Gary Ackerman @garyackerman@qoto.org

If your hypothesis cannot be questioned—it is always correct despite the evidence—it cannot be called science.

"Fairness" in AI can be measured in different ways, such as ensuring similar outcomes for individuals with similar qualifications ("individual fairness") or ensuring groups have proportional outcomes ("group fairness"). #AIFairness

Eliminating variables like race, gender, or origin from data doesn't automatically remove bias, as discrimination can still surface through correlations with other factors. #BiasMitigation

A significant source of bias comes from skewed or incomplete data sets used to train AI algorithms. This can lead to skewed outcomes. #DataBias

One way AI models become biased is through confusing correlation with causation. Two correlated factors changing together don't necessarily mean one causes the other. #DataBias

AI has demonstrated improved performance on tasks like image recognition, chess playing, and medical analysis, but biases persist. #AIDecisionMaking

Bias is an inclination toward or outlook that is prejudiced. In the real world, bias closely relates to discrimination or treatment. #AIBias

Ultimately, the curriculum of the future is a result of actively engineering minds & mentalities to enact a preferred vision of society. It's shaped by complex interactions of economy, culture, expertise, & identity.

Learning is increasingly positioned as a lifelong, lifewide activity, harmonized across formal & informal spaces. It's becoming a lifestyle choice, often commodified, where consuming equals learning. #LifelongLearning #ConsumerCulture

Societies are increasingly understood through a "cybernetic style of thought" – saturated with metaphors like networks, flexibility, speed, virtuality. This style shapes how the curriculum is remade. #Cybernetics #EdTech

Curriculum is like a "saddleback," looking both to the past & promoting a vision for the future, expressing legacy, aspirations, & anxieties. It's far from neutral or nonpolitical. #CurriculumDesign #EducationPolicy

Humanizing pedagogies have historical roots in Paulo Freire's work, emphasizing critical consciousness, power, privilege, and ideology, not just methods.

The AI control problem is paramount: ensuring future advanced AI remains beneficial and aligned with human values, avoiding potential existential risks. #AIControl #ExistentialRisk

In accident scenarios, potential AI ethical programming includes Retributivist (harm responsible party), Selfish (protect AI's occupant), and Utilitarian (minimize overall harm). Each has implications. #AIethics #DecisionMaking

The concept of AI personhood is highly debated. Should AI systems have rights or responsibilities similar to humans or corporations?

The propagation of misinformation online, sometimes amplified by AI, is a serious issue impacting politics and societal trust. #Disinformation #AI

Companies leverage user-generated data to improve services and user experience, but this practice has led to concerns and regulations like GDPR. #DataPrivacy #UserData

As automation progresses, workers will likely need to acquire new skills to remain competitive. This shift could widen the gap in income inequality. #Upskilling #Workforce

Beyond just job displacement, the increasing capabilities of AI, especially in mimicking human intelligence, raise complex ethical questions.

AI has the potential to increase productivity, but its impact on economic inequality is a significant concern.