FEATURE: Why How We Use AI Should Matter
By Mumeenah Abdulrahman,
Somewhere on your phone, there is a conversation you would not want anyone else to read. You told it about your anxiety, your medical results, your business plan, or your relationship. And the person on the other end was never a person.
That is now normal. In Nigeria, with over 109 million internet users, “let me Google it” has quietly become “let me ask ChatGPT.” We ask these tools to rewrite our messages, explain our assignments, create our pictures, draft our CVs, and sometimes help us make sense of feelings we have not even told our closest friends.
Artificial intelligence, or AI as we fondly call it, emerged as a formal field of study in 1956, when computer scientist John McCarthy and other researchers introduced the concept at the Dartmouth Conference. Ever since then, there have been many improvements, and it has moved steadily from academic theory into the mainstream spotlight.
We first saw it with Apple’s Siri in 2011, which helped make voice interaction familiar to ordinary smartphone users, and then again in 2022 with a major shift when conversational generative tools became accessible to everyone with an internet connection. Since then, there has been a continued transformation in how we interact with technology.
There are many ways that AI shows up in our daily lives, even more so in recent times. Today, there are social media accounts run entirely by virtual personalities with millions of followers, and content powered primarily through automation, from scripts to voiceovers to editing.
We now have the creation of fictional characters with storylines designed as prompts which are in turn generated into full stories, humans confiding in chatbots about their most human secrets and desires, and people joining trends that give you a prompt to tell your assistant to roast you, praise you, or plan your life. Whether we realise it or not, we now live a life that is deeply intertwined with these systems.
While all these things are not inherently bad, there is an element of concern in them that cannot be ignored. I did my research on the two main discourses that come up when the topic of being “Pro-AI” or “Anti-AI” technology is being discussed.
These discuss are, “How much water is used to power these models?” and “Is my data being used to train them?” There are no straightforward answers when these two questions are asked, because the reply depends mainly on the factors at play.
*The Water Question*
Some say it uses 1/15 teaspoon of water to generate a single prompt. Others say it uses a whole bottle of water. Interestingly, both answers are correct, but with a different outlook. The 1/15 teaspoon logic is about the water directly used by the data centres that store information and do the processing. These centres use water-based systems to cool the servers that get extremely hot. The bottle of water logic is broader. It includes the water consumed by the power plant generating the electricity that powers the chip in the first place, and it also accounts for longer conversations, multiple retries, and less efficient systems.
In reality, there is no exact amount of water consumed during an interaction. It heavily depends on three key factors: the model, the media being generated, and the environment of the data centre. First, the model itself. A larger, more advanced system requires much larger processing power, thereby consuming a lot more water and energy.
Second, the media. Generating images and videos requires a far higher consumption of water and energy than generating plain text. A single generated image can use as much energy as charging your smartphone. Third, the environment. The location of the data centre matters a great deal. If the facility is located in a dry and arid region with not enough water, powering these models can put real pressure on local supply and lead to water stress for communities nearby, while the same facility placed in a cooler, more suitable environment with efficient cooling would not have the same impact.
In places where water is already scarce, and we know how real water scarcity can be in parts of northern Nigeria, that extra demand is not just a technical issue. It becomes a community and ethical issue. So, the question is not just how much water is being used, but where that water is coming from, how efficiently it is being used, and who else needs it.
*Does Our Data Train these Models?*
This is the second question that worries many users, and it is important to understand it clearly because there are two distinct phases of interaction that people often confuse. When you converse with a chatbot and it remembers some details about you which you just mentioned in passing, and those details recur in another conversation entirely, this is called the “remember phase”. It is often called memory or personalisation, and the system is designed to make your experience more tailored and continuous.
The second one is when your conversations are used as raw material to train future versions. This is called the “learn phase”. This does not translate to having your private conversations displayed verbatim in another user’s chat. Instead, it means the patterns, word choices, feedback, and corrections from many conversations are aggregated and used to create a smarter next version. The “learn phase” is what is actually used for training, and it is the default setting of many free platforms. In a sense, for free tools, your form of payment is allowing your interactions to help improve the system.
This is why your settings matter. Whether your chats are used for training depends on the platform, its privacy policy, and whether you have opted in or out. The real bone of contention here is not just about your data training a model; it is about the kind of data you paste into these tools.
Information like passwords, API keys, private keys, confidential company documents, unpublished research, or personally identifiable information such as pictures, videos, medical records, and financial information are highly sensitive and deserve to be treated as such.
Once sensitive information is entered into a system that learns, it is difficult to fully retract it, and that is why digital literacy is no longer optional for the average Nigerian youth who lives online.
The main purpose of this article is not to reduce, diminish, or vilify the way we use or see AI in society because, it is a very helpful resource that has improved and can continue to improve a person’s standard of living. It can create job opportunities, allow humans to align and process their thoughts better, improve communication skills, enhance cognitive functions, and provide overall better life assistance if used correctly.
For students, creators, small business owners, and professionals in Lagos or Lafia, it has lowered the barrier to entry in many fields and made complex tasks more approachable.
The problem was never really about the technology itself. It is mostly about how humans choose to use it. We have a responsibility for how we use the resources at our disposal, and we need to move from passive consumption to intentional use.
Therefore, the questions we should be asking ourselves are not simply “How much water is used to power these systems?” or “Is my data being used to train them?” Rather, our questions should be more personal and practical, “How do I use these tools in a way that does not reduce or devalue my intelligence, but rather optimises it?” and “How do I get the best from them without compromising my privacy, while also being environmentally conscious?”
That means learning to write better prompts instead of copying and pasting blindly. It means double-checking facts instead of assuming the answer is always right, not pasting sensitive data, reading privacy settings, and understanding when to switch off model training if the platform allows it.
It also means being conscious of our usage: do we need to generate twenty images when two will do? Can we combine our requests into one clear instruction rather than ten vague ones? These systems are powerful resources, much like electricity or the internet.
How we use them is still our responsibility, and the future will not be about choosing between human intelligence and AI systems. It will be about learning how to use both wisely, ethically, and sustainably.
