Published in 2/2024 - Matter and Intelligence
AI Is Not Born out of Thin Air
Before we can go into the possibilities offered by AI applications, we need to understand what it takes to produce artificial intelligence, writes Ville Paananen.
The history of artificial intelligence is eventful, but in broadly sweeping terms, the development of information technology can be viewed as a gradual increase in the level of abstraction: complex AI applications are becoming less and less transparent because recognizing and understanding them requires an awareness of deep learning, data mining and data classification, as well as of decentralized global IT systems – not to mention the social and environmental impacts of information technology. Due to the difficulty of grasping what artificial intelligence truly entails, I would propose that, before we can even get into the possibilities offered by AI applications, we should first seek to understand artificial intelligence from a purely material perspective: what is artificial intelligence and what does it take to produce it?
On the surface level, the user may only be aware of an AI application as the result of a specific request or task, but there is always a great deal of data mining and processing, as well as highly energy-intensive computing, going on in the background. In addition, a large proportion of all AI-generated data is actually produced manually. The manual data classification is monotonous and mentally straining labour; global subcontractors recruit crowd workers who typically work for a minimum wage and with inadequate mental health services to filter out, for instance, violent and pornographic content – of minors, even – from the datasets scraped from online. Conceptually speaking, “artificial intelligence” filters this human input into valuable raw material for computer processing, but it also overlooks and oversimplifies the value of human cognition. Data-based AI tools, then, largely continue to utilize manual labour and rely on the outsourcing logic of the global market economy.
In addition to the human-powered data classification, the AI applications also need to be trained, which requires massive amounts of processing power and physical data centres built for this purpose. It has been estimated that, by 2027, the energy used by AI will be comparable to the annual consumption of Sweden or the Netherlands. However, as a result of the connectedness of information technology networks, it is difficult to make precise energy calculations, which also complicates the impacts assessment. Still, if artificial intelligence is to become a part of architectural design work, the tools used in the design process will inevitably begin to have an effect on the carbon footprint.
Would architects who utilize AI also have the responsibility of promoting fairer and more sustainable artificial intelligence?
Does AI offer the kind of design options or processes that would compensate for the used resources? Can there be room for AI systems in a world that needs sustainable design? If we cannot even describe the operating principles of artificial intelligence, how can we teach future designers to make choices when it comes to using AI tools? Instead of focussing on specific AI applications, answering these questions calls for a wider examination and understanding of the role that technology companies, such as OpenAI, play in establishing and maintaining global power dynamics. Alternatively, if we appreciate the data-based processing and syntheses offered by AI applications, can we also work towards making sure that the AI data represents a world view that is aligned with our own values? Would architects who utilize AI also have the responsibility of promoting fairer and more sustainable artificial intelligence?
This type of global and material examination also brings forth the matters of ecology and post-humanism, which have been widely covered in this journal in recent years, to the forefront of the AI discussion. As stated by James Bridle in his 2022 book Ways of Being, what we need is not artificial intelligence, but the real intelligence of all of the world’s living creatures big and small, human beings included. From this perspective, intelligence that is based on an abstract AI with an obscure origin feels more or less exploitative and impoverishing. I, for one, am not convinced that we can afford the kind of AI that we are currently witnessing. ↙
VILLE PAANANEN is a doctoral researcher of computer science with a Bachelor of Science degree in Architecture from the University of Oulu.