Not so artificial intelligence
The term artificial intelligence (AI) has been around since 1956, when a group of scientists gathered at a summer project camp known as the Dartmouth Conference. They came up with the term to describe a new study field. The term was rooted in these scientists’ vision to create machines that could replicate human intelligence.
Decades later, we’re still living with the naming legacy of these happy campers. What we nowadays understand as AI, is not artificial, nor intelligent. (I heard this thought from Kate Crawford whose work I really urge everybody to check out! We seem to want to think that AI is not material but somewhere up in the cloud, some bits or digits flying across the sky.
Even though AI systems can replicate some processes of the human mind and intelligence, they do not have a sense of self, purpose or consciousness. They are not sentient and will not be for a long time if ever. The systems only do what they were trained to do and are only as good – or biased as the data they were trained on.
But of course, as these systems become more self-learning, we cannot rule out the possibility that they might evade human control and become a threat. Science Fiction has further fed us with the notion of killer robots taking over humanity. When growing up one of my favourite films was Terminator 2 – and I learned that Skynet becoming self-aware was the worst thing and inevitable as those sequels and Terminator models kept being released!
These existential threats make great headlines but distract us from the very real and current harms of AI. Training and deploying AI models impact our natural resources massively and can harm humanity much more than any Terminator wishing us ‘Hasta la vistas’ might.
The carbon footprint of training a (simple) large language model is estimated being the equivalent of 125 roundtrip flights from New York to Peking. Cooling down the central processing units that keep AI services running requires huge amounts of clean water. One carefree ChatGPT discussion is estimated to equal pouring 0,5 litres of drinking water down the drain (source Forbes).
We’re also running out of crucial minerals that we need for our batteries and devices. International Energy Agency has estimated that we are fast running out of key minerals like Lithium and will face shortages already in the next few years.
The harms of these systems are actual and taking a toll on our planet as we speak. Making the decision to embrace AI tools at large should always include a look at the environmental cost of doing so.
Ps. It is true that AI can also help solving these issues, look for more energy efficient ways to cool down CPUs or discover alternatives for minerals. That would require the will to invest in those projects.. To finish off on a hopeful note, let’s quote Sarah Connor from T2: "If a machine, a Terminator, can learn the value of human life, maybe we can, too.”
