Question of the year — 2025

ENGLISH

Artificial Intelligence in Journalism: Reshaping Newsrooms in Argentina

The rapid evolution of artificial intelligence is transforming industries worldwide, and journalism is no exception. In Argentina, organizations like Telecom Argentina, Chequeado, and the newspapers Clarín and La Nación are leveraging AI to enhance journalistic practices, improve information accuracy, and deliver content tailored to diverse audiences. These advancements not only streamline workflows but also reshape the media landscape, fostering a more informed and connected society.

AI tools like ChatGPT are becoming increasingly prevalent in Argentinian newsrooms.

A report by the Universidad Nacional de Buenos Aires¹ revealed that 82.2% of journalists surveyed have used generative AI tools in their workflows. The study highlights that professionals rely on AI for tasks such as brainstorming, drafting headlines, writing summaries, SEO optimization, and even producing full articles. In Argentina, where newsroom budgets are often tight and resources limited, these tools offer an efficient way to optimize workflows and meet deadlines. The survey also identified a pressing need for ethical guidelines and continuous training, with 71% of respondents emphasizing their importance. This shows how journalism in Argentina is balancing the potential of AI to enhance productivity while identifying the need to preserve editorial integrity in a fast-evolving media landscape.
Telecom Argentina’s publication “Periodismo IA“² serves as a comprehensive guide for journalists navigating the integration of AI into their workflows. The guide addresses themes like automation, data-driven reporting, and personalized content delivery.

Automation is particularly transformative, with AI tools handling repetitive tasks such as transcription, fact-checking, and data sorting. For instance, real-time transcription tools accelerate content production, allowing journalists to focus on deeper analysis. When consulted about the impact of AI in journalism, Álvaro Liuzzi, author of Periodismo IA, highlighted the importance of ethical AI use, stating that “The integration of artificial intelligence in newsrooms is not only about efficiency or technological innovation; it also raises profound ethical questions that impact how information is reported and how the world is understood.”³ In this regard, Periodismo IA devotes a chapter to the ethics and transparency of artificial intelligence, focusing on the frameworks and guidelines for its implementation in news production and editorial workflows. The guide is freely available for download and is accessible in Spanish and English.

Beyond streamlining workflows, AI is also playing a pivotal role in analysing cultural and social phenomena in Argentina, offering new perspectives that extend beyond efficiency improvements. The leading newspaper “La Nación” has implemented AI tools to uncover deeper insights into topics such as music trends and election monitoring. For example, AI algorithms have been used to analyse the lyrics of “traperos” ⁴ (artists in Argentina’s trap music scene), revealing recurring themes like socioeconomic inequality and urban life. This innovative approach gives readers a data-driven lens to better understand cultural dynamics. During elections, La Nación⁵ utilised AI to examine telegrams and electoral documents, identifying discrepancies that might have otherwise gone unnoticed. These applications illustrate how AI is being leveraged to enhance journalistic analysis, providing deeper insights and fostering a more comprehensive understanding of societal trends.

Clarín, another prominent newspaper in the country, has introduced “UalterAI”,⁶ an innovative AI-powered reading assistant.
UalterAI allows readers to customize their news consumption experience, offering options like concise summaries, chronological story formats, and categorized lists of key numbers or statements. For example, the tool can present the top 10 points of a story or display data in a more digestible, columned format. This level of customization improves accessibility, and at the same time empowers readers to interact with news in ways that suit their preferences and needs. By adopting user-centered innovations like UalterAI, Clarín demonstrates how AI can make journalism more inclusive and adaptable.

Meanwhile, “Chequeado“,⁷ Argentina’s foremost fact-checking organization, exemplifies the potential of AI in combating misinformation. The organization blends traditional journalistic rigor with cutting-edge AI tools in their operations, enabling the organization to process vast amounts of online content, detect patterns of misinformation, and flag dubious claims for review. Chequeado also provides a range of AI driven tools for journalists, like Chequeabot,⁸ which identifies fact-checkable phrases in texts and monitors trending content across multiple social networks, among other features. Other tools include a transcription app, an inflation calculator, and cost-assessment functionalities. Beyond its internal processes, Chequeado collaborates with other media outlets, offering training and tools to integrate fact-checking into their workflows. This collective effort fosters trust and transparency in media, essential for combating the spread of misinformation.

AI enhances efficiency and allows journalists to focus on meaningful storytelling.
By fostering transparency and enhancing content accessibility, AI tools are reshaping journalism to better serve the public good. The integration of AI into Argentinian journalism marks a pivotal moment in the industry’s evolution. The organizations mentioned above are leading the way, demonstrating how AI can enhance journalistic practices while maintaining integrity. At the same time, ethical considerations remain paramount, with ongoing efforts to develop guidelines that address concerns like misinformation, bias, and the reinforcement of narrow perspectives. As Álvaro Liuzzi aptly puts it, “Reflecting on how to balance automation with preserving humanity in journalism is crucial for the future of the profession.” A responsible and thoughtful approach to AI implementation will ensure that technology continues to support journalistic excellence, reinforcing the profession’s commitment to truth, accountability, and impactful storytelling.

References
1. Usos del CHATGPT en la Prensa Escrita de la Argentina | POLÍTICAS DE LA CONVERGENCIA.
2. La revolución de la IA en el periodismo | telecom.
3. Álvaro Liuzzi: “La IA está revolucionando el periodismo” | PARENTÉSIS MEDia.
4. Argentina: Inteligencia artificial en el periodismo de La Nación and Así suenan y de esto hablan los traperos en la playlist de los centennials | Global Investigative Journalism Network.
5. Telegramas electorales bajo la lupa: Qué halló LA NACION al analizar más de 100.000 documentos | La Nación.
6. La Inteligencia Artificial llega a Clarín con UalterAI, un asistente de lectura | Clarín.
7. Chequeabot | chequeado.com.
8. Álvaro Liuzzi: “La IA está revolucionando el periodismo” | PARENTÉSIS MEDia.

SPANISH

Inteligencia artificial en el periodismo: la reconfiguración de las redacciones en Argentina

La rápida evolución de la inteligencia artificial (IA) está transformando industrias en todo el mundo, y el periodismo no es la excepción. En Argentina, organizaciones como Telecom Argentina, Chequeado y los diarios Clarín y La Nación están incorporando IA para optimizar sus prácticas, mejorar la precisión informativa y adaptar el contenido a distintos públicos. Estos avances no solo agilizan los flujos de trabajo, sino que también están redefiniendo el ecosistema de los medios, fomentando una sociedad más informada y conectada.

Las herramientas de IA, como ChatGPT, están cada vez más presentes en las redacciones argentinas.

Un informe de la Universidad de Buenos Aires reveló que el 82,2 % de los periodistas encuestados ha utilizado IA generativa en su trabajo, principalmente para la generación de ideas, redacción de titulares, elaboración de resúmenes, optimización SEO e incluso la producción de artículos completos.
En un contexto donde los presupuestos de las redacciones suelen ser reducidos y los recursos limitados, estas herramientas permiten optimizar los tiempos de producción y cumplir con los plazos de entrega. Además, el 71% de los encuestados subrayó la necesidad de establecer lineamientos éticos y ofrecer capacitación continua en el uso de la IA. Esto refleja cómo el periodismo argentino busca equilibrar el potencial de esta tecnología para mejorar la eficiencia sin descuidar la integridad editorial en un entorno en constante evolución. El informe Periodismo IA, publicado por Telecom Argentina, sirve como guía para los periodistas que buscan incorporar la IA en sus procesos de trabajo. La guía aborda temas como la automatización, el periodismo basado en datos y la personalización de contenidos.

En particular, la automatización está cambiando la dinámica en las redacciones, ya que la IA se encarga de tareas repetitivas como la transcripción, verificación de datos y clasificación de información. Por ejemplo, las herramientas de transcripción en tiempo real agilizan la producción de contenido, permitiendo a los periodistas llevar a cabo análisis con más profundidad. Al ser consultado sobre el impacto de la IA en el periodismo, Álvaro Liuzzi, autor de Periodismo IA, destacó que “la integración de la inteligencia artificial en las redacciones no solo se trata de eficiencia o innovación tecnológica, sino que también plantea profundas cuestiones éticas que afectan la manera en que se informa y se comprende el mundo”. En este sentido, el informe dedica un capítulo completo a la ética y transparencia en el uso de la IA, el cual se centra en los marcos y lineamientos para su implementación en la producción de noticias y procesos editoriales. La guía está disponible en español e inglés y puede descargarse de forma gratuita.

Más allá de optimizar los procesos de producción, la IA también permite nuevos enfoques en el análisis de fenómenos culturales y sociales en Argentina. Por ejemplo, La Nación, uno de los principales diarios del país, ha implementado herramientas de IA para profundizar en tendencias musicales y monitoreo electoral. Un caso notable es el uso de algoritmos para analizar las letras de los traperos argentinos (artistas de la escena musical trap), lo que reveló temas recurrentes como la desigualdad socioeconómica y la vida urbana. Este análisis basado en datos ofrece a los lectores una comprensión más profunda de las dinámicas culturales. Durante las elecciones, el diario utilizó IA para examinar telegramas y documentos electorales, identificando discrepancias que podrían haber pasado desapercibidas. Estas aplicaciones muestran cómo la IA puede mejorar el análisis periodístico, proporcionando una perspectiva más detallada sobre los fenómenos sociales.

Mientras tanto, Clarín, otro de los diarios más importantes de Argentina, ha desarrollado UalterAI, un asistente de lectura basado en IA que permite a los usuarios personalizar su experiencia informativa. Ofrece funciones como resúmenes concisos, lectura cronológica y listas de cifras o declaraciones clave. Por ejemplo, puede extraer los 10 puntos más relevantes de una noticia o reorganizar la información en columnas para facilitar su comprensión. Este nivel de personalización mejora la accesibilidad y permite que los lectores interactúen con las noticias según sus necesidades y preferencias. Con innovaciones centradas en el usuario como UalterAI, Clarín demuestra cómo la IA puede hacer que el periodismo sea más inclusivo y adaptable.

Por su parte, Chequeado, la principal organización de verificación de datos del país, representa un ejemplo del potencial de la IA para combatir la desinformación. La organización combina el rigor periodístico con herramientas de IA para procesar grandes volúmenes de información en línea, detectar patrones de desinformación y marcar afirmaciones dudosas. Además, ofrece herramientas para periodistas, como Chequeabot que identifica frases chequeables en textos y monitorea contenido en tendencia en redes sociales, entre otras aplicaciones. También desarrolla un software de transcripción, una calculadora de inflación y funciones que permiten el análisis de costos. Más allá de sus propias operaciones, Chequeado colabora con otros medios ofreciendo capacitación y herramientas para integrar la verificación de datos en sus flujos de trabajo. Este esfuerzo colectivo fomenta la confianza y la transparencia en los medios, elementos esenciales para combatir la propagación de la desinformación.

La IA no solo optimiza la producción periodística, sino que también permite a los periodistas enfocarse en contar historias significativas. Al promover la transparencia y mejorar el acceso a la información, las herramientas de IA están redefiniendo el periodismo para servir mejor al interés público. La incorporación de la tecnología en el periodismo argentino representa un momento clave en la evolución del sector. Las organizaciones mencionadas anteriormente están liderando el camino, demostrando cómo la AI puede mejorar las prácticas periodísticas manteniendo la integridad.
Al mismo tiempo, las cuestiones éticas siguen siendo fundamentales, con iniciativas en marcha para establecer lineamientos que aborden problemáticas como la desinformación, el sesgo y la limitación de la diversidad de perspectivas. Como bien señala Álvaro Liuzzi “La reflexión sobre cómo equilibrar la automatización con la preservación de la humanidad en el periodismo es crucial para el futuro de la profesión”. Un uso responsable de la IA garantizará que la tecnología continúe fortaleciendo el periodismo, reforzando su compromiso con la verdad, la responsabilidad y la narración de historias con impacto.

ENGLISH

AI, Empathy, and the Power of Not Forgetting Who We Are: Tech That Drives Inclusion in Education.

Although in every race there are typically more losers than winners, evolving as a species has always been a shared process. One does not advance alone. We all move forward together. In this world flooded with algorithms, perhaps the greatest innovation is keeping the flame of empathy alive. The kind that isn’t programmed or trained, but that naturally springs forth when we see someone in need of our help. Yet, as the rise of AI accelerates, we sometimes find ourselves caught in a frenetic race, feeling compelled to learn the latest tool or write the newest code. That’s why it’s important to remember that when it comes to education, the true victory is for all of us to cross the finish line.

In Mendoza, Argentina, a project is underway that uses artificial intelligence to detect early on who is at risk of dropping out of school. It analyzes attendance, grades, and class participation to identify students who are falling behind or might abandon their studies, aiming to provide targeted support.

The local government, along with several institutions such as CIPPEC and UBA, has contributed its share to implement this system in more than 100 schools across the province, focusing on those involved in the literacy and mathematics programs promoted by Mendoza’s Ministry of Education.

So far, they have managed to keep around 7,000 students from dropping out, reporting a 20% reduction in school desertion.

How Does It Work?

A teacher uploads the data to the platform, the AI detects a pattern of absenteeism that tends to correlate with dropouts, and it triggers an alert for the school to investigate the underlying reason. Perhaps the student lives far away, needs to work, or has to take care of siblings. Maybe the cause is more serious. Or perhaps it’s much simpler. The truth is, without the alert, that student would have been lost in the statistics. Now, instead, ones and zeros are arranged in such a way that they transform into concrete support. A little push. A “you got this” moment.

As in every part of the world, the adoption of AI in Argentina depends on connectivity. Nearly 90% of the population has internet access, although quality and stability vary by region. In some rural areas, the connection remains intermittent or even non-existent. In recent years, programs that distributed netbooks and promoted digital culture in public schools have been implemented, but the landscape isn’t always equal. Some institutions suffer from poor Wi-Fi and limited infrastructure to sustain large-scale technological projects.
At first glance, this might seem like just another case of AI solving a problem: analyze data, cross variables, light up a signal when a student is at risk of dropping out. But it’s also an example that, while looking forward is vital for discovering new horizons, noticing those around us is key to reminding us we’re in this together. To remain competitive, we need to embrace new technologies. However, if even one student is overlooked, the system fails. Rather than matching the pace of the slowest, innovation can be designed to drive our collective forward.

Projects like these provide hope for a tech-powered education system where no one is left behind. In that expected and near future, AI will have served to reinforce what makes us human.

SPANISH

IA, empatía y el poder de no olvidar quiénes somos: La tecnología que impulsa la inclusión en la educación.

Aunque en toda carrera es normal que haya más perdedores que ganadores, evolucionar, como especie, siempre fue un proceso compartido: no avanza uno solo, avanzamos todos.

En este mundo inundado de algoritmos, quizás la mayor innovación sea mantener viva la llama de la empatía, la que no se programa ni se entrena, la que brota sola cuando vemos que alguien necesita de nuestra ayuda. Sin embargo, a medida que crece el auge de la IA, a veces nos vemos inmersos en una carrera frenética, con la sensación de tener que aprender la última herramienta o escribir un código nuevo antes que nadie. Por eso es importante recordar que, cuando hablamos de educación, la verdadera victoria es que la meta la crucemos todos.

En Mendoza, Argentina, se está llevando a cabo un proyecto que usa inteligencia artificial para detectar tempranamente quién está en riesgo de dejar la escuela. Analiza asistencia, notas y participación en clase, lo que permite identificar a quienes están quedando atrás o podrían abandonar, con el objetivo de brindar apoyo específico.

El gobierno local y varias instituciones, como CIPPEC y la UBA, pusieron su granito de arena para que este sistema se aplique en más de 100 escuelas de la provincia, concentrándose en aquellas que participan de los planes de alfabetización y matemáticas impulsados por el Ministerio de Educación provincial.

Hasta ahora, lograron que alrededor de 7000 estudiantes no abandonaran sus estudios, reportando una reducción del 20% en deserción escolar.

¿Cómo funciona?

El docente carga los datos en la plataforma, la IA detecta un patrón de inasistencia que tiende a correlacionar con abandono y dispara la señal para que la escuela estudie la razón. Tal vez el alumno viva lejos, necesite trabajar o cuidar a sus hermanos. Tal vez la causa sea más grave. O más simple. Lo cierto es que, sin la alerta, ese chico se hubiera perdido en la estadística. Ahora, en cambio, unos y ceros se disponen de tal forma que se convierten en apoyo concreto. Un empujoncito. Un “dale, vos podés”.

Como en todo el mundo, la adopción de IA en Argentina depende de la conectividad. Casi el 90% de la población tiene acceso a internet, aunque la calidad y estabilidad varían según la región. En algunas zonas rurales, la conexión sigue siendo intermitente o inexistente. En los últimos años hubo programas que entregaron netbooks y fomentaron la cultura digital en las escuelas públicas, pero el panorama no siempre es parejo. Hay instituciones con Wi-Fi deficiente y poca infraestructura para sostener proyectos tecnológicos a gran escala.
A simple vista, éste parece un caso más de IA resolviendo un problema: analiza datos, cruza variables y avisa cuando un alumno está por dejar la escuela. Pero también es un ejemplo de que, si bien mirar hacia adelante es vital para descubrir nuevos horizontes, observar a nuestro alrededor es fundamental para recordarnos que no estamos solos. Está claro que para ser competitivos debemos adoptar nuevas tecnologías. Pero si le soltamos la mano a un solo alumno, el sistema falla. En lugar de avanzar al paso del más lento, la innovación puede diseñarse para impulsar nuestro progreso colectivo.

Este tipo de proyectos ofrecen la esperanza de un sistema educativo impulsado por la tecnología, en el que nadie quede excluido. En ese futuro esperado y cercano, la IA habrá servido para reforzar lo que nos hace humanos.

ENGLISH

Artificial Intelligence and the Promotion of Lengua de Señas Argentina

Artificial intelligence has emerged as a transformative force in enhancing accessibility and fostering inclusion for individuals with disabilities. In Argentina, several innovative projects harness AI to facilitate communication for the deaf and those with speech impairments — a community that holds significant recognition in the country. The deaf community, in particular, is institutionally represented by the Confederación Argentina de Sordos (CAS), which advocates for their rights and well-being. As stated on its official website, CAS’s mission is to “improve the quality of life for deaf people nationwide and to preserve, protect, and promote their human, civil, and linguistic rights”.¹ Aligned with CAS’s vision, these AI-powered initiatives not only enhance social integration but also contribute to the broader discourse on the ethical and practical applications of AI for social good. In response to Imminent’s Question of the Year, this piece explores key AI-driven projects in Argentina that aim to teach and translate Lengua de Señas Argentina (LSA) (the official sign language in Argentina), promote inclusivity in educational settings, and support individuals with communication disabilities through mobile technology.

AI Applied to the Teaching of LSA

The Universidad Nacional de Buenos Aires (UBA) has developed an AI-powered platform for teaching LSA, which will be integrated into the online courses offered by the Centro Universitario de Idiomas (CUI).² Coordinated by Romina Aza and supported by the Universidad Tecnológica Nacional (UTN), this platform seeks to expand access to and dissemination of LSA. The platform enables students to learn through instructional videos with sign language instructors and to practice with AI-powered technology that analyses hand and facial movements, providing real-time feedback. A key collaborator in this project is Enseñanza de Lengua de Señas (ELDES), an organization specializing in LSA education that pioneered the development of Argentina’s first interactive sign language teaching platform. This initiative represents a significant breakthrough, as it is the first in Argentina to incorporate AI into sign language learning.
In parallel with UBA’s advancements, the Government of the Province of Córdoba, in partnership with the company Contraseñas S.A.S – Dillo, has announced the implementation of an interactive application designed to teach LSA in schools across the province.³ This initiative aims to promote educational inclusion and equal opportunities by ensuring that all students, not just those who are deaf, learn LSA and are able to communicate. A pilot test will be conducted in select schools before expanding the program province-wide. According to Manuel Flores, a representative from Dillo, the application is not meant to replace sign language interpreters but rather to promote interaction and foster awareness, respect, and social integration of deaf individuals.

Enhancing Communication for Deaf Individuals

Beyond LSA education, two additional AI-based projects are advancing communication accessibility for deaf individuals in Argentina. One such project is an LSA interpreter robot, developed by students Santiago Vázquez, Juan Cruz Ledesma, and Tomás Donoso from the Tomás Alva Edison School.⁴ This robot is designed to decode LSA signs and translate them into Spanish in both written and spoken formats, thereby improving interactions between deaf and hearing individuals. It operates through a combination of programming, AI, and 3D printing. Using an AI powered image interpreter trained on a database of sign language gestures, the robot translates signs into text and speech in real time. Although still in development and currently capable of recognizing only a limited set of letters- particularly those with simpler movements-the project holds significant potential in fields such as healthcare and education. The AI system behind the robot continuously learns from LSA sign patterns, enhancing its accuracy and adaptability over time.
Another noteworthy project is a mobile application designed to translate LSA into Spanish in real time, developed by PhD. Pedro Dal Bianco at the Universidad Nacional de La Plata (UNLP).⁵ This application utilizes AI algorithms to analyse images and videos, identifying hand shapes and movements to facilitate translation.

In addition to improving communication for individuals with hearing impairments, the project is contributing to the creation of an LSA database, which will serve as a valuable resource for future technological developments. The application was showcased at the Universidad Nacional de Quilmes (UNQ) during Disability Awareness Week and is expected to benefit over one million people in Argentina with hearing difficulties.

Háblalo: Inclusive Communication Technology for People with Disabilities

While numerous innovations surrounding LSA have emerged in Argentina, none have achieved the scale and impact of Háblalo.⁶ Created by Mateo Salvatto in late 2016 as a personal initiative to help deaf students communicate more effectively, Háblalo began as a small-scale project that quickly gained traction, reaching 20,000 users in Argentina.

As the project grew, Salvatto partnered with Sergio Cantarovici and Laura Muchnik to establish Asteroid, the company that transformed Háblalo into a global platform. Today, Háblalo supports more than 350,000 individuals with communication disabilities across 65 countries on five continents. The application remains free of charge, functions without requiring an internet connection, and is available in 40 languages.

The philosophy behind Háblalo centers on redefining conventional approaches to inclusivity, aiming to create meaningful, long lasting social impact. This commitment led to the development of Háblalo for Business, a model designed to help companies and institutions become more inclusive by offering accessible services to an audience of over 500 million individuals worldwide.
Over the past six years, Háblalo has received numerous prestigious awards, including the Humanitarian Innovation of the Year Award from MIT and the Best Social Development

Award from Peking University. Additionally, Asteroid has been recognized as one of the Top 100 Startups in the World by the Global Entrepreneurship Network and secured first place in the Santander X Global Awards.
The success and impact of Háblalo have been widely acknowledged by global media outlets, including CNN⁷ and Forbes Argentina,⁸ which have highlighted its role in transforming communication accessibility for individuals with disabilities.

Bridging Communication Gaps: AI as a Catalyst for Inclusivity and Accessibility

The integration of artificial intelligence in sign language education and communication technology marks a transformative shift in accessibility for deaf individuals. The projects discussed here illustrate how AI-powered tools can bridge communication gaps, foster inclusivity in both educational and social environments, and provide vital resources for individuals with communication disabilities. By leveraging AI for social good, these initiatives exemplify how technology can be harnessed to build a more equitable and inclusive society. In the words of Mateo Salvatto: “People with disabilities need to be able to access all types of services, and companies need to consider accessibility as an investment that adds value in many ways; only then will we achieve real inclusion.”⁹

To make accessibility a widespread reality, sustained collaboration among the deaf community, policymakers, educators, and technology developers will be crucial in refining these tools and ensuring their long-term impact. As AI-driven innovations continue to evolve, their role in dismantling communication barriers and strengthening human connections will only grow more profound.

References
1. Confederación Argentina de Sordos | CAS.
2. La UBA desarrolló una inteligencia artificial para enseñar lengua de señas: ¿cómo funciona? | iproup.com.
3. Implementarán la inteligencia artificial para enseñar lengua de señas en escuelas provinciales | prensa.cba.gov.
4. Inteligencia artificial y Lengua de Señas Argentina: presentaron un proyecto de robot que interpreta | Facultad de Educación.
5. La Inteligencia artificial aplicada a la Lengua de Señas Argentinas | UNQ.
6. Háblalo – Sobre nosotros | Háblalo.
7. Mateo Salvatto creó “Háblalo”, una app para ayudar a las personas sordas | CNN.
8. Santander reconoce a las startups más innovadoras del mundo (entre ellas, hay una argentina) | Forbes.
9. La innovadora aplicación de Mateo Salvatto se expande en el mundo para derribar barreras de comunicación | ambito.

SPANISH

La inteligencia artificial y la promoción de la Lengua de Señas Argentina

La inteligencia artificial se está convirtiendo en una fuerza transformadora para mejorar la accesibilidad y fomentar la inclusión de las personas con discapacidad. En Argentina,
varios proyectos innovadores hacen uso de la IA para mejorar la comunicación de las personas sordas y con problemas del habla, una comunidad que goza de un importante reconocimiento en el país. La comunidad sorda, en particular, está representada institucionalmente por la Confederación Argentina de Sordos (CAS), que defiende sus derechos y aboga por su bienestar. Como se indica en su sitio web oficial, la misión de la CAS es “[c]ontribuir a una mejora sustancial de la calidad de vida de las personas sordas de todo el país y a la vez, preservar, proteger y promover los derechos humanos, civiles y lingüísticos de todas las personas”.
En línea con la visión de la CAS, estas iniciativas que incorporan la IA no solo fortalecen la integración social, sino que también contribuyen al debate más amplio sobre las aplicaciones éticas y prácticas de la IA para el bien social. En respuesta a la Question of the Year de Imminent, este texto explora los principales proyectos impulsados por IA en Argentina que tienen como objetivo enseñar y traducir la Lengua de Señas Argentina (LSA), promover la inclusión en los entornos educativos y apoyar a las personas con discapacidades comunicativas a través de la tecnología móvil.

La IA aplicada a la enseñanza de la LSA

La Universidad Nacional de Buenos Aires (UBA) desarrolló una plataforma basada en IA para la enseñanza de la LSA, que se va a integrar a los cursos en línea que ofrece el Centro Universitario de Idiomas (CUI). Esta plataforma, que está coordinada por Romina Aza y cuenta con el apoyo de la Universidad Tecnológica Nacional (UTN), busca ampliar el acceso a la LSA y su difusión. La herramienta permite que quienes están estudiando aprendan a través de videos instructivos con docentes de lengua de señas y puedan practicar con tecnología impulsada por IA. Esta tecnología analiza los movimientos faciales y de las manos y brinda comentarios en tiempo real, para permitir la autoevaluación del estudiantado. En el proyecto, resulta clave la colaboración de Enseñanza de Lengua de Señas (ELDES), una organización especializada en la enseñanza de la LSA que fue pionera en el desarrollo de la primera plataforma interactiva de enseñanza de lengua de señas en Argentina. Esta iniciativa representa un avance significativo, ya que es la primera que se incorpora la IA al aprendizaje de la lengua de señas en el país. Paralelamente a los avances de la UBA, el Gobierno de la Provincia de Córdoba, en alianza con la empresa Contraseñas S.A.S – Dillo, anunció que va a implementar una aplicación interactiva diseñada para la enseñanza de la LSA en escuelas de toda la provincia. Este proyecto busca promover la inclusión educativa y la igualdad de oportunidades para garantizar que todo el estudiantado, no solo quienes tengan problemas de audición, aprendan LSA y sean capaces de comunicarse. Antes de expandir el programa a nivel provincial, se van a realizar una prueba piloto en determinadas escuelas. Según Manuel Flores, representante de Dillo, la aplicación no pretende reemplazar a los intérpretes de lengua de señas, sino promover la interacción y fomentar la toma de conciencia para contribuir al respeto y la integración social de las personas sordas.

Mejorar la comunicación de las personas sordas

Más allá del ámbito de la educación, existen otros dos proyectos basados en IA que están mejorando la accesibilidad a la comunicación de las personas sordas en Argentina. Uno de ellos es un robot intérprete de LSA, desarrollado por los alumnos Santiago Vázquez, Juan Cruz Ledesma y Tomás Donoso del Colegio Tomás Alva Edison. Este robot está diseñado para decodificar las señas de la LSA y traducirlas al español, tanto en formato escrito como hablado, con el objetivo de mejorar las interacciones entre personas sordas y oyentes. Funciona mediante una combinación de programación, IA e impresión en 3D. Aunque todavía está en fase de desarrollo y, por el momento, solo es capaz de reconocer un número limitado de letras, sobre todo las que tienen movimientos más sencillos, el proyecto tiene un gran potencial en campos como la salud y la educación.
Otro proyecto destacado es la aplicación móvil que traduce de la LSA al español en tiempo real, desarrollada por el PhD. Pedro Dal Bianco, de la Universidad Nacional de La Plata (UNLP). Esta aplicación usa algoritmos de IA para analizar imágenes y videos, identificando las formas y los movimientos de las manos para generar la traducción. El sistema aprende de form continua los diferentes patrones de señas de la LSA, lo que le permite mejorar su precisión y adaptabilidad a lo largo del tiempo. Además de facilitar la comunicación de las personas con problemas de audición, el proyecto está generando la primera base de datos de LSA del país, un recurso sumamente valioso para futuros desarrollos tecnológicos. La aplicación se presentó en la Universidad Nacional de Quilmes (UNQ) en el marco de la Semana de la Discapacidad y se espera que beneficie a más de un millón de personas con dificultades auditivas en Argentina.

Háblalo: tecnología de comunicación inclusiva para personas con discapacidad

Si bien en Argentina surgieron numerosas innovaciones en torno a la LSA, ninguna alcanzó la escala o el impacto de Háblalo. Háblalo surge en 2016 como una iniciativa personal de Mateo Salvatto para ayudar a la comunidad sorda de estudiantes a comunicarse de manera más efectiva. Lo que comenzó como un proyecto a pequeña escala rápidamente ganó popularidad y llegó a 20 000 usuarios en Argentina. Cuando el proyecto creció, Salvatto se asoció con Sergio Cantarovici y Laura Muchnik para fundar Asteroid, la empresa que transformó Háblalo en una plataforma global. En la actualidad, Háblalo ayuda a más de 350 000 personas con discapacidades comunicativas en 65 países de los cinco continentes. La aplicación es gratuita, funciona sin necesidad de conexión a Internet y está disponible en 40 idiomas.
La filosofía de Háblalo se centra en redefinir los enfoques convencionales de la inclusión, con el objetivo de crear un impacto social significativo y duradero. Este compromiso llevó al desarrollo de Háblalo for Business, un modelo que pretende ayudar a empresas e instituciones a ser más inclusivas ofreciendo servicios accesibles a una audiencia de más de 500 millones de personas en todo el mundo. En los últimos seis años, Háblalo recibió numerosos galardones de prestigio, como el premio a la Innovación Humanitaria del Año del MIT y el premio al Mejor Desarrollo Social del Mundo de la Universidad de Pekín. Además, Háblalo fue reconocida como una de las 100 mejores startups del mundo por la Global Entrepreneurship Network y obtuvo el primer puesto en los Santander X Global Awards. El éxito y la repercusión de Háblalo se reflejan en medios de comunicación de todo el mundo, como la CNN y Forbes Argentina, quienes destacaron su papel en la transformación de la accesibilidad a la comunicación de las personas con discapacidad.

Tender puentes en la comunicación: la IA como catalizador de la inclusión y la accesibilidad

La integración de la IA en la enseñanza de la lengua de señas y la tecnología de la comunicación marca un cambio transformador en la accesibilidad para las personas sordas. Los proyectos que se exponen aquí ilustran cómo las herramientas basadas en IA pueden cerrar brechas comunicativas, fomentar la inclusión en entornos educativos y sociales, y proporcionar recursos esenciales para las personas con discapacidades comunicativas. Al aprovechar la IA para el bien social, estas iniciativas ejemplifican cómo puede aprovecharse la tecnología para construir una sociedad más equitativa e inclusiva. En palabras de Mateo Salvatto “Las personas con discapacidad tienen que poder acceder a todo tipo de servicios y las empresas tienen que considerar el ser accesible como una inversión que suma en un montón de sentidos, solo así logramos inclusión de verdad”. Para que la accesibilidad sea una realidad a gran escala y resulte posible perfeccionar estas herramientas y garantizar su impacto a largo plazo, va a ser imperante la colaboración sostenida entre la comunidad sorda, quienes tienen una responsabilidad política, las personas del ámbito de la educación y quienes desarrollan tecnología. A medida que las innovaciones basadas en IA sigan evolucionando, su papel en el desmantelamiento de las barreras comunicativas y el fortalecimiento de las conexiones humanas va a resultar de mucha más relevancia.

ENGLISH

Rethinking Inclusion and Education in Mali through Artificial Intelligence

Is Artificial Intelligence an opportunity for education in developing countries like Mali?
Mali is one of the West African Countries with the lowest literacy rate. Many people, especially young people, have heard about or know what artificial intelligence is but unfortunately ignore how it is commonly used. In 2023, the President of the Republic of Mali laid the foundation stone of the Centre for Artificial Intelligence and Robotics after the impressive success of the Malian team in the USA during the 2023 Vex Robotics World Championship, an international robotics competition.
The aim is to make the country one of the global leaders in AI. Since then, there are more and more young people interested in AI. Today, there are very few young people or companies that implement artificial intelligence in their day-to-day businesses.

“Dicken” is an agency specializing in the automation of customer support. Dicken Media creates AI Chatbots “for customer service, in house training, after-sales service and AI consulting.” The company helps its customers get the most out of AI, by helping them save time and money.

AI may be a great tool for learning in schools and even in everyday life. It can provide students and teachers with quality educational content and open up new perspectives for inclusion and equal opportunities in education. In the coming years, it will be a great asset that will help improve the accessibility and quality of education in developing countries like Mali. That is why I think the integration of AI into the curriculum of higher education will enhance and build the capacity of many students to use it. An example is an innovative project developed by Malirobots that seeks to transform education in Mali using cutting-edge technologies. At the heart of the initiative is the desire to strengthen proficiency in the local language, Bamanankan (or Bambara), as an essential part of learning. Bamanankan is one of the most widely spoken languages in Mali, and the project believes that it is essential to develop quality educational resources for children in their mother tongues, starting with Bambara. The approach is based on Generative AI. By leveraging AI, they create compelling stories, books, and learning materials in Bamanankan that are perfectly suited to the skill levels of aspiring readers. The stories are rooted in Malian culture, which makes them relevant and engaging.

I think that artificial intelligence will help clear schools that are overcrowded in Mali and might promote online teaching.
In addition to education, Artificial Intelligence can help create more jobs and make research easy.

It can also be useful in healthcare, security and agriculture. EWAATI is a specialized startup in Artificial Intelligence technologies and robotics. It provides a range of facial recognition technology and data analysis for the sectors of security and public safety, transparent personnel management in enterprises, finance, education, marketing a series of platforms for data analysis as well as a number of terminals for scanning and facial recognition. For the past few months, the Point G Hospital, a public hospital, in Bamako has been at the forefront of e-health. In partnership with HealthTech Mali, it has introduced AI systems in its cardiology department to optimize the diagnosis and treatment of patients with heart disease. Through this collaboration, cardiologists benefit from advanced AI tools, making it easier to interpret medical data and make informed decisions.
On the other hand, I think AI will have a negative impact on its users over time and hinder the development of their skills. It can promote mental laziness, and create a technological dependence on people who use it and obstruct their ability to think or make in-depth analyses. For instance, what will happen in the next years if there is a shutdown in AI data centres around the world as did WhatsApp or Facebook in recent years? Today, it’s important for our countries to have national policies setting limits on the use of AI.
Is AI an efficient tool in the promotion of cultural diversity?
The advent of artificial intelligence can be a great asset in promoting the 13 official languages spoken throughout Mali. Although there are few online educational resources for some languages spoken in the country, AI may help translate and popularize them. This will help speakers of those languages access online information in their mother tongues. Artificial Intelligence may help break down language barriers by making it easier for a non-native speaker to learn a language without the assistance of a native speaker. For example, chatbots can be developed to act as virtual tutors for people who would like to learn new languages. Today, it’s important that all languages, whether they are minority, majority or even foreign, are given equal treatment. For example, on online machine translation platforms, many African languages are not available. There are few African, even Malian languages with databases available online to facilitate their learning through AI. Also, there is no software available for the translation of those languages. This situation raises a key question: is artificial intelligence a solution to cultural diversity? I think the answer is yes, provided that some languages are not discriminated against on digital platforms. To cope with that situation, it’s necessary to find ways to translate them from Artificial Intelligence as we are living in an era where cultural diversity is promoted. The development of machine translation could significantly help develop a linguistically and culturally rich society. In other words, breaking down language barriers can change the world.

References
1. A la découverte du centre Robots Mali | YouTube. robotsmali.org.
2. The President Assimi Goïta launches the Centre for AI and Robotics Mali | YouTube.
3. Dicken AI Media Agency.
4. E-santé: l’IA change la donne au Mali – Resilient Digital Africa | resilient.digital-africa.co.
5. eWaati Agency.

TAMASHEQ

Ăsǝtbăt n-ăsǝrtǝy ǝd teɣăre dăɣ Mali s tăyǝte ta-d exlăk awădim

Ăǧǧen tăyǝte ta-d exlăk awădim t-ămos ăroš dăɣ ezarăg wa-n teɣăre i ǝdduwlăn wi wi wăr ǝyyiwăn? t-ămos Mali ǝyyăn dăɣ ǝdduwlăn n Ăffreqiya ta-n atarăm s wǝr ǝǧǧităn eddinăt wi ăɣranen. eddinăt ǝǧǧotnen hullăn tǝmmǝdrit ǝssanăn meɣ ǝslăn a-wa-s itawănna tăyǝte ta-d exlăk  awădim mušăn wăr ǝssenăn a-wa tǝnfa ălxǝdmăt sǝr-s.  dăɣ ăwatăy wa-n 2023, ămanokăl n ǝdduwl wa-n Mali ǝssǝnsa tekadăyt ta-n edey n ehăn wa-n  tăyǝte ta-d exlăk awădim ǝd robotăn darăt ǝrrizăɣ wa tǝkrăš tăggayt n Mali dăɣ tămadăšt n  ălɣalim ta tigăt dăɣ ăwatăy wa-n 2023 dăɣ Ămrik s itawănna “Vex Robotics World  Championship.
ittus n ăwen wăr ămos ăr-a tumăst Mali ǝyyăn dăɣ ǝdduwlăn wi-n muzarăn n  tăyǝte ta-d exlăk awădim dăɣ ălɣalim. A iǧa ăwen dăɣ,ăǧǧen dăɣ tǝmmǝdrit tăssoɣăl-in  ǝnniyăt net tăyǝte ta-d exlăk awădim. dăɣ wityăn wi dăɣ n-ăha wǝr tǝǧǧit tǝmmǝdrit ǝd dăɣ ălxǝdmaten nasăn.  

Diken t-ămos tăggayt s ălxǝdmăt net wăr ămos ăr ăsǝrɣǝs n tǝdhǝld n eddinăt.Diken Media  ɣăn a-wa-s ităwanna chatbots i ălxǝdmăt n eddinăt, teɣăre ɣor hanăn, tămašɣult ta-n darăt  inaššăn ǝd “.t-ămos hăr ăǧudi a-idăhhalăn inummušɣăl net dăɣ ăkarăš n tǝnfa n tăyǝte ta-d  exlăk awădim s ăggăz n alwăq d ăzrǝf.  

tăyǝte ta-d exlăk awădim t-ămos tǝǧittewt măqqorăt i ăsǝssaɣǝr dăɣ lăkkolăn ǝd dăɣ ǝdduniya.  ăddobăt ăd ekf inittulăb d-imăsaɣrăn tolăs ǝtir hassăn dăɣ tărtit ǝd tǝgdăht dăɣ teɣăre.dăɣ tǝlkamăt, ăd eqqǝl tǝnfa măqqorăt he tǝdhǝlăt ăsǝrɣǝs ǝd tulaɣăy n teɣăre dăɣ ǝdduwlăn wi  wăr nǝyyiwăn šund Mali.ăwen dăɣ ăs iǧa ɣur-i s ugǝz n IA dăɣ lăkkolăn wi măqqornen ăd  isǝmmuti tolăs isǝssuhit musnăt n-inittulăb dăɣ ălxdǝmăt net.  

Ălmităl ǝyyăn ămos broje wa-d eɣna Malirobots s ittus net ămos tămuttăy n teɣăre dăɣ Mali s  ălxidmăt s teknoloji. dăɣ ămmăs n ittus wen ille derhăn n ăsǝssuhi n musnăt n teɣăre dăɣ ăwal  n ăkal wa-s ităwanna tabănbarăt.tabănbarăt t-ămos ăwal wa ǝssăwalăn eddinăt s ǝǧit dăɣ Mali  tolăs broje orda-s ămos harăt măqqorăn efes n isuǧ n teɣăre olăɣnen i ilyadăn dăɣ tăwalen  nasăn dăɣ as ezzăr s tabanbarăt.  

tǝkărse ten-hi tiǧa făl tăyǝte ta-d exlăk awădim. taɣăn hăr ăǧudi tǝnfusen olăɣnen, ilkittabăn d  ǝsuǧ n teɣăre dăɣ tabănbarăt iǧănen i musnăt n imaɣrăn. tǝnfusen tin-hi iǧanăt făl ăttarex n  Mali ăwen dăɣ a tanăt isahuskiyăn  

tăyǝte ta-d exlăk awădim ăd idhǝl dăɣ tănnakmăma ta tiǧăt dăɣ lăkkolăn dăɣ Mali tolăs  ăddobăt efes n teɣăre făl ănternet.  

darăt teɣăre, tăyǝte ta-d exlăk awădim tăddobăt ăddahalăt dăɣ eɣăn n ǝššǝɣǝlăn d ăsǝrɣǝs n  farăk. tăddobăt tolăs ăd t-ănfu dăɣ ezarăg wa-n ăăssexăt, ǝd tăzǝggăwt d tǝwgăs.  

EWAATi t-ămos tǝsǝssăhǝrt s ălxǝdmăt nent iǧa făl tăyǝte ta-d exlăk awădim ǝd  robotăn.Tahăk teknoloji s ălxǝdmăt net ihuk n salăn făl udmawăn d ămǝyyǝz n salăn d 

ăqqalnen ihăndaggăn wi-n tăzǝggăwt d eddinăt, tamăšɣult n eddinăt dăɣ tǝsǝssăhǝrt, d-ăzruf,  teɣăre ǝd sǝblisitetăn d ihăndaggăn wi-n ămǝyyǝz n udmawăn d ăfarăd  

a ilăn orăn,lăɣtur wa-s itawănna Point G ehăn Bamako inta deɣ issǝnta ălxǝdmăt s ăssexăt s  alhăndasăt.Lăɣtur wen-dăɣ tǝdhăl tǝsǝssărhǝrt s itawănna Health Tech Mali ǝssǝnta ălxǝdmăt  s ǝsuǧ wi-n tăyǝte ta-d exlăk awădim dăɣ ehăn n ăssexăt net wa-n ulhawăn i ăsǝrɣǝs n ămǝyǝz  ǝd ăsafăr n imărhinăn n torhinnăwen ti-n ulh.ănmǝšɣăl wen-dăɣ issǝrɣăs ǝd i imăssafarăn n  torhinnăwen ti-n ulh musnăt s ălxǝdmăt s isuǧ wi-n tăyǝte ta-d exlăk awădim hasăn-d  ăssǝrɣasăn ăšǝnšǝš n ămǝyyǝz d ătkul n ăddarabatăn olăɣnen.  

dăɣ ăšrut ǝyyăn, ordeɣ hullăn as dăɣ tilkamăt as tăyǝte ta-d exlăk awădim a tǝkrǝš tikma  lăbasăt făl eddinăt wi sǝr-s ăxdămnen d efes n tudabăten nasăn.Addobăt a sǝr-săn d-ăruw  tăzzayt n tăyǝtte nasăn d eɣăn n ătteɣmăr iǧăn făl teknoloji d ăgadăl n imǝyyǝzăn ăssohătnen. ălmităl n ăwen ămos as ma he eǧin a făl dăɣ tǝlkamăt imăl-t uɣšăd dăɣ ihanăn wi-n tăyǝte ta-d  exlăk awădim šund a-wa-s kăla iǧrăw Watsăb ǝd Feysbuk ăhin okayăn dăɣ iwityăn?.  

i olaɣăn as dăɣ iwityăn wi dăɣ năha, ăd issimilăn ǝdduwlăn nănăɣ folitik d-ǝd tăqqalăt  ălxǝdmăt n tăyǝte ta-d exlăk awădim  

ăǧen tăyǝte ta-d exlăk awădim tăqqăl isuǧ olaɣăn i efes n ăgnatăn ?  

emel n tăyǝte ta-d exlăk awădim ămos ăsuǧ măqqorăn i efes n mărawăt tăwalen n ǝdduwl wa n Mali.kuddăɣ as illăn-t ǝlkittabăn n teɣăre făl ănternet iǧănen dăɣ tăwalen t-ǝyyăd n ǝdduwl,  tăyǝte ta-d exlăk awădim t-ăddobăt ăd ǝdhǝl dăɣ ăfǝssǝr d ălmud făl-assăn.ăwen ăd ǝdhǝl kel  tăwalen ti-n hi dăɣ ăǧarăw n salăn dăɣ tăwalen nasăn.tăyǝte ta-d exlăk awădim ăddobăt ăd erz iɣalayăn ǧer ăwalăn s ăsǝrɣǝs n i eddinăt n tăwalen tǝyyăd ălmud năsnăt dăɣ as sǝsaɣrăn imăn  nasăn.ălmităl n ăwen ămos as ille a-wa-s itawănna chatbots s a făl-d aɣnăn ăddoben ăd  ăqqilăn imăssaɣarăn i edidnăt wi ărhanen ălmud n tăwalen tǝyyăḍ.

ămos a olaɣăn as tăwalen ɣǝrǝd năsnăt ogdăh-năt ǧer ɣăs ăbrur-năt,ăssohăt-năt meɣ ti-n ǧer  kallăn.ălmităl hăndaggăn wi-n tăsfsǝr n tăwalen făl ănternet, tăwalen ti-n Ăfreqqiya wăr-t illenăt.illanăt tăwalen n Afreqqiya hăkǝd dăɣ ti-n Mali ilănen ilkittabăn făl ănternet  ăddobătnen ăsǝrɣǝs n ălmud năsnăt s tăyǝte ta-d exlăk awădim.tolăs wăr-t-illa ăsuǧ n ăfǝssǝr n  tăwalen ti-n hi. tălɣa ten-hi torăw-d ăsǝstăn: ăǧen tăyǝte ta-d exlăk awădim t-ămos ăddabara i  ăgnatăn?.iǧa ɣur-i as ălwajăb n ăwen ămos iyya a făl wăr eqqel ăwen hăsnăt tǝkma făl  hăndaggăn wi-n ănternet.Ǝlzăm utǝr n ăddabaratăn i ăfǝssǝr n năsnăt s tăyǝte ta-d exlăk  awădim făl-as nǝzzăɣ dăɣ ăzzamăn n efes n ăgnatăn ǝǧǧotnen. efes n tăfsǝr eɣna ălhandasăt  ăddobăt ăd eɣăn efes n ăwalăn d ăgnatăn dăɣ tămazdoq tăssohet. dăɣ ăwen terăzze n iɣalayăn  ǧer ăwalăn ăddobăt ăd ăwǝy-d tămuttăyt n ălɣalim. 

ENGLISH

Bridging the Digital Divide: The Philippines’ AI-Driven Language Revolution

The enthusiasm for artificial intelligence is remarkably high among Filipinos. A recent study by Linkee from early 2024 positions the Philippines as the fourth most AI-obsessed country globally – mainly driven by the monthly search volume for all AI tools combined. Nevertheless, a significant digital divide remains, with over 30 million individuals lacking internet access. This challenge is primarily caused by poor digital infrastructure, high costs, and a shortage of internet services. In 2017, long before ChatGPT was on the scene, a project centered on natural language processing for the two predominant languages in the Philippines began to develop. By 2022, it had matured into a three-year project called “iTanong.” This name is a witty play on words, as “itanong” means “to ask” in Filipino, and the initials I and T for information technology are seamlessly weaved into it.
What sets iTanong apart from other AI language models is its ability to let Filipinos ask questions in their own language. However, the initial roll out will be in Filipino, English, and Taglish. Created by the Advanced Science and Technology Institute (ASTI), it serves as an essential tool in a country where English is the primary language used for official communication. It is utilized across various domains, including official government websites, meetings, reports, educational materials, and print media, among others. This contradicts the stipulations of the 1987 Constitution, which mandates that the government must actively encourage and support the use of Filipino as the official language of communication.

Enhancing Inclusivity and Accessibility
in Information Sharing

Not until recently, the Philippine government is taking proactive measures to highlight the importance of using Filipino, recognizing the need for a language that is inclusive and accessible to all Filipinos. A significant example of this initiative is a proposed bill that requires government agencies and local government units (LGUs) to translate public service announcements into Filipino and various languages used across the country. Senate Bill No. 680, known as the Language Accessibility of Public Information on Disasters Act, points out that disseminating information in our local languages is more effective, as most individuals find these languages easier to grasp than English. It must be noted that the use of the English language is highly concentrated in the metropolitan cities. The bill encompasses advisories related to natural disasters, extreme weather conditions, social welfare programs, emergency support, other social protection efforts, disease outbreaks, military and police security matters, and emergencies caused by human actions.

The Merits of Localized AI

Despite the fact that most Filipinos can read and write in English, only 55% of adults are fluent in the language. iTanong’s dedicated and localized approach showcases its potential to address the specific requirements of Filipino users. With its tailored features, it effectively navigates the linguistic diversity in the country, providing access to crucial information with the help of artificial intelligence. Based on the information presented on the ASTI website, the project’s primary objective is to enhance information accessibility through the following:
• Develop a natural language querying engine (NLQE) capable of interacting with any database or flat-file tables to extract valuable information or insights.
• Conduct and facilitate cutting-edge research in machine learning, deep learning, data science, natural language processing, and artificial intelligence.
• Provide capacity-building programs for researchers and technology users.
• Promote the widespread implementation of technology across other government entities.
• Commercialize the products and technologies resulting from the research.

Although the system shows great promise, it will initially be rolled out in government offices to succeed the Citizen’s Charter before releasing it to the public. This document currently serves as an essential tool for these offices, helping inform visitors about their service standards. Once iTanong is implemented, it will allow for straightforward and prompt updates as necessary. The increasing impact of AI in the Philippines is promising, yet accessibility poses a significant challenge. Filipinos are enthusiastic about adopting the technology, but language and infrastructure issues hinder its widespread use. This is where initiatives like iTanong play a vital role, providing AI-driven information in languages used locally, to benefit a broader audience. However, achieving real advancement goes beyond just cutting-edge technology, it necessitates government backing, improved digital infrastructure, and widespread acceptance. If implemented successfully, such projects could redefine the landscape of AI interaction for Filipinos, making technology more inclusive and relevant to everyday life rather than an exclusive advantage for the digitally savvy.

References
1. iTANONG-DS : A Collection of Benchmark Datasets for Downstream.
2. Natural Language Processing Tasks on Select Philippine Languages.
3. DOST-ASTI teams up with NU, BU to bring iTANONG to Filipinos.
4. The Philippines is creating a ChatGPT rival that speaks Filipino and Taglish.
5. Solon wants public service advisories in Filipino, dialects | Philippine News Agency.
6. 19th Congress – Senate Bill No. 680.
7. Advancing Language-Driven AI Solutions in the Philippines | OpenGov Asia.

FILIPINO

Pagtulay sa aksesibilidad ng wika sa pamamagitan ng AI

Ang sigasig ng mga Pilipino para sa artificial intelligence (AI) ay kapansin-pansin. Ayon sa isang pag-aaral ng Linkee noong unang bahagi ng 2024, ang Pilipinas ay ikaapat sa pinakamahilig sa AI sa buong mundo, batay sa dami ng buwanang paghahanap ng AI tools. Gayunpaman, nananatili pa rin ang isang malaking digital na agwat, kung saan higit sa 30 milyong Pilipino at walang access sa internet. Ang mga pangunahing dulot nito ay ang mahinang imprastrakturang digital, mataas na gastos, at kakulangan ng mga serbisyo sa internet.

Noong 2017, bago pa mainlunsad ang ChatGPT, nagsimula ang isang proyekto na nakatuon sa natural language processing para sa dalawang pangunahing wika sa Pilipinas. Pagsapit ng 2022, nagsimula itong maging isang tatlong taong proyekto na tinawag na “iTanong”. Ang ibig sabihin ng pangalan na ito ay “magtanong” sa Filipino at naaayon din dahil may mga titik I at T mula sa information technology. 

Ang iTanong ay naiiba sa ibang AI language models dahil pinapayagan nitong magtanong ang mga Pilipino gamit ang sariling wika. Gayunpaman, sa unang yugto ng paglulunsad nito, magagamit muna ito sa Filipino, Ingles, at Taglish.

Isang proyekto sa ilalim ng Advanced Science and Technology Institute (ASTI), layunin nitong maging isang mahalagang instrumento sa bansa kung saan ang Ingles ang pangunahing wika sa mga opisyal na komunikasyon. Malawakang ginagamit ang Ingles sa mga opisyal na website ng gobyerno, pulong, ulat, materyales sa edukasyon, at iba pang midya. Subalit, ito ay taliwas sa isinasaad ng 1987 Konstitusyon na ang gobyerno ay dapat aktibong hikayatin at suportahan ang paggamit ng Filipino bilang opisyal na wika ng komunikasyon. 

Aksesibilidad sa Impormasyon

Kamakailan lamang, nagsimulang gumawa ng hakbang ang gobyerno ng Pilipinas upang bigyang-diin ang kahalagahan ng paggamit ng Filipino, na kinikilala bilang isang wikang mas inklusibo at madaling maunawaan ng lahat. Isang mahalagang hakbang sa inisyatibang ito ay ang panukalang batas na nag-uutos sa mga ahensya ng gobyerno at mga lokal na yunit ng pamahalaan (LGU) na isalin sa Filipino at iba’t iba lokal na wika ang mga anunsyo ng pampublikong serbisyo.

Sa ilalim ng Panukalang Batas sa Sendado Bilang 680 o Language Accessibility of Public Information on Disasters Act, binibigyang-diin na mas epektibo ang pagpapakalat ng impormasyon gamit ang mga lokal na wika, lalo na sa panahon ng sakuna, matinding lagay ng panahon, programang panlipunan, suporta sa emerhensya, at iba pang usapin ng seguridad at kalusugan. 

Benepisyo ng AI na Lokal ang Pokus 

Bagaman karamihan sa mga Pilipino ay marunong magbasa at magsalita sa Ingles, 55% lamang ng mga nasa hustong gulang ang may kasanayan sa pagsasalita nito. Dahil dito, ang pokus ng iTanong sa lokal na wika ay nagbibigay ng mas angkop at epektibong solusyon para sa mga Pilipinong gumagamit nito. 

Ayon sa website ng ASTI, pangunahing layunin ng proyekto ang pagpapahusay ng akses sa impormasyon sa pamamagitan ng mga sumusunod:
• Pagbuo ng isang natural language querying engine (NLQE) na kayang makipag-ugnayan sa anumang database upang makakuha ng mahahalagang impormasyon
• Pagsasagawa ng makabagong pananaliksik sa machine learning, deep learning, data science, natural language processing, at artificial intelligence 
• Pagsasagawa ng pagsasanay at capacity-building para sa mga mananaliksik at gumagamit ng teknolohiya 
• Pagtutulak sa malawakang paggamit ng teknolohiya sa iba pang ahensya ng gobyerno
• Pagkokomersyalisa ng mga produkto at teknolohiyang nagmula sa pananaliksik 

Bagaman malaki ang potensyal ng sistemang ito, uunahin munang ipatupad ito sa mga tanggapan ng gobyerno bilang kapalit ng Citizen’s Charter bago ito gawing available sa publiko. Ang Citizen’s Charter ay isang dokumento na tumutulong sa mga opisina ng gobyerno na ipaalam ang kanilang mga pamantayan ng serbisyo sa mamamayan. Sa sandaling ganap nang maipatupad ang iTanong, magiging mas mabilis at madali paggawa ng mga pagbabago ng impormasyon.

Pagbabago sa Digital na Agwat

Maraming Pilipino ang sabik na yakapin ang teknolohiyang ito, ngunit ang mga isyu sa wika at imprastrakturang digital ay humahadlang sa malawakang paggamit nito. Kaya nananatiling isang hamon ang aksesibilidad. 

Dito pumapasok ang mga inisyatiba tulag ng iTanong – isang AI-driven na sistema na gumagamit ng mga lokal na wika upang makapagbigay ng impormasyon sa mas maraming mamamayan. Gayunpaman, hindi sapat ang makabagong teknolohiya upang makamit ang tunay na pag-unlad. Kinakailangan ang suporta ng gobyerno, pagpapabuti ng imprastrakturang digital, at pagtanggap ng mas nakararaming Pilipino. 

Kung maisasakatuparan nang maayos, ang mga proyektong tulad ng iTanong ay maaaring magbago ng paraan ng pakikipag-ugnayan ng mga Pilipino sa AI – ginagawang mas inklusibo at mahalaga ito sa pang-araw-araw na buhay, sa halip na maging isang eksklusibong instrumento lamang sa mga may kakayahang digital. 

ENGLISH

Tracking the Evolution of Filipino
Words in the Digital Age

Language constantly evolves, adapting to new ideas, cultures, and technologies. Humans instinctively create new words, repurpose old ones, and shape meanings to fit their ever-changing world. A prime example is the Filipino word “gigil”, which describes the trembling sensation of overwhelming emotion — unique to the language and untranslatable in a single English word. Meanwhile, polysemy, the ability of words to carry multiple meanings, is a universal linguistic phenomenon. Consider the English word salvage — while it traditionally means “to rescue,” in the Philippines, it has taken on a darker meaning: “to harm or kill.”These shifts reflect historical, social, and cultural changes embedded in language. Understanding word senses is crucial for Natural Language Processing (NLP) systems. Accurately identifying a word’s intended meaning improves the performance of applications such as sentiment analysis, question-answering, and stance detection. However, for Filipino and the Philippine variety of English, existing lexicons, such as the Filipino WordNet, primarily document words’ basic definitions, morphology, and synonyms. While this lexicon contains over 14,000 words, it lacks coverage of evolving meanings and context driven shifts, making it difficult to track the dynamic nature of language.

Computational Approaches to Word Evolution

Two computational methods have proven useful in studying language change: word embeddings and word co-occurrence networks. These techniques analyze how words are used in different contexts over time, capturing historical shifts in meaning. Studies show that new meanings emerge to reduce cognitive effort and that dominant senses stabilize through social interactions. However, no single meaning remains static: words develop new, often unrelated meanings influenced by social groups, digital media, and evolving discourse. Language change is evident in various domains, such as armed conflicts, socioeconomic class markers, and shifting stereotypes. In the digital age, polysemy is even more pronounced on social media, where words gain new meanings within specific online communities. For example, on Philippine Twitter, kalat — which traditionally means “mess” — is now used to refer to personal or controversial revelations. These rapid semantic shifts present challenges for NLP models, which must adapt to stay relevant.

Aiming for a Smarter, Context-Aware Filipino Lexicon

This research project seeks to bridge this gap by:
1. Developing and training diachronic (historical) representations of Filipino words and senses using word embeddings and temporal multiplex networks.
2. Analyzing the emergence of polysemy and semantic shifts across digital platforms and social contexts.
3. Building a continuously updated Filipino WordNet that captures these evolving meanings.

By investigating how Filipino words change over time — especially in the fast-paced digital landscape — this project will provide valuable insights into linguistic adaptation. The resulting lexicon will offer richer semantic information, allowing NLP models to interpret language more accurately.

Enhancing NLP for the Filipino Language

A key challenge in building effective NLP models for Philippine languages is the limited availability of training data. Additionally, current models struggle to capture subtle semantic nuances and emerging word meanings. Traditional models experience performance drops over time due to linguistic drift-where words acquire new meanings faster than models can be retrained.
By integrating context-aware temporal word representations, this research aims to enhance NLP capabilities in sentiment analysis, language identification, topic detection, and machine translation. A newly developed Filipino WordNet, enriched with evolving meanings, will improve model accuracy, helping bridge the gap between computational linguistics research and real-world applications. With better handling of semantic drift and improved model accuracy, this research will contribute significantly to advancing AI-driven language technologies for Filipino and Philippine English. As language continues to evolve, building adaptable and intelligent NLP systems ensures that our digital tools remain as dynamic as the people who use them.

References
1. Senti AI.

FILIPINO

Paano Nagbabago ang mga Salitang Filipino sa Internet 

Ang ating wika ay laging nagbabago. Sumusunod ito sa mga bagong ideya, kultura, at  teknolohiya. Tayong mga tao ay likas na gumagawa ng bagong mga salita, gumagamit muli ng  mga lumang salita, at nagbibigay ng bagong kahulugan para umangkop sa mundo natin.  Halimbawa, ang salitang “gigil” ay naglalarawan ng matinding emosyon na nanginginig ka — ito  ay kakaiba sa wikang Filipino at walang direktang katumbas sa Ingles. Samantala, ang isang  salita na maraming kahulugan ay pangkaraniwan sa lahat ng wika. Tulad ng salitang Ingles na  “salvage” — ang ibig sabihin nito ay “iligtas,” pero sa Pilipinas, ang ibig sabihin nito ay  “patayin.” Ipinapakita ng mga pagbabagong ito ang mga nangyari sa kasaysayan, lipunan, at  kultura. 
Mahalaga na maintindihan natin ang iba’t ibang kahulugan ng mga salita para sa Natural  Language Processing (o NLP). Kapag alam ng computer ang tamang kahulugan ng salita, mas  magiging maayos ang mga application tulad ng sentiment analysis, question-answering at stance  detection. Pero sa Filipino at sa Ingles na ginagamit sa Pilipinas, ang mga lexicon ngayon ay  karaniwang naglalaman lang ng mga pangunahing kahulugan, istruktura, at kasingkahulugan ng  mga salita. Kahit na may higit sa 14,000 salita ang leksikon na tinatawag na Filipino WordNet,  kulang ito sa mga bagong kahulugan na lumalabas, at sa mga kahulugan na nagbabago depende  sa konteksto. Kaya mahirap masundan ang mabilis na pagbabago ng ating wika. 

Paano Natutukoy ng Computer ang Pagbabago ng Salita 

May dalawang paraan na ginagamit ang computer para pag-aralan ang pagbabago ng wika: word  embeddings at word co-occurrence networks. Sinusuri ng mga ito kung paano ginagamit ang  mga salita sa iba’t ibang sitwasyon sa paglipas ng panahon, para malaman ang mga pagbabago sa  kahulugan. Ipinapakita ng mga pag-aaral na nagkakaroon ng bagong kahulugan ang mga salita  para mas madali itong maintindihan at nagiging matatag ang mga madalas gamiting kahulugan  dahil sa pakikisalamuha natin sa iba. Pero walang iisang kahulugan ang nananatiling pareho— nagkakaroon ng bagong kahulugan ang mga salita dahil sa impluwensya ng iba’t ibang grupo ng  tao, social media, at mga usapan.
Makikita ang pagbabago ng wika sa iba’t ibang bagay, tulad ng giyera, mga simbolo ng estado sa  buhay, at mga nagbabagong pananaw. Sa panahon ng internet, mas madalas na nagkakaroon ng  maraming kahulugan ang mga salita sa social media, kung saan nagkakaroon ng bagong  kahulugan ang mga salita sa loob ng mga grupo online. Halimbawa, sa Twitter, ang salitang  “kalat”—na ang ibig sabihin ay “gulo”—ay ginagamit na ngayon para tumukoy sa mga personal  o kontrobersyal na rebelasyon. Ang mabilis na pagbabagong ito ay nagdudulot ng problema sa  mga computer na tumutulong sa atin na maunawaan ang wika, dahil kailangan nilang umangkop  para manatiling updated.

Pagbuo ng Mas Matalinong Filipino Leksikon 

Layunin ng proyektong pananaliksik na ito na punan ang pagkukulang na ito sa pamamagitan ng: 
1. Pagbuo ng mga paraan para maipakita kung paano nagbago ang mga salitang Filipino sa  paglipas ng panahon, gamit ang word embeddings at temporal multiplex networks.  
2. Pagsusuri kung paano nagkakaroon ng maraming kahulugan ang mga salita at kung  paano nagbabago ang mga kahulugan nito sa internet at sa iba’t ibang sitwasyon.  
3. Pagbuo ng Filipino Wordnet na patuloy na ina-update para maisama ang mga bagong  kahulugan. 

Sa pamamagitan ng pag-aaral kung paano nagbabago ang mga salitang Filipino—lalo na sa  internet—ang proyektong ito ay magbibigay ng mahalagang impormasyon tungkol sa pag angkop ng wika. Ang bagong leksikon ay magbibigay ng mas maraming impormasyon tungkol  sa kahulugan ng mga salita, para mas maintindihan ng mga computer ang Filipino.

Pagpapabuti ng NLP para ng Wikang Filipino 

Ang isang malaking problema sa pagbuo ng mga NLP models na nakakaintindi ng wikang  Filipino ay ang limitadong dami ng data na kailangan para matuto ang mga ito. Bukod pa rito,  nahihirapan ang mga modelo na maintindihan ang mga maliliit na pagkakaiba sa kahulugan ng  mga salita at ang mga bagong kahulugan na lumalabas. Ang mga lumang modelo ay hindi na  masyadong maayos ang performance dahil mas mabilis ang pagkakaroon ng bagong kahulugan  ang mga salita kaysa sa pag-update ng mga modelo.  Sa pamamagitan ng paggamit context-aware word representations, layunin ng pananaliksik na ito  na mapagaling ang NLP capabilities sa sentiment analysis, language identification, topic  detection, and machine translation. Ang bagong Filipino Wordnet, na may kasamang mga  bagong kahulugan, ay magpapabuti sa accuracy ng mga models, para mas magamit ang mga ito  sa totoong buhay.
Sa mas maayos na pag-handle ng pagbabago ng kahulugan ng mga salita (semantic drift) at mas  mataas na accuracy, ang pananaliksik na ito ay makakatulong nang malaki sa pagpapabuti ng  mga teknolohiyang AI para sa Filipino at Ingles na ginagamit sa Pilipinas. Habang patuloy na  nagbabago ang wika, ang pagbuo ng mga NLP systems na matalino at kayang umangkop ay  nagsisiguro na ang ating mga digital tools ay kasing-dynamic ng mga taong gumagamit nito. 

ENGLISH

Implementation of AI in Poland and its Linguistic Aspects – Prompt Overview

Poland is a perfect example of openness to the arrival of new technologies and consciously using them in a responsible and people oriented way. The topic of AI has become a fixture in government activities and research centers, in implemented and planned programs, projects, and visions – as well as in its linguistic aspects.

In Polish operational steps, AI is seen primarily as a tool in the service of people – in public administration, initiatives aimed at increasing accessibility for people with disabilities, the development of small and medium-sized enterprises, education, and cybersecurity.
One of the major aspects discussed is the issue of language – the role of AI in communication, access to information, learning paths, or building native large language models (LLM), but also avoiding disinformation and confusion, for instance by counteracting fake news.
There are already several entities operating in Poland that oversee the development and implementation of AI. The Artificial Intelligence Portal¹ is developing dynamically, and the Working Group on Artificial Intelligence (GRAI)² and the SoDA AI Research Group (SAI)³ are actively functioning. The fruit of the latter’s work is, among others, the “Human Cooperation with AI: Perspectives for the Polish Public Sector”⁴ report containing examples of the implementation of AI-based systems in Poland and around the world.

Polish Language Model – PLLuM

A great success story in the field of AI implementation in Poland is the creation of – by far – the biggest Polish open language model, PLLuM (Polish Large Language Universal Model),⁵ based on artificial intelligence technology.
This year, PLLuM is supposed to power the government application mObywatel.⁶ PLLuM is the result of cooperation between leading Polish scientific institutions, research centers,
and experts from various fields. An important element of its creation was the development of an extensive and diverse data set that reflects the complexity of the Polish language – taking into account its social, cultural, and historical context, along with its nuances, idioms, and phraseologies. PLLuM – to put it simply – thinks in Polish.
The project collected the highest quality text data, including works not publicly available – about 150 billion tokens – in Polish, but also in other Slavic and Baltic languages and in English, in order to serve a better generalization of models.⁷ Thanks to this, the largest Polish- language collection of organic prompts and manually prepared answers were developed, as well as a corpus of preferences and original evaluation sets. In addition, the collection was created using a high-quality Polish language corpus and lexical resources, as well as synthetic instructions created using large language models.⁸ A whole family of language models was created, ranging in size from 7 to 70 billion parameters, adapted to various tasks in Polish.⁹ Specialized Retrieval Augmented Generation models were also created, which are to help build intelligent assistants for Polish public administration.
Other Polish language models worth mentioning include Bielik or Qra.

AI in the Service of Accessibility and Inclusiveness

Another program being implemented in Poland is FERS (European Funds for Social Development 2021-2027), guaranteeing support and actions to increase accessibility for people with special needs, improving the quality of education and healthcare, increasing their chances in the labor market, as well as promoting social integration and development of competencies. One of the affiliated projects, whose the common denominator with FERS is increasing digital accessibility, aims to support active social inclusion to promote equal opportunities, non-discrimination, and active participation in particular groups with special needs and disadvantaged backgrounds to increase their employability.¹⁰ In addition, the action plan for the project “AI DC – Artificial Intelligence Supporting Accessibility” was approved under Measure 03.02 Digital accessibility in Priority 3. Accessibility and services for people with disabilities.¹¹
Its main goal is to create an artificial intelligence-based virtual assistant that will serve as a tool for digital solutions and help creators identify and fix digital accessibility issues, including linguistic and communication errors, on websites and mobile applications. The assistant will not only detect errors but also suggest fixes, explain issues related to digital accessibility, suggest alt texts for graphics, and generate captions and subtitles for multimedia materials.
The user will be able to benefit from this artificial intelligence solution by talking in a simple, straightforward manner to a virtual assistant whose operation will be based on a principle of an advanced chatbot. It will draw attention to detected errors or other issues and doubts related to digital accessibility.

Ethical Approach to AI Development
and Implementation

Poland is one of the most active countries in developing ethical use of data according to the concept of Trustworthy AI. The Polish government has developed a dynamic ethical code of AI, establishing a “Virtual Institute of AI research and experts, taking into account art. 30 of the Constitution of the Republic of Poland and the Charter of Fundamental Rights of the EU, promoting human-centric AI and its reliability (e.g. by counteracting errors such as algorithmic bias), transparency and explainability.”¹²
It should be acknowledged that activities related to language and AI also include systematic prevention of disinformation and fake news. Poland places great emphasis on cybersecurity and monitoring undesirable activities in this area. Efforts are also carried out to monitor large language models and prevent hallucinations in the case of chatbots and GenAI. The aim is to create safe and responsible conditions for introducing AI in the public sector and state administration bodies.¹³

Will We Find a Common Language with AI?

The power of language transcends all limitations, including technological ones. Joint actions by scientists, government and society – all properly educated and prepared to cooperate with AI – can help create a state without barriers. Although the development of technology carries with it obvious threats, joint efforts can create a safe space for its use. Language models can be used to create helpful assistants, expand educational opportunities and lead to the creation of new career paths. The actions currently being observed in Poland are slowly but surely heading in the right direction.

References
1. Portal sztucznej inteligencji | Gov.pl.
2. Grupa robocza ds. sztucznej inteligencji (GRAI) – Portal sztucznej inteligencji | Portal Gov.pl.
3. About Us | SoDA.
4. Raport SoDA AI Research Group: Współpraca Człowieka z AI Perspektywy dla Polskiego Sektora Publicznego.
5. PLLuM.
6. PLLuM – A tool that teaches machines to think… in Polish.
7. mObywatel 2.0.
8. Zakończenie projektu PLLuM.
9. Fundusze Europejskie dla Rozwoju Społecznego – Centrum Projektów Polska Cyfrowa | Portal Gov.pl.
10. Działanie 03.02. Dostępność cyfrowa – Centrum Projektów Polska Cyfrowa | Portal Gov.pl.
11. AI DC – sztuczna inteligencja wspierająca dostępność cyfrową – Centrum Projektów Polska Cyfrowa | Portal Gov.pl.
12. Polityka dla rozwoju sztucznej inteligencji w Polsce od roku 2020 | Gov.pl.
13. Sztuczna inteligencja w służbie obywateli i administracji rządowej – Portal sztucznej inteligencji | Gov.pl.

POLISH

Wdrażanie AI w Polsce i aspekty językowe tych procesów – wybrane działania

Polska jest doskonałym przykładem otwartości na nadejście nowych  technologii oraz świadomego korzystania z nich w sposób odpowiedzialny  i zorientowany na ludzi. Temat AI na stałe zagościł w rządowych działaniach  i ośrodkach badawczych, w realizowanych i planowanych programach,  projektach i wizjach – również w kontekście językowym. 

W polskich praktykach AI postrzegane jest przede wszystkim jako narzędzie  w służbie ludzi – w administracji publicznej, inicjatywach mających na celu  zwiększanie dostępności dla osób z niepełnosprawnościami, rozwoju małych  i średnich przedsiębiorstw, w edukacji, a także w problematyce cyberbezpieczeństwa. Jednym z istotnych poruszanych aspektów jest kwestia  językowa – rola AI w komunikacji, dostępności do informacji, ścieżkach nauki czy  budowania natywnych dużych modeli językowych (LLM), jak również unikanie  dezinformacji i dezorientacji, na przykład poprzez przeciwdziałanie fake newsom.

W Polsce istnieją dedykowane podmioty i ośrodki badawcze obejmujące pieczę nad  rozwojem i wdrażaniem AI. Prężnie rozwija się m. in. Portal sztucznej inteligencji, aktywnie działają Grupa Robocza ds. Sztucznej Inteligencji (GRAI) i SoDA AI Research Group (SAI) . Owocem pracy tej ostatniej jest między innymi raport „Współpraca Człowieka z AI: Perspektywy dla Polskiego Sektora Publicznego” – w którym znaleźć można przykłady implementacji systemów opartych na AI w Polsce i na świecie. 

Polski model językowy – PLLuM  

Wielkim sukcesem w pracy nad wdrażaniem AI w Polsce było stworzenie  największego dotychczas polskiego otwartego modelu językowego PLLuM – Polish  Large Language Universal Model, opartego na technologii sztucznej inteligencji. Już w tym roku PLLuM ma zasilić rządową aplikację mObywatel. 

PLLuM to efekt współpracy wiodących polskich instytucji naukowych, ośrodków  badawczych i ekspertów z różnych dziedzin. Istotnym elementem jego powstania  było opracowanie obszernego i zróżnicowanego zbioru danych, który oddaje złożoność języka polskiego – zarówno w kontekście społecznym, jak i kulturowym  oraz historycznym, wraz z jego niuansami, idiomami i frazeologizmami. PLLuM –  w najprostszym ujęciu – myśli po polsku.  

W ramach projektu zgromadzono najwyższej jakości dane tekstowe w języku  polskim, w tym utwory niedostępne publicznie – ok 150 mld tokenów – ale także  w innych językach słowiańskich i bałtyckich oraz w języku angielskim, co ma służyć lepszemu uogólnieniu modeli. Dzięki temu opracowano największy polskojęzyczny zbiór organicznych promptów i ręcznie przygotowywanych odpowiedzi, a także  korpus preferencji i autorskie zbiory ewaluacyjne. Przy tworzeniu zbioru korzystano  ponadto z wysokiej jakości polskojęzycznych zasobów korpusowych i leksykalnych  oraz z instrukcji syntetycznych, tworzonych z użyciem LLM. Utworzona została cała rodzina modeli językowych w wielkości od 7 do 70 mld parametrów, dostosowanych  do różnorodnych zadań w języku polskim . Powstały również wyspecjalizowane modele Retrieval Augmented Generation, które mają pomóc w stworzeniu  inteligentnych asystentów dla polskiej administracji publicznej.  

Inne polskie modele językowe warte uwagi to m.in. Bielik i Qra. 

AI w służbie dostępności i inkluzywności 

Jednym z realizowanych w Polsce programów jest FERS – Fundusze Europejskie  dla Rozwoju Społecznego 2021-2027, gwarantujący wsparcie i działania na rzecz  zwiększenia osób ze szczególnymi potrzebami, podnoszenia jakości edukacji,  ochrony zdrowia, zwiększenia szans na rynku pracy, a także promujący integrację społeczną i rozwój kompetencji. Jeden z projektów, których wspólnym mianownikiem  jest zwiększenie dostępności cyfrowej ma na celu „wspieranie aktywnego włączenia  społecznego w celu promowania równości szans, niedyskryminacji i aktywnego  uczestnictwa oraz zwiększanie zdolności do zatrudnienia, w szczególności grup  o specjalnych potrzebach i w niekorzystnej sytuacji”.

Zatwierdzony został ponadto plan działania dla projektu „AI DC – sztuczna  inteligencja wspierająca dostępność” w ramach działania: 03.02 Dostępność cyfrowa  w priorytecie 3. Dostępność i usługi dla osób z niepełnosprawnościami. 

Jego głównym celem jest stworzenie wirtualnego asystenta opartego na sztucznej  inteligencji, który będzie służyć jako narzędzie w kwestii rozwiązań cyfrowych oraz  pomagać twórcom identyfikować i naprawiać błędy dostępności cyfrowej na stronach  internetowych oraz w aplikacjach mobilnych. Asystent ma nie tylko wykrywać błędy,  ale też sugerować rozwiązania niezbędne do ich naprawy, wyjaśniać kwestie  związane z dostępnością cyfrową, proponować alt teksty do grafik i generować podpisy oraz napisy do materiałów multimedialnych.  

Użytkownicy będą mogli skorzystać z pomocy sztucznej inteligencji poprzez  możliwość rozmowy (w nieskomplikowanym, zrozumiałym języku) z wirtualnym  asystentem, którego działanie będzie opierało się na zasadzie zaawansowanego  chatbota. Będzie on zwracał uwagę na wykryte błędy lub inne kwestie oraz  wątpliwości związane z dostępnością cyfrową. 

Etyczne podejście do rozwoju i wdrażania AI

Polska jest jednym z najbardziej aktywnych krajów w rozwijaniu etycznego  wykorzystywania danych według koncepcji Trustworthy AI. Polski rząd wypracował dynamiczny etyczny kodeks AI, powołał „Wirtualny Instytut badawczy i ekspertów AI,  uwzględniający art. 30 Konstytucji RP oraz Kartę Praw Podstawowych UE, promuje  human-centric AI oraz jej niezawodność (np. poprzez przeciwdziałanie błędom takim  jak algorithmic bias), przejrzystość i wyjaśnialność”.

Należy pamiętać, że działania związane z językiem i AI to również systematyczne  zapobieganie dezinformacji i fake newsom. Polska kładzie bardzo duży nacisk na  cyberbezpieczeństwo i monitorowanie niepożądanych praktyk w tym zakresie. Na  bieżąco prowadzone są również działania monitorujące modele LLM czy  zapobiegające halucynacjom w przypadku chatbotów i GenAI, mające na celu  stworzenie bezpiecznych i odpowiedzialnych warunków wprowadzania AI do sektora  publicznego oraz organów administracji państwowej.

Czy znajdziemy wspólny język ze sztuczną inteligencją? 

Moc języka pozwala na przekraczanie wszelkich barier, również tych  technologicznych. Praca i działania przedstawicieli świata nauki, rządu  

i społeczeństwa – odpowiednio przygotowanego i posiadającego kompetencje  pozwalające na współpracę ze sztuczną inteligencją – pozwolą na stworzenie  w przyszłości państwa bez barier. Choć rozwój technologii AI niesie ze sobą oczywiste zagrożenia, wspólne wysiłki mogą stworzyć bezpieczną przestrzeń do jej  wykorzystania. Modele językowe mogą być stosowane do tworzenia pomocnych  asystentów, rozszerzać możliwości edukacyjne i kształtować nowe ścieżki kariery.  Działania, jakie obecnie możemy obserwować w Polsce, krok po kroku zmierzają we  właściwym kierunku.