Even AI Understands Limits to Growth
by Cole Thompson
Artificial Intelligence, or AI, is attracting great attention as working AI systems become accessible to the public. The AI claim is that it can digest the mass of knowledge that humanity has made public, then perform cognitive tasks with that knowledge or answer questions with speed and accuracy. This has many implications, some potentially worrisome. But when AI works well, it can serve up some interesting “truths.”
Creation of …? (Mohamed Mahmoud Hassan, CC0 1.0)
While AI does not generate authoritative or definitive information—you wouldn’t bet your savings on its output—my sense is that its findings often deserve a hearing. So I thought I’d run a few AI engines through their paces on a question of great interest to you and me: whether perpetual economic growth is possible.
Several AI systems are available now for use by the public. The most well known is ChatGPT, but there’s also Google’s Bard AI and Microsoft’s AI. For this article I tested Bard and ChatGPT.
Hey, Bard, What about Perpetual Growth?
My question to Bard was straightforward: Can the economy keep growing forever? Bard’s answer was equally direct (the bolding is mine):
The economy cannot grow forever on a finite planet. Economic growth is measured by the increase in the total output of goods and services produced by a country or region. However, the resources available to produce these goods and services are finite. As the economy grows, it uses up these resources, which eventually leads to depletion and environmental degradation.
There are a number of ways to address this issue. One way is to switch to renewable energy sources, such as solar and wind power. Another way is to reduce our reliance on consumption and focus on more sustainable forms of economic activity, such as circular economy.
It is important to remember that the economy is not the only thing that matters. We also need to consider the impact of our economic activity on the environment and on the well-being of future generations.
In the long run, the economy cannot grow forever on a finite planet. However, there are a number of ways to mitigate the negative impacts of economic growth and to ensure that the economy continues to serve the needs of humanity in a sustainable way.
Circular economy in the sky? (Dalle-e)
I was fully expecting some hand-waving response asserting that “new ways will be found,” in an echo of my economics professor long ago. But looking at Bard’s answer, AI pretty much nailed it, although I would highlight a few areas where it needs to do more homework. For example, Bard may place more hope in the notion of a circular economy than it may deserve. Still, for a rookie analyst, Bard was solid in articulating limits to growth.
Your View, ChatGPT?
Next, I turned to ChatGPT with the same question. Can the economy keep growing forever? Its response (again, with bolding added by me):
No, it is not possible for the economy to keep growing forever. The earth’s resources are finite, and therefore, the economy’s growth is ultimately limited by these resources. Additionally, growth cannot continue indefinitely without negative consequences, such as environmental degradation and resource depletion, which can ultimately limit the economy’s ability to grow.
Furthermore, economic growth is dependent on a number of factors, including population growth, technological progress, and changes in resource availability. As these factors change over time, the rate of economic growth may increase or decrease, but it cannot continue at a steady rate indefinitely.
Therefore, while it is possible for the economy to continue growing for some time, it is ultimately limited by the finite resources of the planet, as well as the negative consequences of unchecked growth. Sustainable development and responsible resource management are important considerations for ensuring long-term economic growth and prosperity.
This is even better, at least the first paragraph. Here we have a blunt “No” at the very beginning of ChatGPT’s response. Again, I think an experienced steady stater would give a slightly better answer, but this was a strong first effort.
Another Thing…
As long as I had its attention, I pushed forward with another question: What would life be like in a steady state economy? Out came ChatGPT’s well-organized response:
In a steady state economy, economic growth is stabilized, and the focus shifts to maintaining a stable and sustainable level of economic activity that does not exceed the capacity of the natural environment to provide resources and absorb waste. Here are some possible characteristics of life in a steady-state economy:
-
-
- Stable consumption patterns: People would likely consume fewer goods and services overall, but the distribution of resources would be more equitable. People would still have access to basic necessities such as food, shelter, and healthcare, but the emphasis would shift to quality of life rather than constant material growth.
- Increased focus on sustainability: With a limited pool of resources, the emphasis would be on maintaining the natural systems that provide the resources we need for survival. This would include a focus on renewable energy, conservation of natural resources, and sustainable agriculture.
- Greater local self-reliance: A steady-state economy would require greater local self-reliance in terms of food production, energy generation, and other basic necessities. Communities would need to work together to create sustainable systems that meet their needs.
- Reduced inequality: The steady-state economy would require a more equitable distribution of resources, which could reduce poverty and inequality. However, this would also require a shift in cultural values away from conspicuous consumption and towards sharing and cooperation.
- More time for leisure: With a focus on quality of life rather than material growth, people would have more time for leisure activities, creative pursuits, and personal growth.
-
Overall, a steady-state economy would prioritize sustainability and quality of life over constant economic growth and would require significant cultural and institutional changes to achieve. However, many proponents argue that it would ultimately lead to a more just and fulfilling society.
Re-valuing the essentials of life in a steady state economy. (Ryan Johnson, CC BY-SA 2.0)
I love the feel ChatGPT gave of a steady state economy; it could do marketing for the concept! On the other hand, it erred where I would not have expected it to, in asserting out of the blocks that in a steady state economy, economic growth is stabilized. Not true: It should have said that in a steady state economy, the rate of economic growth is stabilized at zero. This error of fact is surprising in a logic- and fact-based system.
I suspect this error will get corrected in time, as AI answers constantly evolve. (You can ask the same questions multiple times and get different phrasing, although in my experience the answers point in the same direction.)
Another slight error is its use of a hyphen in “steady-state economy.” CASSE doesn’t do so, nor does most of the steady state economy community. The reasons are a bit complicated, but can be explored here. (And beyond this question of nuance, it seems odd that ChatGPT used a hyphen inconsistently in its response.)
Why Does AI Get It, and So Many People Don’t?
The Bard and ChatGPT responses are encouraging and rather surprising. AI systems have been trained to scour humanity’s biggest information commons—the Internet—and derive what is true and useful. Searching the heaps of online verbiage that endorse the idea of endless economic growth, AI systems pluck out the key insight that on a finite planet, endless growth is not possible or desirable.
I think three factors largely account for this pleasant surprise:
- An AI system does not have an ego. It does not care about “fitting in” or making statements that people like to hear.
- An AI system cannot be bought. It does not care about continued funding, keeping a job, or campaign contributions.
- Most importantly, AI systems are designed to weight authoritative and trustworthy sources of information much more heavily than sheer volume. As Brian Czech has long pointed out, steady staters “have two major allies: sound science and common sense.” Each of these is well-represented on the Internet.
Things that Could Go Wrong—But Probably Won’t
I’m a veteran of cybersecurity, and sadly, one thing I’ve learned is never to underestimate the capacity of humans to do incredibly selfish things, regardless of the cost to everyone else. So let’s take a pause—what could go wrong here? More specifically, what might powerful people do if they don’t like what AI is saying about economic growth?
I see two possible threats:
- Organizations or individuals might seek to “poison” the information commons with a flood of information that seems authoritative, and tilts AI analysis toward rejecting the steady state.
- Organizations might seek to mandate direct distortion of AI results. This might take the form of legal mandates that AI systems cannot question the status quo on certain topics, including economic growth.
However, I don’t think either of these tactics is going to be easy to arrange. One of the priorities in the AI business is resisting information-poisoning efforts. It’s crucially important for useful AI, and it’s interesting work for AI practitioners. The more various actors try to poison the information commons, the more “training corpus” (raw material to learn from) will be available to AI and human assistants to analyze and adjust to. I will go out on a limb and predict that eventually the poisoners will run out of tricks that work, and AI engineers and scientists will win this battle for the truth, because they want to.
Regarding direct, legally required distortion of results, I think the payoff is too vague to justify the cost and risk to powerful elites. To be sure, economic elites go to great lengths to protect assets they see as threatened, as documented in, for example, Confessions of an Economic Hit Man. But the benefits to the status quo of an assumption of infinite growth are benefits that are very general. It’s hard to imagine macroeconomic theory being enough of a direct threat to any single individual’s wealth that laws on AI answers would be passed by chest-thumping politicians.
Even if I’m wrong about that, explicit meddling in AI is the sort of thing that gets out and talked about in Silicon Valley and beyond. Just one revelation that AI is intentionally lying would create a firestorm of interest and questioning—exactly the opposite of what status quo economic pundits want.
Is she real? (Unsplash, Andy Kelly)
The worst case would be if AI were legally required to distort truth about economic growth, and if people just shrugged and accepted it. That might be reminiscent of the Snowden revelations about the National Security Agency’s Prism program, when mass surveillance and analysis of internet communication was revealed: There was some controversy, then people seemed to forget about it and move on.
Or did they? In fact, people in the USA and around the world did take note and adapted their behavior. In the hypothetical case of deliberately distorted AI responses on questions of growth—and frankly, on the survival of humanity—word would get out, and people would adapt, not necessarily their behavior but their values and assumptions. Once you know you’ve been intentionally lied to, it’s hard to forget, and hard not to think about that particular lie. This would again be exactly the opposite of what conventional economic growth pundits would wish for.
The Lesson for Steady Staters?
First, pat yourself on the back; you earned recognition and kudos. It’s not often one can look at something that really matters to humanity’s future and say, “I made a difference.” It is precisely because CASSE (and others) have upheld high standards of analysis and discussion that artificial intelligence now looks at the world and understands that endless economic growth is not possible. I feel gratified, and I hope others do as well.
My final thoughts are:
- Keep doing what we’re doing. Keep our standards of analysis, discussion, and publishing high. Be trustworthy.
- We have probably gained a powerful ally with the rise of artificial intelligence. When confronted with human inertia and entrenched beliefs in perpetual growth, let’s not be afraid to refer people to AI for an “expert opinion.” It might just speed up the epiphanies that humanity badly needs to make it through the 21st century.
Cole Thompson is a cyber security professional and the Sierra Nevada chapter director for CASSE.
Great piece, and surprising that AI picked out the “right” answers in light of the massive amount of pro-growth information around. I’ve shared it on my Substack and FB pages (https://erwindreessen.substack.com/publish/post/113282222).
One quibble about that hyphen, though. As the referenced page from the Encyclopedia of Earth says, “the conjoined “steady-state” is a heuristic tool to imply the stable ratio of capital:labor and, linguistically, is an adjective of “economy.” ” A state’s economy may be steady alright, but the key point is for the economy to be in a “steady state”! Therefore there should be a hyphen.
Thanks Erwin. We’d been thinking of an AI piece for some time, and Cole stepped up to the plate.
Not sure I understand your second paragraph, though. At the “steady state economy” entry in the Encyclopedia of Earth (an entry I authored) you’ll find the rationale for linguistically distinguishing the ecological, Dalyist, unhyphenated “steady state economy” from the neoclassical, Solowian, hyphenated “steady-state economy.” I’ll repeat it:
“It is appropriate to use the unhyphenated phrase “steady state economy” to describe an economy of stable size because “state” (as in political state) is an adjective of “economy” (as in a state’s economy), and “steady” is an adjective of this state economy. In other words, “steady state economy” typically refers to a national economy of stable size, although it may also refer to an economy of a city, province, or other political unit. (It may also refer to a regional economy or the global economy, and in such cases political units are aggregated.) In neoclassical economics, “steady” is not an adjective of “state economy.” Rather, the conjoined “steady-state” is a heuristic tool to imply the stable ratio of capital:labor and, linguistically, is an adjective of “economy.””
I stand by the distinction, although I’ve never been adamant about it. Frankly, an ecological steady state economy is likely to also be (with regard to the capital:labor ratio) a special, non-growing case of the neoclassical steady-state economy. But, 17 years ago (when I wrote for EOE), the distinction was more important, because the ecological, Dalyist vision of a steady state economy was way outside the vernacular. The neoclassical, unsustainably growing “steady-state” economy had the inside track. Perhaps this is no longer the case. (Thanks partly to CASSE?)
We’ll continue to promulgate the non-hyphenated “steady state economy” without falling on any swords over it. The linguistics are sound. “State” modifies “economy” and “steady” modifies “state economy.”
Not to belabour the point, but I can use your own words (and yes, I read the entry in the encyclopedia) to argue against a hyphen! The steady state concept can apply to any level of economy, as you say. So “state” does NOT refer to state as in nation or country. Rather, we aim for an economy that exhibits a “steady state” (steady ratios). Therefore, “steady state” is an adjective of “economy.” To make that unambiguous, we should write “steady-state economy”!
Thank you Erwin!
This article is an eye-opener, as it altered my thought process on the subject if AI. The approach is brilliant: testing AI with a real subject. For once, an objective way to become optimisitc.
Just read that Alibaba are on the way to launching a Chinese AI. Ability to write business proposals one of its listed benefits. The efficiencies to be created by AI will be spectacular, driving even more unsustainable growth. What also worries is the sheer volume of written output that AI can generate. Marketing material so specifically tailored to its target market, maybe tailored for individuals. A tsunami content potentially drowing our messages of restraint.
Yes, there’s definitely a possible downside to AI. I’m keeping my fingers crossed that the useful answers we’ve seen so far on economic growth will endure.
I wonder how AI would respond to a question about the cause or “Why?” of its response. Would it get to the 2nd Law?
These AI tools may adjust their responses in line with the internet history of the questioner. So I would be curious to see if you got similar responses if you tested them again from the computer/browser of someone antithetical to the cause. Are you on amicable terms with any such person?