Perfect folly: the Race for Artificial Intelligence is the wrong race at the wrong time. The race we urgently need is a Race to Sustainable Cognition.
The world’s largest companies are now frantically developing AI in total disregard of global climate and ecological crises, for AI triggers vast new demands for energy, water and certain minerals at the worst possible moment.
“We need way more energy than we thought” said Altman recently, while Zuckerberg has mused that new data centres may require dedicated nuclear power plants. This at a time when the ‘clean transition’ is not reducing GHGs anything like fast enough, which ought to be focusing would-be ‘great minds’ on energy reduction, not on new, discretionary, forms of energy consumption. (AI companies invariably have some line about AI’s potential to improve ‘energy efficiency’, which only betrays their ignorance of Jevons’ Paradox.)
In short, the AI race is the enthusiastic bringing of petrol to a fire – just as climate scientists report more concern and despair than ever before (see last week’s Guardian survey).
There is a tendency to portray Huang, Nadella, Cook etc. as ‘leaders’ because they develop technologies few of us understand, but in fact they are more in the grip than most of a deeper tide they either cannot see or do not have the courage to resist.
That deeper tide is ‘externality-denying capitalism’ – the system most of the world now lives under and which has lifted human decision-making entirely off its biophysical moorings. Because the economy does not recognize the real cost of energy, AI leaders are highly incentivized to not reflect on what they are doing to the planet. Per one Twitter wag: “it is difficult to get a person to understand second-order effects, when their stock-based comp depends on their not understanding it.”
In turn, externality-denying capitalism is the result of a long lineage of errant thinking including, in reverse chronology, shareholder value maximization, ‘complete’ and ‘efficient’ markets, disregard of externalities, economics’ physics-envy, and the primacy of reductionism over systemism. For 400 years, the Western world has been enacting a slowly accelerating cognitive-behavioural feedback loop in which how we choose to think about the world gradually shapes how we behave in the world and, eventually, how the world becomes.
In other words, we have been doing ‘artificial intelligence’ for a long time. Its most consequential outputs are heat and entropy. What is new is the pace and scale of consequences, and the strange inclination to celebrate the fact.
The race we need to run, and which ought to be attracting ‘leaders’ is the Race to Sustainable Cognition. This might start by asking whether blindly following a decision-making system that so incentivises AI at this time is a mark of intelligence or stupidity.
Postscript
The structure of the cartoon is an ‘action-justification’ loop, in which a certain way of thinking has justified the development of AI, for which the current high level of enthusiasm despite its evident sustainability and climate problems now prompts an impulse to double down on the enabling justification…
In other words, ‘externality-denying capitalism’ has the basic dynamic of a reinforcing or runaway feedback loop: errant thinking encourages ideas that boost share prices, which to be viewed as ‘sustainable’, rely on re-committing to the errant thinking that spawned the idea…
Runaway feedback loops are the essence of ‘unsustainability’.
To understand this, consider the opposite: the ‘sustainability’ of a system – from small organism to ecosystem to human society – is the ability for the system to self-regulate in the face of internal or external disturbance, i.e. to suppress potential runaway feedback loops that might lead to collapse or unravelling of the system. This is ‘homeostasis’, which is the very essence of sustainability. A system is ‘sustainable’ to the extent that it achieves homeostasis without interruption.
In the human body, for example, we know those runaway processes that overpower the body’s efforts to achieve homeostasis as ‘pathologies’. Alas, one or other pathology eventually achieves enough momentum to bring down the whole system.
Ecologists, biologists, medical doctors etc. know all this as a matter of training and professional experience, but many other disciplines and professions have no inkling of this basic phenomenon of life. In fact, disciplines like economics and finance cultivate a mindset that denies the relevance of this.
The irony of the AI action-justification feedback loop is that one of the disciplines that most cultivates awareness of feedback loops is… computer science! ‘Do…loop’ problems that cause computer ‘crashes’ are well known to coders. Software sometimes gets stuck in loops it cannot get out of. The screen goes blue. You have to turn the computer off and start again.
Human society is effectively stuck in such a loop, because our cultural software – norms, habits, incentives – is stuck in such a loop.
Alas, while AI technicians are presumably highly proficient at recognizing and fixing reinforcing loops that appear in the code in front of them, they would appear to be oblivious to the larger culture-scale reinforcing loop of which their thought and action is merely one link in the chain. They are part of the loop! As is anyone just trying to get by in the system of externality-denying capitalism…
How do you stop a reinforcing loop that is doing damage? You see the loop for what it is, you step outside of the loop, and you stop it.
Leave a Reply