Problems with Artificial Intelligence

By Kurush Dordi, July 25, 2026

Artificial Intelligence is an umbrella term. It regroups various types of computing and algorithmic models. There are three types of AI: the Weak or Narrow AI; Artificial General Intelligence or Strong AI; and Super or Self-aware AI. So far only the first one has concrete applications and the last two are strictly theoretical. All the so-called AI applications and algorithms available on the internet are Narrow or Weak AI models. And they are not new. They have just become more sophisticated as the processing power of integrated circuits has been increasing exponentially over the past 75 years.

The exponential trend turned a corner in the 2010s with the development of artificial neural networks models and parallel processing. Since then, the number of AI applications has soared, from Siri in 2012 (Reactive AI model) to the latest Large Language Models (LLM) like ChatGPT (Generative AI model). Some of the most recent AI models have required billions of petaFLOPs for their training, a FLOP being a floating-point operation or a single arithmetic operation such as addition, subtraction, multiplication or division. And a petaFLOP is a quadrillion (10¹⁵) FLOPs. Such is the scale of computation required by modern AI models.

So as you have probably figured out by now, the number of microprocessors needed for all those applications has also been increasing exponentially. All these microprocessors are gathered in large data centres that are commonly called the Cloud, a term that invites you to imagine something light and almost immaterial, but this is far from being the case. It requires land, steel, concrete, copper, aluminium, rare earth metals, and a constant supply of energy and water. All of which create an additional stress on the already gigantic demand for all these resources and of course has a significant impact on the environment.

If you thought living near a chemical plant or wind farm was a problem, you wouldn’t want to live near a data centre, where the constant hum of the climate control units and the polluting fumes of the diesel generators has become a growing concern. The use of water for the cooling systems and the electricity demands are also a growing issue for local farmers and residents.

And finally, the actual human cost – ranging from the exploitation of cheap labour required to train the AI models with astronomical amounts of data, which have often been taken from the web without the consent of their creators – to the flooding of social media and internet platforms with meaningless AI slop, misinformation and porn, is also very high.

All that being said, AI in itself is not the issue, and the same goes for most technology which, if used ethically, will be in the best interests of humanity in the medium to long term. However, we can see that this is not generally the case today, apart for some scientific and research purposes. Knowing, for example, that using an LLM such as ChatGPT for a simple question will use ten times more energy than a basic Google search, what can we do? Such is the case with any technology, we could ask ourselves: Do I need to use it? Could I use something else with less footprint?

The easy access to technology in general makes it hard not to use, especially if it makes our life easier. And it can be easy to forget the actual cost of using those things. This is not me trying to chastise anyone tempted to use various technologies, as I do the same. But it might be a good reminder to dial it back when we see how dependent and existential a piece of technology, such as AI, has become. Going back to simpler ways, even if it can be sometimes less convenient, can be a way to reconnect with others, nature and especially ourselves.

Image Credits: By rawpixel | CC BY PD