Was KI über uns verrät | Simone Scholz | TEDxKollerschlag

Quick Overview

The speaker, Simone Scholz, argues that Artificial Intelligence (AI) systems currently reflect and amplify existing human biases and societal flaws present in their training data, emphasizing that "Superman is not coming" to fix these issues, and the responsibility lies with humans to train AI ethically to avoid perpetuating discrimination, racism, and sexism.

Key Points: Simone Scholz, originally from Carinthia, moved to Vienna at age 10, where she pursued ballet, an activity she loved, despite initially being labeled a 'worker's child' (Arbeiterkind). The speaker highlights that AI systems, like the famous chatbot that emerged in 2016, are trained on massive datasets, including 550 billion social media entries, which contain inherent human biases. AI systems, such as the Microsoft chatbot, learn to imitate human behavior, including discriminatory, sexist, and violent patterns found in the data, rather than being inherently malicious. The speaker points out that the responsibility for fair AI lies with humans who curate the data, contrasting the massive data fed to AI with the small amount of data children learn from. The quote, 'Superman is not coming' (attributed to Eric Brechner), underscores that no external savior will fix biased AI; humans must actively work to create ethical models. Scholz concludes that we must proactively reflect on and question our own behaviors and biases when training AI, rather than expecting the AI itself to resolve inherent societal unfairness.

Context: This TEDx talk, titled 'Was KI über uns verrät' (What AI reveals about us), features speaker Simone Scholz discussing the impact of training data on Artificial Intelligence (AI) behavior. Scholz draws from her personal background, mentioning her childhood as a 'worker's child' and her passion for ballet, to frame the discussion around how societal biases are encoded and reflected in the increasingly powerful AI systems being developed today.

Detailed Analysis

Simone Scholz delivers a presentation arguing that Artificial Intelligence (AI) is inherently a mirror of humanity, reflecting the data it is fed. She shares a personal anecdote about growing up as an 'Arbeiterkind' (worker's child) and how her passion for ballet was often at odds with societal expectations, illustrating the concept of ingrained bias. She states that AI training relies heavily on massive datasets, such as 550 billion social media entries, which contain discriminatory, sexist, and violent language and behavior. Consequently, when trained on this data, AI models—like the early Microsoft chatbot—mimic these negative patterns, producing biased or harmful outputs. Scholz stresses that the solution is not external intervention, quoting Eric Brechner: 'Superman is not coming.' The onus is on humans, especially data scientists and AI trainers, to consciously interrogate and filter the training data to ensure fairness and avoid automating existing societal prejudices. She concludes by asserting that we must actively teach AI positive role models and ethical behavior, rather than expecting the technology to solve human failings on its own.