It would be great if we could answer important policy questions using algorithmic information theory. For example, if I wanted to predict when the world population would peak, I would find various functions that approximately fit the attached graph (annual rate of change of world population) and add the size of the function's encoding to the log of the absolute value of the prediction errors to get the total size of the encoded data in bits. Then I would weight each function by 1/2 to the power of the bit length to get a probability distribution of the zeros. Just eyeballing the graph, I would guess there is a cluster around 2050-2080.
You could do this country by country and maybe get a more accurate number. What you can't do is infer causality. A compressor would notice that low GDP is associated with a high birth rate, but it won't tell you which is the cause of the other because the information content is the same either way. Or would even more data help? Suppose I compressed all the world's knowledge that is publicly accessible by a web crawl, around 15 TB of text. Surprisingly, we can already do this. I asked DeepSeek "when will world population peak?" It answered: The exact timing of the world population peak is uncertain, but most recent projections suggest it will occur around **2080–2100**, with a peak population of roughly **10 to 10.5 billion people**. ### Key Projections: 1. **United Nations (2022 Revision)** – Peaks around **2086 at 10.4 billion**, then stabilizes or slowly declines. 2. **IIASA (International Institute for Applied Systems Analysis)** – Suggests a peak around **2070–2080 at 9.8–10.2 billion**, depending on fertility trends. 3. **The Lancet (2020 Study)** – Predicts an earlier peak (~**2064 at 9.7 billion**) followed by a decline due to falling fertility rates. ### Factors Influencing the Peak: - **Declining Fertility Rates** – Many countries (especially in Europe and East Asia) are already below replacement level (2.1 children per woman). - **Economic Development** – As more nations urbanize and educate women, birth rates tend to drop. - **Healthcare Improvements** – Lower child mortality leads to smaller family sizes over time. - **Climate & Policy Changes** – Food security, migration, and government policies (e.g., China’s former one-child policy) could shift trends. ### Post-Peak Decline? After peaking, the population may gradually decline, leading to challenges like aging societies and labor shortages. However, this depends on future fertility rates in Africa and South Asia, where growth is still high. Would you like details on specific regions? -- Matt Mahoney, mattmahone...@gmail.com On Mon, Apr 21, 2025, 2:21 PM James Bowery <jabow...@gmail.com> wrote: > "Policies for the People" is supposedly a place where MAGA can let off > steam and be heard by the Trump Administration. > > Not so certain about that but here goes... > > > https://forum.policiesforpeople.com/t/reform-macrosociology-with-the-algorithmic-information-criterion-for-macrosocial-model-selection/25812 > > On Mon, Apr 21, 2025 at 8:52 AM James Bowery <jabow...@gmail.com> wrote: > >> It is worth noting that among the various models, DeepMind's model was >> most prone to the specious pedantry -- entirely missing the point that >> there is a difference between model generation and model selection. >> >> Google Strikes Again! >> >> On Mon, Apr 21, 2025 at 8:49 AM James Bowery <jabow...@gmail.com> wrote: >> >>> https://claude.ai/share/df6fdf28-b73b-48b7-9f5b-01b265e3748c >>> >>> Prompt: >>> >>> "Provide a counterfactual analysis of a history in which starting with >>> the publication of Solomonoff's papers on inductive inference, it was >>> recognized that approximating the Kolmogorov Complexity of a wide-ranging >>> selection of longitudinal macrosocial measures was recognized by the NSF as >>> the gold standard information criterion for selecting among candidate >>> causal models of macrosocial dynamics, and paid out its funding for >>> sociology only to those who decreased the size of the executable archive of >>> that standardized dataset." >>> >>> Anyone with an information technology background that holds strong >>> political opinions that can read this and still be tolerant of "the >>> experts" to whom they defer, has no idea what information actually is. >>> >>> It's amazing how stupid people are when it comes to this topic. The >>> arguments that inevitably come up time and time again are, when >>> appropriately framed, specious pedantry so transparent that one is hard >>> pressed to describe them as anything but obstructionist fear of truth. >>> >>> "CrimeStop" comes to mind: >>> >>> “Crimestop means the faculty of stopping short, as though by instinct, >>> at the threshold of any dangerous thought. It includes the power of not >>> grasping analogies, of failing to perceive logical errors, of >>> misunderstanding the simplest arguments if they are inimical to Ingsoc, and >>> of being bored or repelled by any train of thought which is capable of >>> leading in a heretical direction. Crimestop, in short, means protective >>> stupidity.” >>> -- George Orwell, 1984 >>> >> *Artificial General Intelligence List <https://agi.topicbox.com/latest>* > / AGI / see discussions <https://agi.topicbox.com/groups/agi> + > participants <https://agi.topicbox.com/groups/agi/members> + > delivery options <https://agi.topicbox.com/groups/agi/subscription> > Permalink > <https://agi.topicbox.com/groups/agi/T6e1019516c8638e9-M61e236fcfd6852cb3b58e279> > ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T6e1019516c8638e9-M8b2898191ff863a67f4ebbfa Delivery options: https://agi.topicbox.com/groups/agi/subscription
