Alan Fortuny Sicart

Solow stated in 2008 that there is no reason for capitalism to exist in a steady state, and this is the main conclusion after we analyse the neoclassical literature. There are multiple conditions where economies could function without growth, and such economies almost always have better environmental state than those that grow. Those theories also called supply side economics, state unanomously that the condition for stable output and emissions is a constant supply of effective (actual input * productivity) input remains constant.  That could be done by either reducing the input in line to the technological progress on that input or redirecting input usage and technology to the clean ones (labor instead of resources, and renewable sectors instead of fossil fuel based).  To achieve that, regulation is needed. A cap on input use have to be in place for example for resource use or taxes on the dirty sectors will deviate production to cleaner sectors that are more intensive in labor and ca

  Dasgupta-Heal-Solow-Stigliz Model (1974) The following model adds limited natural resources to a model similar to the one developed by Sollow. The stock of the non-renewable resource is depleted at the following rate and overall lower than the initial stock: These model assume constant susbtitution elasticities between the factors K,L, R. As Resource availability is strictly decaying, economic growth require significant amounts of factor substitution . So significant as natural resources need to become inesential in the production process.  Another alternative of the model is resource augmenting technology, where that leads to an increase in effective natural resource avaialability bigger than the decline of the resource stock . The third option to explain long term economic growth is by permanently increasing the other factors, capital and labor. That would be easier the higher the productivity of those factors is. Green Solow Model with abatement The green Solow model keeps the ma

  AK Model of Humand Capital In this model, there are no decreasing marginal returns on capital as in the models presented in part 2.1.If capital K is consider to reflect the accumulated knowledge and physical capital, this can be consider to provide constant and not reducing returns to scale. That lead to the following equations for the determinats of the growth of output: That ultimately means that a s long as the technological progress growth is higher than the disccounted dereciation of capital, output will grow at a pace equally to the growth of that general concept of capital (human and physical). Zero growth conditions In this model, increases in the productivity t have to be compensated by reductions in the accumulation of capital at the same rate . That would happen if depreciation or saving preferences change in the direction to lower savings and higher depreciation. While here the determinants of growth are engogenous determined, the casual separation between technology de

 Foundations In this theories, two agent types determines the level of production, consumption, investment, labor demand...which are homogeneous firms and homogeneous workers. Firms maximize profits producing more till the point where marginal revenue equals marginal costs. Households are workers decide how much to work given the interest rates and wages, together with their preferences. Their income comes from labor, savings and firm's profits (they are both workers and owners). Due to the decreasing marginal productivity of the input factors and also consumption in the utility function, more work and consumption ultimately contributes less and less to the utility. Labor supply is positively related to the level of wages (the opposite happens with labor demand). Consumption decisions over time depend on the interest rates, as higher interest rates makes more appealing saving. In this context there is no unemployment, as all labor is used and adjusted to the productivity levels on

 Motivation Economic Growth is a relative recent thing, as it barely exists before the XIX century.  The division of labor and use of dense/cheap but polluting forms of energy explain our increases in productivity, but also a destruction of the environment and the exhaustion of the commons that is imprecendented in human history at least. Pursuing a steady state economy, and the study of the the macroeconomics of the steady state is not about going back in time, but rather comply with resources and environmental limits, achieve sufficiency, in a economic system that avoids the growth mandate while improve wellbeing intra/inter generations. In my first paper [1], I showed the limitations of growth in multiple dimensions and the impossibility to achieve aggregate growth without facing social, resource, energy and environmental crises. While the case that we need to stop growing has been made, what remains to  be investigated is the macroeconomic conditions of economies without growth

 Momentum Contrast for Unsupervised Visual Representation Learning(MOCOv2) In the previous post, we present SIMCLR as one of the most promising self-supervised image representation methods. As the caviat of long training and large batch sizes could be too much for many applications, Mocov2 and its subsequent versions are both exploration. https://arxiv.org/pdf/1911.05722.pdf Introduction Self supervised methods in NLP are the norm, and almost how they have been built. Language tasks haven been made more and more efficient with the usage of dictionaries and keys on attention models, where memory is manage very efficiently on very large texts.  The obvious continouous, highly dimensional nature of image data should not be a block for leveraging the learnings from NLP in self supervised tasks and representations in computer vision. Moco is a great proof of how it can be done. If we reframe keys and tokens in the forms of group of images that are similar in the same query and disimilar to

 SIMCLR: simple framework for contrastive learning representations As from the previous post, this is aimed to summarize the following paper, in the next post we cover moco and its benefits from a computational standpoint. Let's get started: https://arxiv.org/pdf/2002.05709.pdf Introduction Viual representations can be done with generative (VAE or GANS) or discriminative methods (like SIMCLR or MOCO) when labels are not available. Generative methods are expensive to compute, and have some issues as shared in the part 1 of this post. In this paper we see how a discriminative approach, which replicates a supervised tasks on self generated labels is performing state of the art results. Not only that, but SIMCLR is also more intuitive, requiring a much simple architecture. The major components of that framework are: contrastive tasks based on data augmentations the contrastive loss contains a non-linear component to improve the quality of the representation large batch sizes and long t

 It is fair to say that there are two main challenges in applying computer vision in industry and in research groups. The first one is the complexity of the topic and the methods used so far. The second is the immense amount of labeled data to get sufficient good results , particularly when transfer learning is not an option. While the first problem have multiple solutions, the second have for many a much less clear roadmap to go. How can we properly extract features and maximize information retrieval from an image, if we do not have labels? How can we ensure we are building a generalized and human aligned encoding of an image if we have no labels to "bias" or direct the learning of our neural net?  In the following posts I focus on how can we create high quality image representations that are useful, unbiased, generalizable and do not required an insane amount of training or gpu compute. Let's get started. Key definitions and concepts before we get started It is importa

Recognition Problems There are plenty of applications where we need to recognize a face, specie or car plate in a single image. This is challenging, as we need to get it right, while seeing only once that object specifically. Training that objects with one sample for each would not work neither. What we can do instead is to try to stablish the similarity between two images to stablish the likelihood of the same object being in both. In that way we can stablish if the object is in the training data based or not, given the similarity score and a threshold. Siamese Networks anf the Triple loss In order to calculate that similarity function, siamese networks tend to work fairly well . Those networks encode the image into a lower dimensional vector, from which we can compute a distance metric (cosine, euclidean...). The learning of that networks requires a specific loss function, called triplet loss. The triplet loss we have an anchor (the image to evaluate) and a positive and negativ

While neoclassical economics explain growth based on the supply of the factors of production and keynesian economics as the result of effective demand , marxist theory understand the economic sytem as both political and physical. Instead of giving exact conditions, the theory provide a dialectic description of a new system that does not need to expand. On top of the supply of factors and the effective demand, the competition and ownership setting is key to explain the development of the economic process.  Marxist theory provide   a valuable framework to explain the growth mandate and the last economic crises with competitive capitalism . Production expansion is required to increase profits with economies of scale and keep employment . Crisis are regularly happening as the capital intensive expansion of production leads to excessive supply and weak demand . On the negative side those theories consider profits as the only source of investment, which have little support in economies whe