Motivation

Beliefs about the future structure every economic decision any actor does take. Thus, a change of beliefs is bound to change decisions and behavior. From the perspective of AI safety as a discipline, the foundational belief about the future is that there will be a point in time when recursive self improvement will enter a closed loop, and the range of possible futures will fundamentally widen. This assumption is currently not shared with the majority of society.

How the current techno-economical situation relates to this belief is a debated topic and extremely well discussed in “Where’s the “intelligence explosion”?” by Ramez Naam. How much it changes geoeconomics and geopolitics is laid out by Anton Leicht. What can not be debated is the amount of influence this mental model already had. Understanding it can restructure one’s outlook on what is actually possible in those futures, and the least that can already be said is that this has significantly changed a lot of career trajectories of young, talented and ambitious people.
I would argue that this is due to a combination of emotional, normative and economic reactions to this change in belief of the space of possible futures:

Process

In our project, we tried to use the tools of economic theory and modeling to understand what difference this belief exerts on decision space and behavior of individual actors. We read wide to understand the theoretical implications of RSI on long run economic growth (I plan to write another post on this paper specifically), the present and short term implications, rigorous models of labor market impact ( I still need to read “Messy Jobs”) and if and how labor markets are already impacted.
The underlying assumption was that using these methods would enable us to disentangle these 3 reactions from each other and clarify analytically what exactly this belief changes through 2 channels: increased range of widely disparate future scenarios, and increased uncertainty.
In a second step we applied decision theory with traditional utility maximizing behavior on this problem set in a normative agnostic way.

In this process, we encountered a number of noteworthy problems

1. The belief in RSI disposes every actor who perceives themselves to have a shot at triggering and “owning” this process to understand themselves as in a race with everyone else they deem a potential contender. This is, given the justified beliefs about potential payoffs, even when facing a discount rate, economically rational1

2. One of the reason why Ramez’s post on the empirical evidence of takeoff potential and reality is so critical results from the fact that the assumed time until takeoff leading to branching timelines influences expected utility calculations significantly

3. The range of future scenarios are very hard to fit into the standard models and methods of economics (to the extent that maybe aggregating over averages be very reductive and explicit discrete scenario planing a much better approach), and had a number of interesting implications:

   1. Once extinction level events enter the possibility space, utility value assignment becomes extremely hard to model non-normatively

   2. The first best economic reaction to such a problem individually would be to buy insurance. Extinction scenarios are not insurable as there is no credible counterparty.
      This implies the impact of beliefs onto behavior is much more pronounced as individuals try to smooth their expected utility curves through behavior influencing the distribution of futures and their societal and individual impacts

   3. A number of scenarios that most normative theories would judge bad do not necessarily look so through the lens of a potential investor for example. We iterated a number of times trying to keep the model itself value neutral and letting actors explicate their normative preferences, which worked but it felt like syntactically correct code doing something semantically incorrect

   4. Uncertainty about ones own beliefs actually has a smoothing effect on expected utility curves as well because the divergence of possibilities is so wide, and long time horizons have a number of implications that advise intense epistemic humility2

Output

We then operationalized these results in a web app to make the gained analytical clarity accessible to others, asking users to explicate their beliefs about the future and the values they hold:

Using those inputs to produce a coherent map of their specific space of different futures and which of the 5 assumed options when choosing their career they should pursue according to their own criteria:

We had to use a lot of simplifying assumptions to shrink possibility and decision space in order to enable concrete guidance and ease informational overload instead of increasing it. This pressure towards simplicity lead us to ask users to specify their normative stances and their impact on non-monetary utility in monetary terms, exemplifying extremely well the intense tension between economics and ethics.

SAIGE was kind enough to allow us to use their domain for the web app, which you can try out right now.

Darlings killed

Topics we considered but axed due to time constraints and complexity:

1. Given the importance of actions vs insurance, is it ever economically rational to invest to change distributions of futures? (probably yes, but complicated to model and very value judgement dependent)

2. If belief in RSI is expectations destabilizing due to higher uncertainty, could there be a role for state capacity forecasting expertise to stabilize expectations similar to central banks anchoring inflation expectations?

   1. How would this work in theory and practice?

   2. What would be the distributional implications, specifically would there be a difference between labor and capital markets?

   3. What could be benefits apart from reducing friction and anchoring? For example, could technocratic forecasting expertise in semi-independent statistical agencies be useful to mobilize private capital needed to finance compute expansion of middle power strategies such as laid out in the latest

      “Transformative AI strategy for Europe”? If yes, under what conditions?

Outlook

The project was intensely stimulating, intellectually rewarding and made me extremely curious and very happy.
I would like to thank very deeply everyone who made it possible for me to take part in this wonderful exploration:

• The SAIGE team:

  • The founder, Jessica

  • Manon Kempermann, who organized the incubator program

  • Johannes Koch, for providing scientific guidance and is now managing the coming SAIGE fellowship program

• My project Mentor Pavel for weekly guidance clearing up my manifold confusion

• Casimir, my co-mentee who provided the policy lens complementing my econ and technical background
  

The incubator was in fact such a great experience that I’m currently doing my next AI safety related project, “Mapping Cost Trajectories of AI-Adjacent Technologies”.
I work in Berlin at the co-working space of AI safety Berlin, feel free to reach out!

2

https://economics.princeton.edu/wp-content/uploads/2025/12/EUT_long_run-Steiner.pdf

A job is a bundle of tasks, and those tasks compose multiplicatively, such that one weak step in a chain caps the entire output. The object of productivity gains are not jobs as atomic units, but the tasks they are composed of. Whether that means your role disappears, transforms, or quietly becomes more productive depends not on how powerful AI is in general, but on the shape of your particular bundle: which steps are O-ring and which are additive, and where AI is strong or weak on each.

The rest of this post lays out: what a job actually is form this perspective, how to read a task bundle for AI exposure, and why the same AI capability produces displacement in one role and empowerment in another.

Here’s the tl;dr:

First: what actually is a job?

The obvious definition of a job is about performing a set of tasks within a social relation with the goal of getting paid. For the purpose of analyzing productivity shocks it is helpful to add a compositional dimension to this: What does a person with a given job title actually spend their paid time doing, and how does that connect to what their employer is paying for? Each atomically coherent unit of those processes is called a task, enabling us to disaggregate a job description into a bundle of tasks. Displayed here are two examples of how a specific job description contains a number of different categories of tasks, like Software Engineer or Management Consultant.

This addition to our analytical categories enables a more fine-grained understanding of

1. Competence

2. How a productivity shock might impact the labor market

The “a job is a bundle of tasks” definition allows us to understand what it means to learn a new job, retraining and how we judge the competence of others:

Adding the process layer

The “coherent integration” did a lot of the heavy lifting in our definition. Another way to “slice” a job is by the desired outcomes and the processes that generate them. Luckily, we can use the task taxonomy on labor processes here as well and categorize each individual step along the process into one (or more) task categories.
A software engineer might be tasked with a number of specific value generating activities like feature development, customer issues, mentoring & onboarding, and each of those can be decomposed into task categories.

One additional thing we can do once we have a labor process modeled as a sequence of specific tasks is define their relationship to each other and the final desired output. A lot of modern job processes have what is called an “O-ring” structure, named after the rubber seal whose failure caused the Challenger disaster. One component, costing almost nothing relative to the whole, failed, and the entire mission was destroyed. The economist Michael Kremer formalized this intuition: in complex production processes, output is multiplicative across tasks, not additive. A bottleneck anywhere caps everything.

Applying a productivity shock to this mental model leads one to the following, focusing exclusively on the supply side (the demand side is another source of optimism, once we do not take its amount to be fixed):

• The object of (AI based-) automation are tasks, not jobs

• Different types of tasks will have differing degrees of automation potential

• Thus, a job [title] might continue to exist but its underlying task composition change quite radically

• How AI automation is changing the wages of individual workers is dependent on the structure of the task in the labor process

  • If a process consists mostly of additive elements with high automation potential, the time spend on them will shrink

  • The overall task composition will then change, as more time of the worker is spent on O-ring process elements

  • A job is at risk of obsolescence if most of its processes are additive and it cannot find new O-ring processes with high value added that were not previously part of it. This would constitute labor displacement

  • If a job contains sufficient amount of O-ring processes whose O-ring elements have little to zero AI automation potential, individual worker productivity will rise and subsequently their pay. This would constitute labor enhancement

In summary, automation potential of a task is not the single factor determining its effect on the labor market. Based on this taxonomy, we can then provide fictional, stylized examples of impact on actual jobs as task bundles.

This is still just to demonstrate how this mental model helps to think the impact through more rigorously. The numbers are purely to demonstrate the structural logic and getting an intuition for what it might look like.

Example scenarios

Intense displacement: Rule based para-legal work

Each bar is the probability of doing a task well. The workflow’s quality compounds across tasks — one weak step drags the whole output down — so totals are the product across all five tasks. On each task, the employer assigns whoever scores higher, so Human + AI takes the max per task.

• Human alone: 0.65 × 0.75 × 0.80 × 0.72 × 0.82 = 0.23

• AI alone: 0.95 × 0.92 × 0.88 × 0.90 × 0.72 = 0.50

• Human + AI: 0.95 × 0.92 × 0.88 × 0.90 × 0.82 = 0.57 (human only swapped in for judgment)

The human is retained only if their marginal contribution (0.57 − 0.50 = 0.07) exceeds their wage relative to AI’s cost. A 10-point edge on a single task isn’t enough of a moat once it’s diluted across the workflow.

This is the scenario that is most often assumed to be true for most, if not all jobs, in a not too distant future by those who are rather alarmist.

Middle Ground: Content generation

In the middle ground example we have the following numbers:

• Human alone: 0.70 × 0.75 × 0.65 × 0.78 × 0.80 = 0.21

• AI alone: 0.88 × 0.85 × 0.90 × 0.52 × 0.45 = 0.16

• Human + AI: 0.88 × 0.85 × 0.90 × 0.78 × 0.80 = 0.42

  (human swapped in for editorial judgment and audience strategy)

The human’s marginal contribution (0.42 − 0.16 = 0.26) is much larger than in the displacement case because AI is genuinely weak on two of the five tasks — judgment and audience strategy. The job stays, the role transforms a lot: routine tasks take much less human time, critical O-Ring tasks in each chain start to dominate the work day.

Worker empowerment: Consulting Work

In this stylized example of labor augmentation, the values are:

- Human alone: 0.72 × 0.70 × 0.82 × 0.85 × 0.88 = 0.31

- AI alone: 0.85 × 0.80 × 0.70 × 0.30 × 0.10 = 0.01

- Human + AI: 0.85 × 0.80 × 0.82 × 0.85 × 0.88 = 0.42 (human swapped in for the three tasks they win)

This is the upside scenario that is to some degree (absent specific, well targeted and implemented policies) the necessary condition that would enable distribution of AI productivity gains to the whole population apart from simple consumer surplus.

Phenomenologically, this might feel like an intensification and densifying of the work: Workers are making critically important, complex and complicated decisions at a much higher number and frequency. A lot of early adopters of agentic coding describe this emotional reaction to their changed work environment and task composition.

Outlook

My impression of public discussions on labor market impacts is that most economists implicitly assume this mental model of jobs as complex task bundles, while most unfamiliar with them assume some kind of intense displacement scenario based on the mental model that AI will literally automate jobs, not tasks.
This complicates the empirical picture, meaning many interesting open empirical questions for research! And additionally more upside for most workers than the displacement story suggests.

What does the future hold for jobs with exposure to AI induced productivity shocks?

• Tasks where AI is good enough to be worth the time savings take much less time, enabling human workers to either do much more of them and keep the percentage stable or spend proportionally less time on them

• Jobs where O-ring tasks are already abundant will just have their percentage of time spent increase massively

• Jobs with up until now little O-Ring tasks will get either automated or find more O-ring tasks, either by intensifying the current setup, or adding new tasks to their process chains or completely new processes

On rollout and impact speed

One of the questions that this thinking puts on hold is about the speed of change and the feedback loops societies typically need in order to rearrange dynamically integrating the change. In labor markets, this is about people leaving the labor market due to retirement, exiting a job and entering a different field with (hopefully) better opportunities and pay. Implicit in our current institutional arrangements are assumptions about maximum speed of change. makes this point very explicitly in this Odd Lots podcast on exactly this topic.

If the current (accelerating) rate of LLM model capability improvements gets directly translated into labor market impact this up until now unknown speed of change could overwhelm a lot of stabilizing parts of current social arrangements and institutions.
But that’s a very different question than the original one, the answer to which is:

"Which is why 'will AI take your job?' is the wrong question. The right one is: what's in your job, and which parts can it actually take?"

Literature / Links

Acemoglu, Autor: Skills, Tasks and Technologies: Implications for Employment and Earnings | NBER+

Kremer: O-Ring Theory of Economic Development* | The Quarterly Journal of Economics | Oxford Academic

Acemoglu: The Simple Macroeconomics of AI | NBER

Eloundou, Manning, Mishkin, Rock: GPTs are GPTs (Science)

Dell'Acqua, McFowland III, Mollick, Lifshitz-Assaf, Kellogg, Rajendran, Krayer, Candelon, Lakhani: Navigating the Jagged Technological Frontier (HBS)

Brynjolfsson, Li, Raymond: Generative AI at Work | NBER

Code can be found on here.

Demarking the subject

My understanding of section 1.1. is as an introduction into the subject via demarkation vis-a-vis already existing ones, outlined in the table below. I have also added categories from where I understand the topics to be originating from in order to fix and show my mental model of the topic we are beginning to talk about.

I found the sequencing of the introduction confusing as we start with the individual motives for having children which I understand to be broadly ethical and cultural and switch rapidly into a lengthy discussion of the economic intentions and consequences of decisions by parents concerning their intended and realised fertility. Following we switch back to a more rigid formulation of the ethical questions we started with. The (sneaky segwaying) addition of Question (iv)(p.6):

How should one value possible polulations so as to decide which would be best?

leads Dasgupta to Parfit stressing the difference between their opproaches. His main criticism consist of the denial of socio-ecological constrains in Parfits’ arguments, a criticism I find reasonable. I read Pasgupta as positioning himself in clear opposition to Parfit, but at this point of the book I would not agree to his understanding. Considering constraints is the definition of economics, regardless of the topic this consideration is applied on. If the optimal number of people in a society would not exceed existing socio-economical limits, we would never find out about them, and they would not be binding. Understanding socio-economic limits as normatively binding independent of the externalities for other present or future persons implied in their overuse seems to me a clear instance of the is-ought distinction. Of course Dasgupta would react to this critique by citing as the source of normativity the actual and potential well-being of present and future persons.. My main point here is that I do not think that Dasgupta works in the same field as Parfit, and I much rather read this book than one by Parfit.

This leads me to the following taxonomy of the topics discussed here:

I named the second column to demark the difference economics would typically be concerned with: micro and macro. I want to stress that this categorisation does not imply any kind of hierarchy - I do not think that practical philosophy is all we will ever have to say about ethics and that the idea of ethical expertise does not stand the empirical test, either by akrasia or worse, leaving me to the conclusion that a dialogue between disciplines and viewpoints concerning ethical considerations to be a wonderful and enriching thing to have.

Futher adding to my confusion was the fact that we get a short introduction into the global demographic transition with the first table strongly suggesting the concept of differing regimes between which populations in their dynamics switch, even mentioning the respective selection theory and regimes having differing evolutionary stable equilibria - I find interesting that this fact is mentioned in the footnote but the main text concerning externalities and the tragedy of the commons does not elude to the game theoretical dynamics governing those. Combining the topic of the book with the demographic transition, I just want to stress that the revealed preference of inhabitants of advanced economy is to have less children, leading an economist regarding the choice for or against children as a choice between different goods to classify them as an inferior good. Apart from the calculations, models and normative assessments regarding the optimal size of populations, the individuals actually making choices seem to have clear priorities.

Dasgupta then lays out the program for the rest of his lecture after his introductionary remarks:

1. Build a model

2. Feed data into the model

My understanding is that the following sections until the start of part one constitute in principle a kind of ontology and other considerations which the model will be based upon or trying to incorporate.

Utilitarian Ethics

I will first try to formalize how Sidgwick is discussed and then what modifications Dasgupta already establishes. I will also try to differentiate differing objects of analysis, namely individual I, actions a and societies S.

Supposed we wanted to model an individual life L the sum of events added chronologically:

Following utilitarianism the way to determine total individual utility U of a life would be by assigning a happiness value h to each event and summing up:

with h(n) having possible values positive to negative infinity (for simplicities sake).

Shifting the object of analysis to individual actions, we would then calculate similar to overall utility of a life:

with a, the first person to consider and b the total sum of people being affected by the given action. Notice how the meaning of the calculus changed without the equation at all. Furthermore we can use U^a to compare potential states of the world to each other as well - that is after all exactly what we do when considering the desirability of an action within a utilitarian framework.

Already we can see that selflessness is mandatory within utilitarianism, as there is no privilege given to the person supposed to do the action in U^a.

Ranking alternative states of the world between each other also becomes easy as we established a uniform currency in happiness into which we assume to subsume every relevant aspect of judgement, leading to a simple way of evaluating actions and states of world in which more utility is better and equal amount of utility denotes indifferent states between which no judgement can be reached upon. Let us assume together with Dasgupta that our criterion for happiness is sufficient complex to avoid all pitfalls for the arguments' sake. I do think that the addition of informed desires can do a lot of lifting by inserting a lot of enlightenment concepts concerning for example self-inflicted tutelage in combination with higher-order volition.

Similar to our simple summation operation computing the value of a single life, we can judge whole societies by summing the overall utility of all members:

Starting at a random person c living within the society and ending with last person d. Of course the same functionality regarding states of the world from U^a applies here as well: we now have a way of determining not only a positive society but also potential changes to it enabling us to do normative prescriptions in addition to descriptions of the world. Of course, we achieved all this by assuming away all problems of epistemics, perspective, individuality, legality and a dozen other things but we want to model with a certain goal in mind to get a better idea about some of our intuitions, and we will discuss ontology and methodology at a later point in the book where we can pick up what we left here. Besides the qualifications of informed desire and using well being instead of happiness I find the discussion about the necessity of consumption of goods and services to achieve a state of happiness noteable. I personally would be quicker to admit this as an epistemic problem we as human societies have not solved satisfyingly, but have no quarrel diagnosing empirically the consumption of goods and services as a necessary condition for happiness in a utilitarian sense for most people and activities in such a way that assuming the corner cases away seems reasonable. To me the epistemic lens of this is much more interesting than potential ethical dilemmas regarding this specific model. The epistemic considerations lead to lots of complicated normative questions of a totally different kind as Mazzucato pointed out.

To summarize, Dasgupta lays out his plans to modifiy what I laid out of Sidgwick here in the following ways:

1. Modifying desires into informed desires to avoid pleasure automata (already done here)

2. Shifting the unit of analysis from a temporal dimension to a material one (or simply adding it to the other) by introducing goods as a necessary condition of the fulfillment of informed desires (both done here on page 7-9)

3. Declaring his intention to at a later stage change the aggregation function for overall well-being by inserting a model of generations,

4. as well as a different understanding of the “neutral” life

Theses four changes to my understanding will lead to his theory of Generation-Relative Utilitarianism.

Ends and Means

This chapter deals with the second point of the list above - the material conditions within which well-being is to be realised and assessed. It starts by introducing in essence an ontology upon which the following modelling will be based. I love how the point of activity and services is stressed throughout, focussing on the dynamical nature of the object of economic measurement generally and the meaning of GDP specifically. I would like to point out that though GDP is mentioned, Dasgupta sets up the whole exercise as I understand it in opposition to conventional economics and/or econometrics, using the tools it itself uses to work on problems that will probably lead to different sociological, political and ethical positions.

Capital Goods

For my own understanding the first distinction to make is that between Economic Activity at a certain point or space in time t and that part of human activity that does not fall in that category, a distinction that Dasgupta in my reading of the text just assumes - if he does not, it is at least needed for me as a helpful frame around the modeling. As non-economic human activity is not concerned with life processes that are relevant for axiology, I find this reasonable.

Economic activity is then split up into services and assets which is derived from their relationship to time: Services being perishable in the sense that the act of consumption of a service through its temporality leaves other people at a later time unable to consume the same unit again. I find fascinating the use of metaphor describing the value of an asset as the (discounted) expected value computed as the services it will be able to perform in the future. Taken seriously, this would make goods ontologically dependend upon services which would be the fundamental category of economics shifted? The emphasis on dynamics is mentioned before and quite new to me. I find it pretty helpful, even if this is not a paradigm or intuition shared by many.

Services as an category seem to be assumed to be understood without further need of decomposition, but assets are further separated into the sum (S) of all types of capital and enabling assets. Here we see a rather stark contrast to conventional economic ontology, classifying financial means not as capital (which would be a lot of peoples intuitive example of capital, and all other forms of capital being ontologically dependend upon it or at least metaphorically derived from there), but as an enabling asset. I think we can see a lot of foreshadowing here how different the conclusions of the developed model may be from typical models that are to be contrasted here.

Following some possible types of capital categories are discussed and we arrived at three different forms: natural, permanent and human capital.

Dasguptas argument for limiting the categories of capital to those three as the ones that we are sufficiently able to price/measure is for the purpose of modeling sensible (even though this is a practical argument). I found Bourdieus’ differeing forms of capital extremely helpful as an epistemic tool understanding society, especially when combined with, paradoxically, Luhmann understanding different forms of capital simple as the corraly of different kinds of currencies one can spend in different systems and in part even convert into each other, but I would consider this a “soft” theory of which I would not be able to give an intuition how to build a model out of it. Dasguptas point here is that a lot of economic activities consist of converting at individually profitable rates different forms of capital into each other, most prominently natural into produced, and produced into human (at least how I picture it intuitively). Even if societal or (softer criterion) individual conversions may be profitable (we will get to well-being soon), it seems intuitive to assume differences between the types of capital. Natural capital may have regenerative aspects to it (gathering firewood out of a forest without cutting trees), enabling a potentially steady state of extraction. Produced capital seems to not regenerate and always degrade, if at different and even negligable rates (roman concrete). Human capital of course is the hope of decoupling, could be potentially infinite and enable non-zero-sum economics even in the field of usage or access in contrast to produced capital where the question of distribution of delivered services will always be at least implicit political.

I visualised the nesting of categories in order to get a better overview (Code available at the repository for this review:

Inclusive Wealth & social well-being

After discussing the means before, we now turn to the ends of it all: How should the extension of h in our starting point for utilitarianism be integrated into the model such that we can use happiness as a currency? (I repeated the formula from above)

The question comes up because what we absolutely do not want to do is simply rebuild a typical (macro-)economic model with money as the central accounting mechanism. To me the answer to this question seems to be given on page 15., with the answer being well-being and it being described as:

A person’s well-being is shaped by the extend to which her projects and purposes are realized. They in turn are rooted in her engagements, both with her own self and with others [..], which means account prices of goods and services are person-specific. It also means that the degree of fairness in the distribution of well-being influences accounting prices.

I would like to make several comments to that:

1. We can see here how in this model wants are theoretical unlimited, but means will always be - the typical problem of economics dressed (down) in utilitarians cloak.

2. There is (yet) no theoretical reason why more basic desires should be prioritised over more complex ones - like our current economies do, where rich people can price out poor people as a result of different means, unrelated to the urgency of their need.

3. There is now clearly a need for temporal discounting, which I will lay out below.

4. The constant stressing of the difference between “real economy”, how things are run at the moment and this model brings me to the merciless normative demands utilitarian poses to people. It is a really demanding normative theory!

5. Combining the third point with the point from the quote about person-specificity: Both the normative demands and the nuance are then just assumed away by modeling the economy with just one representative agent over an infinite time horizon.

6. My impression is that just establishing well-being as the truly relevant moral end to strive to is already revolutionary enough - never mind the inclusion of resource limits or moral considerations for future generations, at this point there is already such a chasm between what is described and implicitly prescribed that is difficult to bear on its own.

The role of temporal discounting:

Let’s imagine we wanted to implement a model how we know up until this point and do a genealogy to finetune if the model worked the way we think it should, and forgot temporal discounting. No primitive accumulation would happen and only rent extraction of those assets would be possible / permissable, as the undiscounted expected value of a natural asset over infinite generations would be infinte, and the undiscounted expected value of the resulting produced capital would necessarily be not infinte due to its main characteristic being degradation over time. No amount of faster population growth could justify the destruction of natural capital without temporal discounting. Amusingly enough, this would result in a behaviour that could be from the outside characterized as animistic, holding most of nature sacred in such a way that may feel like having a personal relationship with certain aspects of it. So we can see here that temporal discounting in effect enables the model to have something like a “developmental” stage of an economy where natural capital is unsustainably converted into other form of capital - with the implied hope that our children will be able to use those other forms of capital to get the whole system into a steady state or even make whole what was lost before.

At the same time, a more complex theory of natural capital would of course try to implement something like local or even global tipping points, with the collapse of the amazonas rainforest due to the collapse of its own climate feedback loop the most prominent example of a combination of both levels at the same time.

Inclusive Wealths’ radicalism

We call the social worth of the three classes of capital goods inclusive wealth [which] [..] includes human and natural capital

At this point my mapping is that within a given society inclusive wealth is counted by adding social worth of all capital goods up (ignoring markets here!), and that social worth does not only consist of the amount of self-realisation (see the longer quote above) of persons living right now, but also in some way the value of the self-realisation of persons not yet living.

That the social value of a produced capital may vary with its ownership or type of access to it (as in the piano having a higher value to someone actually being able to play it) ignores typical market-centered valuation and does not even try to talk about pareto-frontiers. Similarly the point about institutions and practices being a key determinate of the worth of capital within a society - both these points imply that a social planer could increase social welfare by a lot simply by changing the distribution of property and structures of states, social practices, mores, and much more. This is to me on the one hand empirically correct, and still such a bold normative position to take. Serious utilitarianism does not make prisoners!

At the same time the modeling does not talk about distribution, and I would argue that a lot of the defined relationship between social well-being and inclusive wealth rests on the fact that this claimed to be ceteris paribus true, with the c.p. doing the actualy heavy lifting. It also assumes away the (epistemic) compexity of distributional choices within a society: who gets to keep the piano? Being a perfect social planner is really hard, apparently.

How to use it

We did all this contrasting with accounting prices to get a different notion of investment is my impression, especially with the inclusion of natural capital. The strength seem to be that defining it is easy: Every decision made that increases inclusive wealth and consequently over time social well-being is a sound investment, and the opposite being described as a degradation of the capital stock of a society. I love how we now have a heuristic that is utterly familiar and alien at the same time, by using a concept already established and applying it to another topic - a move very similar to the discussion regarding utilitarianism above where the same definition could be used on different objects to solve similar yet different problems. I wrote about the complex trade-offs regarding social investment of societies above when talking about discounting and would like to continue here.

If we assume natural systems to be on the whole relf-regenerating, we have formally three types of capital with fundamental different relationships to time: Natural capital will (assuming harm done to nature by humanity before) increase with time, produced capital depreciates and human capital will change in relation to decisions by society, including those regarding overall fertility. If a society decides to grow while not living in a malthusian world, I would assume human capital to grow, at equilibrium replacement human capital would behave similar to the amount of investment made into it in relation to the generation before, and with a declining population the investment would have to overpower the fact that the number of “goods” into which human capital is accumulated is shrinking overall - literally a basis effect.

Scenario 1: Modeling one-time conversion

Regarding the idea of investment decisions it is already said that some amount of produced capital would be needed to form and maintain human capital, but that relationship seems to be quite opaque. The conversion of natural into produced capital seems rather clear cut that I felt invited to do some simple calculations and plot them to get a feel for the dynamics at play.

Let us assume first that we have a given stock of natural capital with a value of 100, and one generation extracts a certain amount of it and converts it at a certain rate into produced capital. That produced capital then depreciates according to a constant and the natural capital regenerates asymptotically to the original value.

Lets start with a conversion rate of 1.2, a flat extraction amount of 50, a replenishing basis of 25 (adding to the natural capital stock the base at t times the rate divided by the base to the power of 2), a depreciation rate of 0.5%, over a time horizon of 100.

While this is a “happy story” in the sense that overall capital accumulates, it is probably not a likely story due to intertemporal distributional issues: Generation 2 got the biggest jump in well being, and everyone else is facing a depreciating stock of produced capital and has a regenerating natural capital that increases overall capital. Just imagine the difference in values compared to our current ones needed to make this a politically stable arrangement….

Lets be techno-optimists and assume rapidly better technology at the point of extraction with a conversion rate of 2 and a depreciation rate one fifth of the first simulation!

Natural capital replenishes the same as before, but the growth in overall capital is much more. Notice how the better extraction rate makes the intertemporal distributional conflict more pronounced even against the backdrop of intense imbalance of improvements: We increased the extraction rate by 2/3, but divided the depreciation by a factor of 5! (which if one timestep is supposed to be a generation would be crazy value to start with).

Lets be a little more pessimistic about the gnawing of time and half the regeneration basis of natural capital while doubling depreciation of produced capital to 1%, it really starts to bite:

This intertemporal distributional conflict is no longer about who gets to grow how much but who takes the hit for others to grow, pretty ugly. As the code is available to play with, getting a feel for this scenario is available to all.

Scenario 2: Modeling constant extraction rates

Let us assume the generations would like to avoid intertemporal distributional conflicts by extracting constant amounts of natural capital and converting it into produced capital every step of the simulation and see where this leads us. We start again with a stock of 100, a conversion rate of 1.2, a replenishment basis of 25 and a depreciation basis of 0.5% and extract the same amount of natural capital each time:

We see here how the generations avoid conflict with each other by:

1. Exploiting the externalities - the stock of natural capital will be depleted some day, and we have good reason to assume that degredation is non-linear because of biological system effects, and the same is probably true for the impacts on human well being.

2. Accepting massive opportunity costs in the form of binding their own growth trajectory to that of the biosphere. Consider that scenarios of scenario 1 are theoretically all the time possible for each generation as an alternative. Here as well we can say that for this to be sustainable, the generations would need a radically different kind of morality that the one we exhibit at the moment.

If we halve the extraction rate to 0.25%, degredation of natural capital can be avoided completly with a slow build up of produced capital stock, the generations even gain a little overall compared to the basis scenario 2:

This seems to be strictly dominating the first scenario of constant extraction in that it avoids the intertemporal distributional conflict while being sustainable regarding natural capital, although the second point still stands here.

Next, let us assume that the extraction rate is double that of the base scenario to get an idea about the amount of discipline needed to maintain the natural capital - just assume our generations do exhibit the shifting baseline effect. An extraction rate of 1% per step yields:

This.. does look like biosphere collapse while everyone is still feeling fine as progress is still happening - produced capital is growing after all, right?

What if we run the techno-optimist scenario with a conversion rate of 2 and a depreciation rate of 0.1%?

Obviously the biosphere degrades the same as in the base scenario case 2 (each generation extracts the same in both scenarios) but we get more growth - alas, the dimensions are still quite sobering, about 7 percentage points more from the starting point, but with fundamentally better technology, over a time horizon of 100 periods.

Turning to the last variable left unchanged, if we double the regeneration basis of natural capital for both base scenarios, we get for the one time extraction:

There could obviously some kind of optimization be run on the consumption model where it is done several times in order to exploit the faster regeneration at lower levels of natural capital - but the point of this exercise was to get some intuitions about the relationship between these two types of capital.

The same goes for constant extraction, where we could adjust the extraction rate proportionally to the regeneration base of natural capital:

What this simple modeling shows to me is the radicalism of the introduction of natural capital into the summation of social well-being. Accepting the systemic confines of nature as an equal input into our growth trajectories introduces really biting trade-offs in multiple dimensions. I do find the taxonomy of capital goods pretty convincing and am also not a delusional techno-optimist believing in unbound promethean potential of human ingenuitiy. As such, natural capital seems to be the ground base for all other forms of capital either directly of via transmission for produced to human capital. I would also like to stress at this point that we have not even talked about population growth and fertility decisions yet. I suspect once those enter the picture, things will get much more complicated than that.

I mentioned several times how there would have to be radically different morality in order to keep certain extraction strategies stable across time and would like to stress here that my line of thinking there was to assume lived ethical practice as a solution to intertemporal distributional issues “What kind of generations/societies would be needed to keep this behaviour stable?” as a causal story. The rationalistic wording for such a causal story would be “What kind of incentives would each generation need to keep the exhibited behaviour evolutionary stable within their societies?”.

I have on the one hand quite materialistic beliefs regarding the formation of moral judgements of persons and huge respect for biases and desireability of feeling self-affirmation and a positive self-image. At the same time, I do think moral progress is possible because empirically it happened and do believe that one of mankinds most astounding and redeeming feature to be imagination and a strive to do and be better - leading me to believe in the possibility of quite a different morality than we currently observe, with all the great and horrible implications that entails. I am sure we will return to our musings here at a later time.

Placing a value on opportunity sets

This to me is the hardest section to interact with from an emotionally productive stance. I have no horse in this race, and find the argumentation whose model is the better universal model of “Life, the Universe and Everything” overall pretty pointless. What’s more, I feel unable to learn something new from this pedantic point of view in a pretty specialised debate about modeling assumptions and if one model is theoretically subsumable into another. To me, the whole point of models is them being an epistemic tool, and this has no relation to their ontological relationship to each other. If capability theory enabled policy maker on the ground by providing a more intuitive framework and epistemic shortcuts to estimations of total utility even approximately correct, why would any utilitarian object? That is of course my inner pragmatist speaking - if that was not my emotional stance towards these discussions, I would probably be having them more, even enjoying them.

In particular the mentioned proof that foregoing choosing between different capability sets as akin to option buying feels rather mood. Of course that is a way to understanding it but a way of understanding every decision situation in which waiting for further information at a clear or opaque price is a possibility - so the point of this “reduction” is another than to clarify assumptions, it is to implicitly declare superiority of one model over another. If every model is wrong but some are useful, which I believe (of course I have to have a Meta-Theory for that, but that theory is decidedly not utilitarianism as it is not concerned with questions concerning such topics), I would concern this to be an invalid move in the language game, that being models and positions criticising each other in order to improve each other and themselves.

I furthermore do share some intuitions regarding Qualia and think some parts of the decision made by human being to be a plumber or a teacher really hard to communicate satisfyingly as a result of the fact that the feeling of being a body within a certain society fulfilling certain social roles to not be in toto shareable, even if understandable. If we will never be able to understand how it feels to be a bat, the same has to hold true with place and roles in society. Trying to subsume all this into pure numbers as utilitarianism does seems to me to be an epistemic shortcut valid in the face of the difficulty, but no reason for triumphalist declarations, quite the contrary: I would behoove us to remind ourselves of the limits we encounter everyday and practice (epistemic, ontological) humility in the face of them.

Wrapping up

My initial goal was to write about the next 11 pages as well, but this post is already way too long. It feels great to give a text this amount of care and thought. The topics discussed, especially reflecting upon the normative orders and their relation to types of (primitive) produced capital accumulation resonated with me and reminded me of le Guins wonderful translation of the Tao te Ching, poem 29, Not doing:

"
Those who think to win the world
by doing something to it,
I see them come to grief.

For the world is a sacred object.
Nothing is to be done to it.
To do anything to it is to damage it.
To seize it is to lose it.
[..]
"

This feels like a starting point pondering what societies the modeling this book proposes actually wants us to become. I think we may become better people as a result, but that is of course a value judgement on my part.

I was asked to read and discuss this book by Jacob for which I already am immensly grateful.

I have long wanted to have a place in the ether where I could just write as long about media I consumed as I wanted and without time pressure like social media.

Those two facts combined into the substack you are currently visiting.

What I want to communicate here are my thought processes while reading the book. The axe I am grinding is that most book discussions take place at a point in time when the author of the discussion read the whole book and in the best case was able to judge it in its entirety.

I think the thoughts, difficulties, associations, and questions in the moment of first encounter with a sentence are also of considerable value to others. Thus, I will try to document those as best I can. If I have prejudices, priors or similar about the author or content of a book, I plan to lay those out at the start. While in the review process, I will strive to suspend my judgement until the end when I judge the book. Reading a book precludes the rapport with the author that the topic be valuable and their treatment of it productive. As this is the first instance of me doing so, some things will not work well. I hope to iterate and learn and for you to bear with me through this process. I Currently my plan is to read the entry para-text of the text itself containing of

• Eulogy by Dasgupta for Kenneth Arrow

• Foreword by Robert Solow

• Preface

• Random Thoughts on “Birth and Death”  by Kennet Arrow

and discuss those in posts here as a starting point. I will try to create a parent post on here where I link to all other posts to keep the work, I did accessible organized.

The associated code can be found in this repository.

Table of contents:

1. Para-texts

2. Introduction 1

Subscribe now

As discussed in the parent post, I will first put my cards on the table after having read the whole para-text and discuss what I just read after that.

Before that, I want to talk about how I read. One of the most valuable lessons I took from philosophy is that the reader has to do the work where the author did not or could not do it himself. This is because explicating and reading as strongly as possible the position of a deceased authors like Plato or Aristotle in order to place them in the debate may lead to a deeper understanding of ones’ owns, others and the authors positions regarding the matter at hand. Aside from that, one also learns to make the case for or against an argument regardless of ones’ own position on the subject matter. As such, I try to engage a text on its’ own terms, practice suspension of disbelief and judgement until the end and try to lend a helping hand to the arguments where I can. To me this way of reading is extremely helpful and makes every book and adventure - granted, this is a way of reading that makes discrimination of what to read immensely valuable as well.

This book is about a topic that I would approach differently. Which is great, because I do not want to read a book telling me I do everything right, I want to learn.

While I do have a considerable amount of utilitarian proclivities, I prefer to take the cop out of rule utilitarianism leading into the milquetoast neighborhood of deontologist, though they would certainly object. Through Tyler Cowen I am somewhat familiar with the basic premises of Derek Parfit work and can already see the repugnant conclusion lurking in the background.

Regarding my (personal, emotional) stance on the matter of natality, I oppose anti-natalism quite strongly, though I would not consider myself a pro-natalist as well. I would broadly outline my reasoning as follows:

• Economic: As Arrow himself notes, each human has a brain and I consider another brain much more valuable than its resource consumption at the moment. Though I would certainly try to curb consumption of the super rich and also the global north overall - but this is a very different discussion

• My impression is that shrinking populations are a historically unprecedented fact that have several bad aggregated implications:

  • Economic: a society of rentiers seems to become sclerotic in several ways

  • Political: a society of rentiers is a society turning into a vetocracy

  • Cultural: a society of rentiers is a society struggling with change and new ways of being in the world, new questions, answers, values and thus less progress

  • All in all, I think our societies are much less alive than 50 years ago. And I overall consider this to be a bad thing, regarding our collective ability to accept and take on hard problems in the world, but also because I think (a shift on a much more emotional level than intellectual) that being alive is good. I understand that the counterfactual would imply a drastically higher ecological impact as well - and do not have a good answer

• Connecting with the emphasis on emotion, I think human beings in relationships conductive to them wanting to procreate is a wonderful thing and something I wish everyone to have a true choice between.

• Similarly, my understanding of parenthood is that while rearing children tends to be a huge burden in a number of ways, it is my understanding and belief that parents later in life are overall happier people.

• In summary, my position would be for a number of mostly cultural reasons concerning human flourishing, the main question should be how to get more humans into a place in their life where they feel safe and loved in a way to consider procreation, and not engage with the question of how to optimize overall utility of a yet to be existing population within certain ecological limits. (I beg to excuse the slight polemic here) This is not to discount those limits, but simply a result of my ethics, epistemics and overall outlook on life.

I think those are all the chips I needed to lay down before engaging with the material itself, I hope it to be a helpful orientation from where my engagement is coming from and where I personally would want to go. I am sure this book will give me some tough questions to think about and positions and problems to consider, and I look forward to that!

I am in no way opposed to the kind of modeling and prima facie conceptionalisation this book seems to attempt and though I am always very sceptical about models appealing to unpersonified social planners, I see how valuable this exercise can be. From what I read until now, the discussions are quite inspiring, enlightening and the questions being asked are surely fascinating.

My suspected problems, questions at the back of my mind when I start to read the text itself will be about the influence culture is given to, how boundaries within an optimal population to be calculated will be characterized, and what individuals place in all this is. I also feel that the “true” topic seems to be that of limitations, which as I started this post, I would approach much differently. I hope (and have big reasons to hope) to learn so much with this book, about where I am wrong, where I hate being right, where I am undecided and how the world works and where my inconsistencies in thinking lie. I am happy to have this place documenting this journey and hope others may find it helpful or enjoyable as well.

Para-text

After some deliberation, I have arranged the order of my discussion different than presented in the book. I found my discussion of the foreword to be the most cohesive ending of this post, and have thus ordered:

Eulogy - Preface - Arrows’ Mails - Foreword

In memoriam Kenneth Arrow

In this section two things stand out to me:

The love and admiration with which Dasgupta talks about Arrow - this is how I would like the relations in my life to be described as. It made me want to read more Arrow, which I feel is the point. What an amazing person.

The second point is something that I may be getting back to that does confirm my experiences interacting with other people as well:

At the personal level, though, Arrow was far from being a Utilitarian (p.16)

Being friends with a practicing utilitarian is hard. Consequently, being a practicing utilitarian and having (fulfilling and happy) friendships is hard. Consequently, being a practicing utilitarian is hard. One of my reasons why I prefer rule-based utilitarianism is precisely that it does not force you to alienate yourself from your surroundings. My interpretation of this part is that Arrow understood this well and decided to be a utilitarian in theory and happy human being in practice. I sympathise with that, and do not think it is better to be unhappy and right than the other way around - and not only because I have a hard time laying out what “being right” would mean. Of course, being right and happy is the true goal, and I simply think we do not and should not have to pick.

Preface

It is a classic foreword, laying out the work done before on the same topic, gives an overview concerning structure and content of the work itself and states the motivation. I like the starting point beginning with potential parents decisions to be made concerning their desired number of children and the clear categorisation of factors ideally to be thought about. At this point I am a little sceptical about the modeling decision to keep most other factors steady, most notably technological progress. I am sure Dasgupta will have lots to say about this later in the modeling, at least in terms of justification, implication and results concerning the model itself. Notably to me are already the debates about the point where marginal utility is zero - it seems to me that this is what utilitarianism always claims to solve but in my opinion never does: the input-output problem, such that lots of situations that are not clearly bad can be made to be permissible or even desirable conditional on assumed minimal standard. This is not a problem Malthus has but at least right now I understand the book and Sidgewickian Utilitarianism to solve the “problem” by making the break-off point of Malthus flexible, leading to the problem of where to place the slider. The much more interesting problem to me which I hope the book will be talking about at length is not granting the living standard flexibility but the environmental constraint Malthus describes. The malthusian impulse (this is decidedly not meant derogatorily) that I feel to be at the core of the book is laid out in the closing remarks of this section.

I also look forward to the discussion concerning parfitian considerations - while the type of though experiments and arguments I know from Parfit are one of the reasons for me losing interest in academic philosophy, I do find them fascinating, but am pretty sure I will side with Dasgupta here, this is not a hard sell for me. Also notable to me is the interdisciplinary nature of this book, which I hope to enjoy quite a lot.

Arrows’ Remarks

I will comment on Arrows’ remarks in the same order laid out in the third mail.

• I love the comment concerning preferences, which is how I understand his first three points. My takeaway is to separate the problems really sharp: The collective of a society or the social planer can think about the marginal utility of adding another person, but the person being decided about cannot. At the same time the perspective of the parents necessarily has to differ form that of the collective. I think I will return to this point at a later time and look forward to that

• His fourth point is an excellent observation and precise formulation of the dilemma at hand. To me the sheer existence of it stresses the fundamental shift of the human condition resulting from the industrial revolution. I personally would just declare the theoretical considerations laid out here to not apply to the time before, as their significance rest on the industrial revolution having taken place - but I think this is quite an unorthodox position to take

• I had to read ahead to get the general gist of the last point, but decided to return to this discussion once the model is fully laid out - which I assume will take quite a while

Foreword (by the late Robert Solow)

Solow has great humor - similar to the result of the eulogy, I will have to add him and Arrow to my already skyscraper-high list of authors to read.

He starts with Malthus and then continues from Mill to Sidgewick, getting more theoretical step by step. The question in my mind from here on out is how is this different than Malthus, but with a decent standard of living and a criterion not to over extract resources baked into the model? My feeling is this question will stay with me for the whole of the book and a good contrast to compare the progress made to. To me it feels like a wonderful reflection of what we are about to read with personal and technical comments intermingled in an insightful an empathetic way. I think it may be really wonderful to reread this section once I finished the actual text as a starting point to reflect my understanding of it as Solow does here.

I would like to discuss several specific points further.

The first directly connects to the second point about Arrow and combines with Solows’ question regarding the abstract decision maker. That person does of course not exists. Still, the model laid out here (and most other models concerning at least economics) presupposes the possibility of them, while the actual reality we all live in is one of particularities. The model to the persons it tries to model and declares to care about by optimizing their overall utility strips all of them of their perspectives and experiences. I think this is the reason while being a utilitarian and having friends is hard, and the modeler should keep it into account. A person trying to model the world is already one with a complicated relation to herself, the world and the other people in it. I think Solow reminding me of this is really important - and I also think that his answer that we can only hope that the abstracted “us” is making the decisions strikes me at the same time as correct and delusional - after all, I am also a person trying to model the world with a complicated relationship to myself, the actual world and the people in it. But it presupposes a kind of habermasian shared base for decision making that erases the fact that most people to us are utterly alien in their beliefs, actions and feelings. This does not make me fatalistic in any way although it may sound like it. My impression is Solow could write at least as much about this point as I feel I could, but like him I will stop here.

The second point I would like to discuss concerns the section on page 21, regarding the what I would call “derivate” of the optimal population function derived from the attitudes of the actual population regarding its own size. My first reaction was of puzzlement as this does not seem to be a hard problem. Solow (at least as I read him) seems to implicitly agree: The reasons he gives to ignore or “abstract away” this question are three: Preferences of a population could overwrite the calculated optimum we worked so much for to calculate, the modelers do not have the epistemic means to approach these preferences in any way, but(!) then muses about the utility function nonetheless:

maybe most people would […] dislike a very sparse population and they might also dislike a very crowded population, being more or less indifferent about a wide middle range of densities. In that case, it would be safe to ignore the point that I have been discussing. (p.21)

My rought estimates of such a function would be some kind of inverted quadratic function with a wide plateau like this:

import numpy as np
import seaborn as sns

def function(a, b, c, d):
  return -((a+b)**c+d)
  
linspace = np.linspace(0, 1)
result = function(linspace, -0.5,  4, -10)

plot = sns.lineplot(result)

My internal solution to this apparent incoherence is that the truth of the matter is that this is a political question and un-personified utilitarian planers calculating the optimal number of people supposed to live at a certain moment do not have any political mandate whatsoever, which they simply abstract away. I may sound harsh here but I think this is an absolute defensible position to take for a modeler, Solow seems to say so himself, the fact that he mentions it and leaves this obvious incoherence in the text reads to me like another meta-comment on the whole project of the book - this is a great foreword!

The third really important point to me is the short discussion of discount factors - I will skip the short discussions about tinkering with simplistic models derived from Sidgewick as I assume that will be the main point of the actual text - what Solow seems to imply here looks closely related to the point before: Individuals discount functions regarding the utility of their direct and indirect offspring and those of the general population are probably (to the utilitarian) obscenely disparate. Notice how important it was for us modelers to render preferences of individuals moot as discussed previously, as the un-personal planer does per definition not have this discrimination. At this point, I do not even understand what a discount rate is supposed to represent here - typically discounting representing a time preference concerning consumption. So, who consumes here? Do we have to take that ingrained time preference for granted? Would we not say that a typical developmental strategy based on consumption deference for capital investment like China practiced implies a discount factor of more than 1? What are the consumption options or opportunity costs foregone or incurred here? All those questions are great questions that I hope to be able to answer after finishing this book, as those question vexed me for quite some time even without knowledge of this book. For one once we have an actual model written down, we will be able to toy with discount factors and the like to test our intuitions, get a feel for the topology of the model and understand the key take insights gained from explicating our implicit assumptions in a (hopefully) coherent way.

Finally (at least in my internal structuring of this text) Solow tries to demark clearly the difference this book and the real world:

This is not intended as a description of the way fertility and saving decisions are made by ordinary humans: it is intended as a benchmark against which to evaluated those decisions, and perhaps to make recommendations

All in all, the foreword by Solow is as I already expressed, great. It tells the reader clearly what to expect and how to engage with the text at hand. It is a great meditation and cautionary reflection on the natures of models, the character of modelers and the practice of modeling. Of the hidden assumptions not only in models themselves but in the people writing them. It is a great text, and it praises the actual book for trying to get a handle on hard questions.

Reflections

Two things keep resonating with me having spend quite some time with this relatively short text(s). The first is obviously about modeling - it is a topic near and dear to my heart. We all know models are wrong but can be useful, a factor I think many are in the danger of forgetting. To me, the go to metaphor for all of this is the MONIAC, which I encountered for the first time in Terry Pratchetts’ "Making Money”. Not to be reminded how far modeling has come (it certainly has, not least thanks to computing), but as a reminder and an object of reflection and meditation that can be transferred to other models - its weirdness makes it at least to me really useful for that. I think about the intuition and metaphors that led a modeler to the conclusion that water may be a way to represent the “flows” of capital within a necessarily “closed” economy (one can already see how a first intuition forces one to buy other premises as well), about the inherent limits the model clearly supposed and processes and phenomena it simply could not account for, represent or even point in the direction of. At the same time I consider that most of that was probably understood by the modeler themselves - they knew what their model could and could not do, and used it because they thought it to be useful to them at the time. And it probably was. I personally also love all other implementations of a computational device differing from the classical von-Neumann architecture, but that seems a rather exotic interest.

The second point I want to make it the demarkations of the project of this book already in play in the foreword, and it is more of an emotional point. The extremes between which this book will take place are obscene amounts of wealth and consumption on the one hand and starvation on the other. I do not want to talk about the first as I feel this is not the place, but the second. My understanding is that the implicit fear, even terror of this book is the return of Malthus by slower technological progress, a shrinking hospitality of our ecology and population growth not taking these factors into account. I think this is probably one of the most important questions that we in affluent societies have forgotten but should have not. I may be projecting a lot here, but I wish to imprint upon me and you the reader the terror of Malthusian Dynamics that are so diminuitively described here, by two man who did it well. I will start with the less graphical and more artistic quote by Terry Pratchett in “Snuff” on page 11 where a characters reads out the description of another race of sentient and very discriminated against beings living under those circumstances:

“I must say that goblins live on the edge, often because they have been driven there. When nothing else can survive, they do. Their universal greeting is, apparently, “Hang”, which means “Survive”. I know dreadful crimes have been laid at their door, but the world itself has never been kind to them. Let it be said here that those who live their lives where life hangs by less than a thread understand the dreadful algebra of necessity, which has no mercy, and when necessity presses in extremis, that is the time when woman need to make the unggue pot called ‘soul of tears’, the most beautiful of all the pots, carved with little flowers and washed with tears.”

I think the fear of encountering the “dreadful algebra of necessity” is well found. The book talks about fertility decisions on the basis of modern contraceptives. These methods are not that old, and living not on but behind the point where “marginal utility” of another human being in a society does imply such situations.

But take it not from me or Terry, take if from Brad DeLongs “Slouching towards utopia” describing from what place, the malthusian world, that slouch did start on:

Agrarian Age humans were desperately poor: it was a subsistence-level society. On average, 2.03 children per mother survived to reproduce. A typical woman (who was not among the one in seven who died in childbirth, or the additional one in five who died before her children were grown, sometimes from the same contagious diseases to which her children succumbed) would have spent perhaps twenty years eating for two: she would have had perhaps nine pregnancies, six live births, and three or four children surviving to age five, and the life expectancy of her children remained under, and perhaps well under, thirty. Keeping your children from dying is the first and highest goal of every parent. Humanity in the Agrarian Age could not do so at all reliably. That is an index of how much pressure from material want humanity found itself under. (p.16)

Reflect on the amount of suffering just this statistics entail. How much violence, how much hunger, how much tragedy, how little (if any) freedom life entailed not that much long ago for our species. Brads’ book makes at least to me clear what an incredible miracle the current state of our species as a whole is in - imperfect and unjust as it may be. To me it also left the important impression that this state is not sustained on its own, and that we have not banished the third rider of the apocalypse as well as we would all like to believe.

I think the topic of this book is of immense importance. And I do hope it will have answers or points from where to think further about those hard problems.