AI is commonly used in so much recently, including AI art, programming and writing, so I thought it would be a good idea to explain my opinions on AI and the reasoning behind those opinions. Everything in this has likely been said before, but I want to write down my opinions anyway.
What is AI
There are so many different technologies that are called AI now, and all of them are completely different. Games have used 'AI' for decades, for how characters move around a world and react to events, which could be implemented as something like a decision tree, and only ever do what the authors originally intended to happen.
Machine vision has also been described as AI, for example modern Optical Character Recognition (OCR), or industrial processing, both of which use neural networks to detect what is occurring within an image. These techniques are, in my opinion not particularly controversial and the majority of the criticisms within this post don't apply to them.
When I refer to AI within this post, it is purely referring to generative AI, for example large language models, generative image models and other similar things.
Environmental Issues
The data centres that AI uses have significant environmental impacts. They consume large quantities of electricity, use lots of computer parts and are significant buildings. Especially given how electricity is getting more expensive, with the wars currently happening, increasing energy usage does not seem like a good idea currently, given how much we still depend on fossil fuels.
Large amounts of resources are consumed by construction of data centres, including computer chips, land and building materials. In my opinion, this can be justified if the output of the data centre is useful to society; however, given how many chips are being used for AI prices have increased for all computer parts. This price increase has made a large number of products more expensive, both for companies and individuals, which is increasing the cost of living issues most people are facing.
I have seen lots of reporting regarding the water usage of data centres. They do certainly use water, and it is likely that they could improve the efficiency of water usage through more expensive and advanced cooling systems. However, the actual impact of data centres on water seems to be relatively small compared with any other industrial use. Reducing usage is always a good idea, but this is not a major issue of AI in my opinion.
Providing electricity to new data centres is relatively complicated. In a lot of locations, the existing power generation is insufficient for the data centres. Many have instead installed their own power generation on site, which has many issues. Typically, data centres don't have much pollution, however with their own power generation, both noise and air pollution will be generated at the data centre. This is a significant issue for the communities that they are constructed in. Some data centres have been built in residential areas as they are considered to not cause pollution, however the power generation is still being installed and it is likely that large scale power generation without a data centre would not be allowed within a residential area. This power generation in some cases (xAI) uses gas turbines, which are not renewable and will increase the effects of climate change when the rest of the world is attempting to move away from fossil fuels.
Ethical Issues
In order to create generative AI, large amounts of input data is needed. It is widely reported that LLMs have consumed content on the internet, books, and pretty much anything that anyone has ever written. There is no realistic way that AI can consume all that information, while not counting as breaking copyright. AI models are capable of repeating their input texts pretty much exactly, which to me is exactly what plagiarism is.
Previously, if you produced a copy of an existing book, the author would likely have an issue with that and stop you from copying exactly what they did. So why do AI companies get to have a free pass to copy everything, in my opinion they shouldn't[1]. It seems fair to me that authors should get paid for what they have done.
Potentially, AI models could be seen to be learning from the information used in their input, rather than copying it, however this does not occur. A person learning from what they look at involves the person making opinions and deciding what they think, this is the key step that doesn't happen with current AIs. The memory features in AI platforms simply copy the past conversation back into the input of the AI, a model doesn't learn from anything that has happened. This, in my opinion makes AI use plagiarism, exactly the same as you copying text straight out of Wikipedia.
Personally, if I make something, I will have put effort into it, why would I want someone else to be able to copy what I've done with minimal effort. If they also put in effort, then I have no issue there, but no computer will be able to have the full context towards anything I've made, as everything a person does will depend on the world around them, their emotions and opinions and learning from past experiences, something that AI is not capable of.
The building of data centres does typically create many jobs; however, this is only during the construction period. The maintenance of a data centre once it is built typically does not require many people at all, and therefore a data centre cannot be considered to help any community it is built in in the long term. This has however been publicised by some companies due to the initial number of construction jobs. In the long term, AI companies have explicitly stated that AI is aiming to replace people's jobs, by doing the work on a computer instead of providing jobs. This will, and already is causing problems, with the number of redundancies caused by companies replacing people.
As well as AI's data collection being un-ethical, in my opinion AIs use of the data is unethical too. The use of AIs relegates the thinking and hard parts of work to a machine in a data centre, owned by a big corporation, why would I want mega corporations to be able to control and know everything I do and think, I don't support mass surveillance programs, and I don't want to support companies censoring and controlling everything I do, therefore I won't use AI to do it.
These companies charge significant amounts of money for their services, so if the amount of money you have available changes, or the terms and conditions of the provider change, you may not have access to the services you used. Alternatively, by using LLMs you are requiring the use of very expensive computers, which also are not widely available. Nothing about AIs makes access to writing or art or anything else easier for people, it just adds more friction and issues.
Issues with the Output of AI
Even if I ignore the environmental and ethical issues above, there are other issues I have with AI, in particular relating to what it does.
A lot of people appear to be becoming reliant off of AI and losing the ability to do thing on their own without relying on a computer to do it for them. Every time I rely on a computer to do something for me, it seems like I get worse at doing that thing myself. If I don't do it by hand, then it is so easy to loose the ability you did have. I feel like this has happened with calculators, it is so much easier to just use one, to the extent that I rarely need to know how to do basic maths myself. The same applies to writing, if I don't write with a pen and paper for a long time, it is much harder to do it again. These examples as individual skills are one thing, but I don't want to let it happen to my ability to express myself, as would happen with widespread AI use.
If an AI does something for me, how can I improve what I'm doing? The AI does it, so it isn't me learning anything, I won't know what to do differently next time. Maybe some people are alright with that, but I'm not ok with just staying as I am without changing.
Personally, I don't like reading AI generated text, something about it seems wrong, like there is a lack of emotion and feelings behind it, as if there isn't anyone writing. Every person has an individual style behind what they do, but all AI text feels the same, very bland text.
Some people seem to be getting some kind of 'AI psychosis', where they are believing AI does so much more than it actually does. There are stories of people thinking they have an 'AI partner', or believing that AI knows exactly what is happening, the level of delusion that you can get seems frightening to me, as if it is changing who people are, rather than letting them exist themselves.
This post is written mainly so I can confirm what my own opinions are, If I hadn't personally written this, then the thoughts I have regarding AI would be a lot harder to communicate to even myself. It is the act of writing something like this down that help me to form opinions, I shouldn't be letting a computer form my opinions for me.
The accuracy of any AI produced work is very questionable. The current AI is effectively an advanced autocomplete, there is no guarantee that the output is correct in any way. At work, both me and the company are legally responsible for what work I produce, so I could be questioned about mistakes I make. As a civil engineer, mistakes in what I do can quite easily cause people to die. I am not willing to trust an AI to get every aspect of bridge design right[2].
Outside of work, I still care about the accuracy of everything I'm doing, it is annoying if something is wrong. I don't want to be misleading with anything I write, I want minimal bugs in the code I write. I'm not trying to 'move fast and break things', I want what I do to work. What is the point in making something that isn't right, I might as well have not done it at all. If I do the task myself and make mistakes (as I am certain to), at least I can learn from that and improve in the future.
Benefits of AI
There are, in my opinion, a few benefits to AI, which could be reasons that I should use it.
If I was to use AI, I could make things faster and increase my productivity. However, that comes at the cost of what is produced being significantly worse. Lots of AI written art, code and text is being produced at increasing speeds, however all of it is worse than that which could be done by a person (or it copies what a person has already done). The increase to productivity isn't worth the decrease in quality to me.
AI can be used to automate some of the boring parts of some writing, coding and other tasks, however it seems to me like that could just be masking a problem by getting a computer to do it. Why is there boring repetitive code to write, shouldn't you change how the code is written so that that can be avoided? I'm aware that this doesn't necessarily apply in all cases, but at least personally it does for me (I know it will be much harder in a corporate environment to change anything enough).
AI seems to be, and acts like a very advanced autocomplete, which is using significant quantities of resources. Autocomplete is helpful to me, although I can understand those who don't want to use it, but the existing autocomplete we have is already good enough IMO; and the existing technology is so much easier to run, it has been in use for decades already.
I am aware I have no skills in prompting AI, however this doesn't especially bother me. There are lots of things that I don't know how to do, am I'm sure that I could learn how, but at the moment, why would I? There are no benefits to me personally in getting better at mass producing slop.
Context and Limitations
All the points I've made so far are mainly aimed at me doing things in a personal capacity, i.e. don't apply to working in a corporate environment. Therefore, what I am trying to achieve is likely different to what other people are looking for. That said, a lot of the issues still apply regardless, whether you are in a company or not, AI is still creating global environmental problems.
Conclusion
Therefore, in conclusion I won't be using AI for anything. The main reason for this is the environmental and social issues, as someone who is part of this world, I have a responsibility to not make it worse for other people, it would be morally wrong of me to use a technology that is actively making people's life worse[3].
I am well aware that this will likely make me slower than if I used AI, however personally, I will get more out of doing tasks without the use of AI. I am aware that this is not something that is possible for everyone; however, using AI is less fun to me, I don't get to say I have achieved something, it was just the computer, with minimal involvement from myself.
Sorry if this whole thing feels like a badly edited stream of thoughts oops! (It is) If I had more time, I would have written less.