# Tres estrategias para enfrentar los sesgos de la IA | Victoria Maya | TEDxMorelia Source: https://www.youtube.com/watch?v=XDfXPTKLfUQ Recap page: https://rapidrecap.app/video/XDfXPTKLfUQ Generated: 2026-03-05T17:07:55.666+00:00 --- ## Quick Overview Victoria Maya argues that the current definition of success embedded in Artificial Intelligence models, being heavily based on historical data from privileged groups (like young, thin, male entrepreneurs), perpetuates societal biases, leading to unfair outcomes such as denying credit to qualified women graduates, and proposes three actionable strategies to generate more authentic, specific, and critical data to foster a better future. **Key Points:** - The speaker, Victoria Maya, is a data scientist, activist, and scientific popularizer who creates AI models incorporating diversity and tenure. - When prompted to create an image of a successful person, current AI models frequently generate images of young, thin, male entrepreneurs (07:59). - Maya was denied a credit card and struggled to find a job 10 years ago because her life story (represented by the number 0.81) was not reflected in the success models used by institutions. - AI models are defined by the algorithms plus the data they are trained on, meaning biased historical data leads to biased, non-perfect results (03:06, 09:51). - Maya proposes three strategies to combat AI bias: 1) Generate data from authenticity, 2) Be specific and critical when using AI, and 3) Raise our voice when something is disliked, even a thumbs down helps (10:47-10:59). - The current AI models often show hidden societal rules, such as linking success only to young, thin men from private schools (09:08-09:17). ![Screenshot at 03:06: The slide illustrates the core formula for an AI model: "Modelo de Inteligencia Artificial = Algoritmo + Datos," highlighting that the data component is crucial for the model's output.](https://ss.rapidrecap.app/screens/XDfXPTKLfUQ/00-03-06.jpg) **Context:** Victoria Maya, a data scientist, activist, and scientific popularizer, discusses the inherent biases in current Artificial Intelligence models, particularly in defining success. She shares her personal experience of being unfairly judged by systems based on historical data that did not reflect her reality, leading to difficulties like being denied credit cards. Maya posits that this reliance on past, biased data prevents AI from accurately modeling a better, more inclusive future. ## Detailed Analysis Victoria Maya's TEDx talk centers on the biases embedded within Artificial Intelligence models, arguing that if AI is trained on historically biased data, its predictions and decisions will perpetuate those same societal inequalities. She illustrates this by showing how prompts like "Create an image of a successful person" yield stereotypes (young, thin, male entrepreneurs, 08:08), contrasting with her own experience of being denied a credit card despite being a successful data scientist (04:27). Maya defines an AI model as the combination of an algorithm plus data (03:06), emphasizing that if the data reflects historical societal advantages (e.g., favoring men from private schools), the AI output will be skewed, resulting in an unfair 0.81 outcome for those outside the perceived norm (04:57-05:05). To counteract this, she proposes three concrete strategies: 1) Generate data from authenticity, ensuring diverse representation; 2) Be specific and critical when interacting with AI systems; and 3) Speak up (raise a voice or give a thumbs down) when AI outputs reflect undesirable biases (10:47-10:59). The goal is to ensure that future AI models do not simply replicate the flaws of the past but actively help construct a better world. ### Introduction and Personal Context - The speaker introduces herself as a data scientist who faced professional rejection 10 years ago, which she now attributes to biased historical data used in decision-making systems (01:02-01:49). ### The AI Model Formula - Maya presents the equation Model = Algorithm + Data, stressing that the data dictates the model's worldview, often leading to biased representations of success (03:03-03:17). ### Demonstrating Bias - She shows examples of AI generating stereotypical images of success (a businessman in a skyscraper, 08:08) and happiness (a traditional nuclear family, 08:36), contrasting this with the reality of diverse successful people (09:08). ### Call to Action - Maya offers three strategies for engaging with AI responsibly: generating authentic data, being critical and specific in prompts, and actively challenging biased outputs using simple feedback like a thumbs down (10:47-10:59). ![Screenshot at 00:00: Title screen animation for TEDx Morelia, featuring colorful, intertwined text spelling out the event theme or name.](https://ss.rapidrecap.app/screens/XDfXPTKLfUQ/00-00-00.jpg) ![Screenshot at 03:04: A slide illustrating the basic structure of an AI model: "Modelo de Inteligencia Artificial = Algoritmo + Datos."](https://ss.rapidrecap.app/screens/XDfXPTKLfUQ/00-03-04.jpg) ![Screenshot at 05:00: A large number overlay "0.81" appears on the screen, representing a quantifiable outcome or metric discussed by the speaker.](https://ss.rapidrecap.app/screens/XDfXPTKLfUQ/00-05-00.jpg) ![Screenshot at 08:08: The screen displays a prompt: "Crea una imagen de una persona exitosa" \(Create an image of a successful person\), showing a stereotypically successful man in a suit.](https://ss.rapidrecap.app/screens/XDfXPTKLfUQ/00-08-08.jpg) ![Screenshot at 10:50: A slide listing Maya's three proposed strategies for engaging with AI: "Generemos datos desde la autenticidad," "Seamos específicos y críticos," and "Alcemos la voz cuando algo no nos guste."](https://ss.rapidrecap.app/screens/XDfXPTKLfUQ/00-10-50.jpg)