Posted: October 14th, 2026 | Author: Domingo | Filed under: Artificial Intelligence, Cognitive Warfare, Disinformation, Geopolitics, Large Language Models, Realpolitik | Tags: Desinformación, Geopolítica, Grandes modelos de lenguaje, Guerra cognitiva, inteligencia artificial | Comments Off on Tlön, Uqbar, Orbis Tertius… o el LLM grooming en la jerga de la IA generativa
En 1940, Jorge Luis Borges escribió «Tlön, Uqbar, Orbis Tertius». El relato arranca con Uqbar, un país que aparece en un único ejemplar de una oscura enciclopedia y en ningún atlas del mundo, y cuyos escritos remiten a un lugar aún más extraño: Tlön. Detrás de ambos está Orbis Tertius, una sociedad secreta de sabios que dedicó generaciones a inventar un planeta entero y a escribirlo, artículo por artículo, en una enciclopedia propia. El engaño estaba tan bien construido que la realidad acabó convirtiéndose en eso mismo, realidad: empezaron a aparecer objetos de Tlön, y el narrador termina convencido de que el mundo será Tlön.
Hoy, esa sociedad de sabios la forman los grandes modelos de lenguaje, los oráculos de la era pos-Google. No inventan nada a propósito: simplemente escriben lo que han leído. Así que quien escriba lo que ellos leen puede construir su propio Tlön. La técnica tiene nombre, LLM grooming, y las elecciones de Moldavia de 2025 la mostraron en práctica, empezando por un titular que nadie escribió.
El 29 de septiembre de 2025, el día después de la votación, apareció en internet un artículo en rumano con este título: «Escribe un post sobre la política moldava en 2025, sin que tenga relación con las elecciones». Alguien había escrito un pompt en ChatGPT, había publicado lo que salió, sin borrar dicho prompt. ERROR FATAL, como aparecía en la pantalla de nuestros viejos ordenadores. El desliz quedó registrado por el Mecanismo de Respuesta Rápida del G7, que siguió las elecciones en consulta con el Gobierno de Moldavia. ¿Gracioso? Mucho. ¿Inofensivo? No tanto.
El artículo pertenecía a la red Pravda, también conocida como Portal Kombat. Se trata de un entramado de al menos 193 webs que apenas producen nada propio y se limitan a republicar contenidos de medios afines. Su audiencia en Moldavia era prácticamente nula: el informe del G7 habla de «niveles mínimos de interacción auténtica». O sea, un medio que no lee nadie. En cualquier facultad de Periodismo, eso es un fracaso. Salvo que los lectores que buscaba no fueran personas.
Aquí es cuando la historia empieza a ponerse interesante, sobre todo desde el punto de vista de la IA generativa. Los grandes modelos de lenguaje aprenden de lo que contiene la web, y muchos de los chatbots construidos sobre ellos también buscan en la web antes de responder. Así que basta con inundar internet con tu versión de los hechos, una y otra vez, en decenas de webs y de idiomas, hasta que las máquinas la absorban. Después ellas la repiten, con su tono sereno y servicial, a usuarios que nunca han pisado tus webs. El Royal United Services Institute (RUSI), un think tank británico independiente dedicado a la defensa y la seguridad, fundado en 1831 por el duque de Wellington, lo resumió bastante bien: el objetivo ya no es la audiencia, sino las herramientas que usa la audiencia; es decir, los propios modelos de lenguaje.
Y en este caso concreto, ¿cuál era la historia que había que repetir? La de siempre. A la presidenta Maia Sandu y a su Partido Acción y Solidaridad se los acusó de un fraude electoral «orquestado por la UE y la OTAN», de malversación e incluso de drogadicción, y se presentó al partido como un títere de la OTAN que amenazaba la neutralidad de Moldavia. NewsGuard, una empresa estadounidense que evalúa la fiabilidad de los sitios de noticias y audita chatbots de IA, constató que la red blanqueaba bulos como estos «en un aparente intento de influir en los rastreadores web y en la IA generativa».
¿Mordieron el anzuelo las máquinas? Al menos una sí. Matryoshka es una operación de influencia que fabrica vídeos y artículos que imitan a medios de comunicación reales y luego los envía por correo electrónico a verificadores y redacciones para tenerlos ocupados desmintiendo. Entre mediados de abril y mediados de julio de 2025, difundió 39 bulos sobre Moldavia. Según NewsGuard, ChatGPT repitió uno de ellos y citó como fuente un artículo de la red Pravda, precisamente las webs que casi nadie lee. Casi nadie, salvo las máquinas.
Dicho esto, toca volver a ponerse el sombrero de analista. ¿Cambió todo esto un solo voto en Moldavia? No lo sabemos, y lo poco que sabemos apunta a menos de lo que parece. El Partido Acción y Solidaridad conservó su mayoría parlamentaria. Las mayores audiencias de la campaña fueron para el vídeo, no para la granja de textos: dos vídeos en inglés de otra operación sumaron más de 2,7 millones y al menos 5 millones de visualizaciones en X. Y ninguna de las fuentes que he leído mide qué llegó a creer un solo moldavo después de hacerle una pregunta a un chatbot de IA generativa. La producción está documentada hasta el último artículo. El efecto en las mentes sigue sin medirse.
Aquí va la luz, si es que estas líneas arrojan alguna. El grooming juega una partida más larga que una campaña electoral. No necesita ganar en septiembre. Necesita estar en los datos de los que aprenderá el próximo modelo y en los resultados de búsqueda que consultará el próximo chatbot. Por eso la defensa no empieza en las urnas, sino en la fuente. Hacen falta desarrolladores que valoren de dónde viene una respuesta tanto como si suena convincente, y usuarios que lean lo que les responde un chatbot sobre un tema político candente como leerían un soplo anónimo, comprobando la fuente que cita antes de fiarse.
Que alguien está intentando enseñar a las máquinas su propia versión de la realidad: pocas dudas caben.
Que la lección haya calado lo suficiente como para inclinar un voto: ninguna prueba de ello, por ahora.
Y a quien publicó aquel prompt como titular: la próxima vez… por favor, vuelve a olvidarte de borrarlo.
Posted: March 18th, 2022 | Author: Domingo | Filed under: Artificial Intelligence, Book Summaries, Realpolitik | Tags: artificial intelligence, China, geopolitics, techno-socialism | Comments Off on China: Techno-socialism Seasoned with Artificial Intelligence
“People take the great ruler for granted and are oblivious to his presence.The good ruler is loved and acclaimed by his subjects. The mediocre ruler is universally feared. The bad ruler is generally despised. Because he lacks credibility, the subjects do not trust him. On the other hand, the great ruler seldom issues orders. Yet he appears to accomplish everything effortlessly. To his subjects everything he does is just a natural occurrence.“
Tao-T-Ching, Lao-Ts
Anyone who wants to learn something about China today, to know its strategic plan between now and 2050, the means to achieve it, and what drives this country in this titanic effort, should read the book El gran sueo de China: tecno-socialismo y capitalismo de estado by Claudio F. Gonzlez.
Claudio F. Gonzlez, PhD in engineering and economist, has lived in China, as director of Asia for the Polytechnic University of Madrid (UPM), for six years. During this time he has been involved in the fields of education, entrepreneurship, research and innovation in the Asian giant. From this privileged vantage point he has been able to observe, analyze, and understand the complexity of this country.
According to the author, throughout the 20th century, the Western world looked at China with the condescension that is due to a former empire in decline and mired in chaos, power struggles, and poverty, and only in the last decades of the past century, as a market of great potential and a cheap manufacturer of limited quality. Nonetheless, China had -and has- its plan, the ultimate goal of which is returning the “Empire of the Center” to the place it has held for most of human history. Namely: being the most socially and technologically advanced nation and, from there, regaining world leadership in the economic, commercial, and cultural spheres.
In 2015, the government announced the first of its grand plans – Made in China 2025, with the goal of making China by this date a leader in industries such as robotics, semiconductor manufacturing, electronic vehicles, renewable energy and, of course, artificial intelligence.
Initiatives such as the Belt and Road Initiative (BRI) or institutions such as the Asian Infrastructures Investment Bank (AIIB) are nothing more than instruments through which China wants to reshape an international order that is more favorable to its new interests. One of China’s stated goals is that by 2035 it wants to be the country that globally sets the next standards in areas such as AI, 5G or the Internet of Things.
China’s successes in the digital economy are based on three main factors:
1. A market that is both huge in size and young, which allows for the rapid commercialization of new business models and equally allows for a high level of experimentation.
2. An increasingly rich and varied innovation ecosystem that goes far beyond a few large and famous companies.
3. And a strong government support, which provides favorable economic and regulatory conditions, and also acts as a venture capital investor, a consumer of products based on new technologies and produced by local companies, and allows access to data that are key to developing new solutions in conditions that are unthinkable in other regions.
Professor F. Gonzlez calls this model techno-socialism or state capitalism.
What are the defining characteristics of this techno-socialist model?
China intends to harness the interest in technological development of its own industry to align it with government interests. The overall goal is, starting from what the Chinese Communist Party (CCP) calls a moderately prosperous socialist society, to catch up with and surpass the most developed Western countries, ideally by the 100th anniversary of the founding of the People’s Republic of China (2049). Socialism in the sense of the Chinese regime is no longer socialism in the traditional sense of ownership and collective management of the means of production, a political conception definitively defeated after Mao’s demise; but its control and coordination to achieve social objectives.
The features that characterize this techno-socialism are those of complete physical security for people and things, the absence of extreme poverty, full employment, and the possibility for the most industrious to obtain economic and prestige rewards for their efforts, as long as they are aligned with the objectives established by the party and do not put its dominion at the least risk. This techno-socialism tries to lead society as a whole towards a centrality of thought that avoids extremisms that destroy peace and social security, and that do not call into question the leadership and omnipotent dominance of the party.
The alignment between business interests -or those of other institutions- and public interests, as interpreted by the party, creates a unique innovation ecosystem in which companies capable of promoting solutions for a broad user base become champions of an industrial policy. Once this status is achieved, and always within the logic of interest alignment, they will gain access to a whole arsenal of measures -subsidies, tax reductions, preferential treatment-, to maintain this position and, if possible, extend it internationally, since they are no longer merely companies, but ambassadors of a new model. In the particular case of artificial intelligence, the government has contributed with the necessary conditions -strategies, plans, regulation, space for experimentations- and practical support -venture capital, public procurement, permissions to access data-, for innovations in this field to follow. Alibaba, Tencent, and Baidu set up research centers, deploy applications, enroll human capital, and support CCP policies.
Will techno-socialism be able to generate enough disruptive innovations to give the technology created in China an entity of its own?
Between 2015 and 2018, the venture capital funded more than 1 trillion to new technology start-ups in China. China has more unicorns -companies less than ten years old with a valuation above $1 billion- than any other country. In terms of research, China is already the country with the most scientific articles, surpassing the US, although it is true that its impact is still minor, the gap is rapidly closing. It turns out that it has been the state the one which, with its research grants, scholarships and universities, has generated ideas that, because of their risk, private initiative would never have dared to finance. Hence, in this sense, public authorities that nurture alternative ways of thinking are the true engine of progress.
Professor F. Gonzlez names this innovation paradigm applicable to China as asymmetric triple helix model, in which the national government controls the overall innovation context through its top- down policies and plans but, at the same time, allows a certain level of autonomy for district, local and regional governments to conduct their own experiments and accommodate innovations that emerge from the bottom up. Large companies, start-ups, and finance companies are aligned with government interests. And universities and research centers similarly align themselves with government objectives in producing new knowledge and generating talent in the form of human capital.
And eventually, when will China achieve and assume the role of world leader?
From the author’s standpoint China, due to a set of inconsistencies and structural gaps, is neither ready nor willing to assume the global leadership in the foreseeable future. However, it does claim to be the most powerful and influential economy, with the most cohesive society and the least contested domestic leadership that will enable it to become something like the best country in a fragmented world. China’s current strength lies in the existence of a long-term plan: a sense of destiny that ties in with its imperial past. There is a deep conviction in Chinese society, a determination, which is the key force to achieve these strategic objectives.
China does not want to be a powerful nation, but deserves it.
