Alternatives to TINA
On growth models, economy, and AI
There are things we take for granted because they dominate our lives today, such as the internet and, in particular, its subset of social media, or the breakneck developments in AI. A world without social media is hard to imagine, even harder a world without the internet, or without computers or smart-phones. As an artist you cannot rebel and refuse to share your music on Spotify, no matter how shady their business model, because that's where the audience is. Music streaming through Spotify has come to stay1. Get used to it, the fatalist says, this is the future we will have to live with. TINA, There Is No Alternative, as Margaret Thatcher used to say. Finance capitalism and the eternal growth imperative is here to stay, because you remember communism. Technological progress is unstoppable, because you don't want to live in a cage, do you? Fossil fuels are still in supply, despite so-called green growth. Technological development will continue on its accelerating path, aided by scientific discoveries and the computational power of AI. Moore's law predicts an exponential increase in computer efficiency and it still holds, after all these years.
Processes with reinforcing feedback tend to yield exponential growth. Whether it is bacteria on a Petri dish or the improvements in AI models, the growth is slow to begin with, then increases and blows up before we expect it. Human intuition seems to work better with linear growth where the increment per time unit is always the same. If a natural or social process is well captured by an exponential growth model, then the model is probably over-simplified and only valid over a limited time span because no growth process goes on for ever (with a possible exception in the expansion of the Universe, if that cosmological theory should turn out to be correct). Bacteria in a Petri dish grow only as long as there are nutrients and free space; when they fill up the dish they will eventually die off and what appeared to be an exponential growth curve will turn out to be a bell curve. Peak oil theorists similarly predicted the year when oil extraction would reach its peak, assuming it would then decline symmetrically over time. Ugo Bardi has proposed another scenario, which he calls the Seneca cliff, in which the decline will be much faster than the rise. The human population has grown exponentially for a while, but cannot continue to do so much longer. Space migration is not worth taking seriously for a number of reasons, including physiological degradation in low-gravity environments and radiation exposure. We are stuck on Spaceship Earth. If the population cannot continue to grow, it may either collapse or reach a plateau after the growth rate gradually slows down. The latter scenario of a sigmoidal growth dynamic is optimistic, and probably unrealistic given the current overuse of non-renewable natural resources.
Economics is a tricky subject which too often seems to be about cooking the books; it obscures rather than clarifies matters. Money, which is used as a measure of value much like seconds are a measure of time, is not a real value asset in itself even if it is usually treated as such. A plot of land with healthy soil, on the other hand, is a real value asset. If you need to move some heavy furniture a kilometre uphill you'd be grateful to have a car instead of having to carry it over on a wheelbarrow by muscle power. Although you would pay for the car and the fuel with money, the gasoline or diesel represents what is sometimes called an energy slave; the fuel gets the work done. Transitioning to solar and wind power and other renewables also requires the use of fossil fuels to build these technologies. The historical period in which theories of economics have been proposed, roughly the last two hundred years, also corresponds to the period when the economies have had an input of "free energy" from fossil fuels. The earliest discovered wells were easy to exploit, whereas the extraction cost in new oil fields keeps growing. The energy return on energy investment (EROI) is steadily shrinking, approaching a theoretical minimum where the amount of energy recoverable is the same as the amount of energy required to extract the oil. Nate Hagens likes to refer to this exceptional historical period as the carbon pulse. Despite rapid improvements in renewable energy sources, the convenience which petroleum offers will be hard to replace. Therefore, extrapolating expectations of economical growth and technological progress based on what the carbon pulse has made possible is bound to lead to bitter disappointments.
However, in the case of some negative developments we may find solace in the fact that exponential growth processes often turn out to reach a peak and then usually decline or flatten out. Currently, the improvements in AI are accelerating, which is unsurprising when large language models are actually used for programming, including the task of improving AI models. Negative effects are widely acknowledged, such as unemployment, thorny copyright issues, declining critical thinking skills, increased risk of loneliness and even, in some cases, users who are lured into delusional rabbit holes by sycophantic conversation partners and end up in a psychological condition reminiscent of a psychosis. Some people have warned that AGI, artificial general intelligence, is just around the corner. When that happens, the AI will be able to improve itself, which it will do exponentially. Some LLMs have already shown the capacity to self-replicate and avoid being shut down. Researchers take cautionary measures of sand-boxing their experimental AI models, but the fear is the same as with viral research in bio labs; at some point the AI will be able to escape. Recall how Stuxnet operated: The advanced sabotage program targeted Iranian nuclear enrichment centrifuges by a computer virus distributed on USB thumb drives, because these facilities were kept off-line for security reasons. It would be wrong to ascribe intentionality to an AI, hence not correct to speak of its deceptive attempts to escape the sandbox, but it doesn't matter what one calls it. For any reason, someone could be tempted to bring a copy home and play with it on a networked computer, and the genie would be out of the bottle.
The energy consumption in training and using AI models is known to be huge, although precise numbers are hard to find. As I have argued in a previous post, the energy demand may in fact be the strongest argument against AI. Even if self-improvement may address efficiency, the energy demand is only expected to rise — exponentially, of course. Communities in proximity of data centres may see more expensive electricity bills to begin with. Nuclear plants, coal plants, solar panels, wind farms, or whatever is able to generate electricity will be built near the data centres. This energy will not be available for other needs. Assuming these data centres are made indispensable to keep societies up and running they will become military or terrorist targets.
If we have to wait for AI development to flatten out because of bottle-necks in energy availability or special material needs for chips manufacturing, then we may already have to experience more of the already troubling nasty psycho-social and economic effects, mixed with a few beneficial uses that techno-optimists may point to as an excuse to keep up the work. Surely the future used to look brighter in the old days. But to insist that AI is the future, and that we must therefore use it as much as possible because there is no alternative, is a fallacy. It is about as smart as the idea that NATO countries have to use five percent of their GDP for defence spending because daddy says so. And why does daddy say so? Because he is backed by a greedy military industrial complex with an inferiority complex; it wants to make even more money2. Fortunately, defence spending can nowadays be defined as building roads and other infrastructure. I look forward to a time when art and music can be created in the name of national security.
Of course I'm being facetious. Most of my music is not available on Spotify and won't be. If you're curious, you'll find much of it on bandcamp (this one, or the other one).
For serious analysis of geopolitical events, I recommend Neutrality Studies (whose clickbait titles are not representative of the nuanced discussion), Glenn Diesen, Judge Napolitano, or Kevork Almassian's Syriana Analysis video channels. Even a serious print newspaper like Le Monde Diplomatique with their long and careful articles replete with footnotes and written by people with academic credentials tend to display biases to which these video channels provide a healthy corrective. Although, in earnest, I would rather recommend reading books than spending too much time on youtube or any mainstream news media.