Marcos

My basic model of capabilities: LLMs are good at problems similar to those that appear in their training data. Training data largely reflects the world, and so LLMs are relatively good at problems that are common, relatively bad at problems that are rare.

One of the great promises of LLMs is that they have reduced linguistic barriers for non-native English speakers in academia by at least an order of magnitude. Economics, my own field, is 99% in English. I have written a few papers and a book in Spanish but, for all practical purposes, these count for zero in my CV. I had to spend considerable time and effort to reach a level where I could give a 90-minute seminar in English. I could do it because economics is a math-heavy field (you can always put the equation on the slide and let the math do the work) and because Spanish is not that far from English. But I have seen incredibly brilliant students, particularly from East Asia, crash and burn in the job market because their English was below a minimum standard. I do not complain about it. As my great professor at Minnesota, Chris Phelan, loves to say: “Nobody should have told you that life is fair.” Some are born into upper-middle-class families in Hampstead, London, and some into villages in rural China. Given how different Chinese is from English, what amazes me is how smart my Chinese colleagues must be to give a 90-minute seminar in English at all. If you are a native English speaker, try to learn enough Chinese to give a five-minute talk. Then you will understand what they have accomplished. LLMs mean that an Egyptian student no longer must agonize for hours over a flowing introduction to her job market paper. Or a Korean student can prepare for interview questions without a native speaker holding her hand. We are only beginning to see the effects of this change. Native speakers (or speakers of languages close to English) have, so far, maintained a clear advantage in policy discussions. Does anyone truly believe that Paul Krugman would have become a New York Times columnist if he had been a native speaker of Japanese? Same brain, same graduate education, same contributions to economics, same academic jobs, very different impact on public opinion. Or why is economics or physics much more international than sociology, political science, or history? Because in economics or physics, you can get away with much lower linguistic skills. In history, if you cannot write English prose that meets a particular standard, you cannot publish. The resistance to LLMs in academic writing is, in many cases, the defense of the linguistic advantages of English. Let’s call a spade a spade.

En España tenemos cosas que damos por descontadas y americanos y otros países alucinan • Lubina de estero a menos de 15€/kg • Entrenar al sol en parques de calistenia gratis • Mercado con producto fresco a 10 minutos de casa • Caminar a casi todos los sitios sin pensarlo • Sobremesa de 2 horas • Aceite de oliva virgen extra en todo • Vida social en la calle, no en casa • Siesta cuando el cuerpo la pide • Marisco fresco que en Londres es un lujo • Clima que te permite estar fuera 9 meses al año • Boquerones, sardinas y caballa. Los omega-3 más baratos del mundo • Sanidad pública que te hace analítica sin arruinarte • Bocadillo de jamón ibérico por 3 euros • Luz solar en invierno que el nórdico no tiene • Red social densa: familia, vecinos, bares. El americano lo llama "comunidad" y lo estudia en Harvard Los estudios de zonas azules llevan décadas buscando la fórmula de la longevidad mediterránea. Aquí es simplemente el martes.

The metareview:

Hoy conseguimos que el Parlamento Europeo paralice Mercosur y lo envíe al Tribunal de Justicia de la Unión Europea Mercosur significa reventar el campo para que ganen las grandes empresas. Estamos con los agricultores y ganaderos que cuidan la tierra y ponen comida en la mesa

Es heavy como la IA generativa se ha apoderado del término IA en general. Es como Bimbo con los panes