Reading

**A People’s Guide to AI by Mimi Onuoha and Mother Cyborg(Diana Nucera)**

• Many human intelligences(Naturalist, Musical, Interpersonal, Logical-Mathematical, etc) Which are being prioritized in AI creation?
• AI can be incorporated almost anywhere, like adding salt to food
• There can be different algorithms for accomplishing the same task, which are judged by their efficiency
• AI vs. Machine Learning
  • ML is a branch of AI
  • AI includes tasks we normally think only humans could
  • ML also includes tasks that humans might not be able to do
  • “In AI, humans set the terms for the algorithm that a computer will use, while in machine learning the computers construct the algorithms”… ML uses pattern recognition
• Deep Learning
  • Subfield on ML
  • Requires more data so programs can make more connections + do more complex operations
  • Ability of computers to improve on their own, get better at tasks over time
• Applications of AI in social spaces can be dangerous, affect certain communities disproportionately
• How do we prepare for a future completely opposite of our current reality?

Read Excavating AI and explore ImageNet, ImageNet sample images, Kaggle ImageNet Mini 1000.

• AI systems inherit biases from image datasets used for training, often reflecting social stereotypes which AI perpetuates
• Image categorization + labelling processed lack transparency + ethical oversight
• “Everything is flattened out and pinned to a label, like taxidermy butterflies in a display case. The results can be problematic, illogical, and cruel, especially when it comes to labels applied to people.”
• ImageNet- widely used training set for ML
• Repeats of the same image at different scales?
• Are there efforts towards cleaning up like datasets ImageNet?

Try playing Emoji Scavenger Hunt.

• Asked to find a TV + sunglasses, pointed at blank wall and was “found”