Thoughts and Theory

Public dataset and analysis of the evolution of parameter counts in Machine Learning

In short: we have compiled information about the date of development and trainable parameter counts of n=139 machine learning systems between 1952 and 2021. This is, as far as we know, the biggest public dataset of its kind. You can access our dataset here, and the code to produce an interactive visualization is available here.

Model size of popular new Machine Learning systems between 1954 and 2021. Includes n=139 datapoints. See expanded and interactive version of this graph here.

Getting Started

I spent the last six months researching causal discovery, in the hopes of finding actionable insights that could be transported to other fields (specially econometrics and medicine) to improve our approach to observational studies.

My reasoning was something as follows:

  • Causal inference from observational data is a very important topic for many different fields ; better causal inference concepts and tools can help us disentangle topics such as the long term factors of economic growth or what biological indicators are accurate trackers of biological age
  • There has been much original research done in the last 20 years on representing causality…

Más de una vez has debido encontrarte una imagen parecida a ésta. Decenas de puntos, conectados unos a otros en un intrincado patrón. Pero, ¿qué pueden ser? ¿Y por qué son tan importantes?

Esta imagen es un grafo. Y, como explicaremos más adelante, es la manera en que un matemático puede representar y estudiar conexiones. Y es que el estudio de las conexiones ocupa un lugar privilegiado en las matemáticas.

En este artículo explicaremos cómo se estudian formalmente dichas conexiones — mediante grafos -, y por qué los matemáticos están obsesionados con ellos. …

[Crossposting from]

IN SHORT: We attempt to forecast when quantum computers will be able to crack the common cryptographic scheme RSA2048, and develop a model that predicts less than 5% confidence that this capability will be reached before 2039. You can read the full article at

Advanced quantum computing comes with some new applications as well as a few risks, most notably threatening the foundations of modern online security.

In light of the recent experimental crossing of the “quantum supremacy” milestone, it is of great interest to estimate when devices capable of attacking typical encrypted communication will be…

Professor Ehud Reiter (my PhD supervisor) just wrote a piece on OpenAI’s GPT-3, arguing that its applications will be quite limited.

While I do broadly agree with the claims being made, I noticed several points of disagreement, so I resolved to take this chance to explore why our models of the world differ.

Let me first state our common ground:

  • Without humans aiding it, GPT-3 produces inconsistent results
  • Its context window (its memory span) is quite limited, so it cannot produce long coherent texts on its own
  • GPT-3 will most likely not be able to accurately summarize and explain data…

In this article I summarize the background research I did on Climate Change for an upcoming report for the Spanish government.

I will first look at the expected consequences of climate change. I will then dive into some considered policy solutions. I will finish the article with some suggestions for action for organizations and individuals.

Photo by Callum Shaw on Unsplash

What are the main consequences of climate change?

In 2016, human activity over the past century is estimated to have caused warming of between 0.8ºC and 1.2ºC above pre industrial levels (1). This rate is likely to increase as greenhouse gas (GHG) emissions increase.There is also the possibility of significant changes in the…

Toby Ord no podría haber elegido peor momento para publicar su libro.

Ord trabaja en el Future of Humanity Institute de la Universidad de Oxford. Uno puede imaginarle pensativo en su despacho, contemplando el Precipicio del que nos advierte y hacia el que la humanidad se dirige.

Y a día de hoy, en mitad de una crisis sanitaria global, parece poco prioritario preocuparse por lo que viene después. Pero es una tarea necesaria y fundamental para proteger a la humanidad de los desastres, si cabe aún peores, que podrían estar a la vuelta de la esquina.

Con este artículo no…

ABSTRACT: Not all covariates of treatment and outcome variables in an observational study should be adjusted for. By default, one should doubt studies which blindly adjust for many confounders without justifying their choice on causal grounds.

DISCLAIMER: My knowledge of causal inference is limited enough that I could be saying things that are very wrong. Reach out to me on twitter @jsevillamol if you find a mistake!

The problem of confounders

Suppose that you want to determine the causal effect of a treatment on an outcome. First order of business is determining whether there is a statistical correlation between them.

Albeit still challenging, we…

Jaime Sevilla

ESR NL4XAI. Math and computer science expert.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store