we use can be wrong as well. They need to be treated with caution and not whimsy.

And that is where we veer off into the next section, so let’s continue.

The Bad — Why I Am Cautious Of AI

Due to AI being trained on human-generated content, it can be wrong a lot of the time, and part of this is because of hallucinations. Hallucinations are, put simply by the Wikipedia article I just linked:

In the field of artificial intelligence (AI), a hallucination or artificial hallucination (also called confabulation, or delusion) is a response generated by AI that contains false or misleading information presented as fact. This term draws a loose analogy with human psychology, where a hallucination typically involves false percepts. However, there is a key difference: AI hallucination is associated with erroneously constructed responses (confabulation), rather than perceptual experiences.

In otherwords, occasionally AIs will “generate” information that is inherently false or misleading and present it as fact. This is not just a fault of the technology though, it’s also a fault of the data it was trained on. This is exactly the reason I am cautious of what AI says or presents, and why I insist in the previous section to treat it as any other tool. Always do more research on what an AI says or produces, don’t trust it’s code or comment without further research and evaluation. After I got the explanation of how functions work from Claude, I did more research on my own without using it. I got a rough summary from it, then did further research to make the knowledge I supposedly gained more concrete.

You wouldn’t trust another person to commit to main without first making sure CI pipelines pass and that the code doesn’t contain vulnerabilities, so why would you trust a random computer program spitting out text to? That seems to be a common pattern these days, unverified information being trusted as fact without further research on the subject at hand. This isn’t a fault of the tool, this is a fault of the user. In order to properly use the tool we have been given, we need to first treat it as a tool.

So what’s the solution here? Get rid of AI completely? Improve it over time? No, neither of these are a solution. Removing the use of AI completely would be a technological setback, and improving it over time, takes, as it says, time. We need a solution for now, and that solution is to stop trusting it as fact. Start treating it as a tool, a part of your workflow, not as your entire workflow.

The Ugly — Why I Don’t Support AI

The growth of AI doesn’t have environmental effects in itself, but what does is the datacenters that are hosting the models we use. You can take a look at any of the following research regarding environmental effects from the use of AI here if you wish, just note that these are not of my own writing or opinion:

• https://www.iea.org/reports/energy-and-ai
• https://mit-genai.pubpub.org/pub/8ulgrckc
• https://libguides.ecu.edu/c.php?g=1395131&p=10318505

These go into detail about the environmental effects and energy consumption from the training and use of AI models. Realistically they don’t use more than we as humans do, but that doesn’t mean they don’t contribute to the acceleration of the effects that green house gasses and global warming have.

Do I think we should stop using AI? No, I don’t. What I think is we need a more environmental friendly approach to energy as a whole. This is less an issue with the technology itself and more of an issue with the way we power the technology. But for now that approach does not seem to exist or isn’t used enough to fully be “earth-friendly” in a sense. It’s harmful, but not because of the tech itself.

However, me running DeepSeek or Gwen3 on my personal computer isn’t going to have the same level of effect as a large datacenter running one. A datacenter runs models in mass. They run thousands, upon millions of instances of models. On top of this, I’m not going to be training a model on my local computer, I’m going to be using pre-existing models, and using models uses less energy than it takes to actually train one. I doubt one person using a model on their local computer is going to contribute enough to the problem to be… problematic, and people need to keep that in mind, especially in the short-term.

Conclusion

My thoughts on the matter can be basically summarized into the following:

• Using AI as a tool as it’s intended to be, is actually a good thing.
• We need to stop trusting what AI produces as fact and research more.
• We need to be skeptical and be less reliant on the technology as a whole.
• We need a better solution to power the models we use in the long run.

AI is useful, it can be a good thing to have in your toolchain, but until we start treating it as that– a tool, we will continue to have problems with people trusting what they say as fact, and in result misleading people who use it. I like using AI for personal things, and I add a notice to projects when I use it so people are aware that it was AI assisted. I don’t believe destroying AI is the correct path, I believe we’re the issue, not the technology we use, and that is something we can improve on… but will we?

This article is licensed under the CC BY-NC-ND 4.0 license.