Inspiration

  • We were interested in assisting astronauts on an emotional scale, so we wanted to explore how AI could sense non-explicit cues to help understand what astronauts go through daily.
  • Being on their missions for over 400 days, they consistently experience isolation. We wanted to.
  • Robots such as the KIBO-RPC inspired us to think of ways to include our digital product in a physical setting on the spaceship.
  • We first went for a more serious route until we realized that most people appreciate an entertaining or light-hearted approach to consciousness.
  • So, we decided to work on an AI tool that creates scripts of high-emotion moments for astronauts on the spaceship, which can have research and recreational advancements.

Problems faced

  • Our biggest problem was ideation; we had great concepts, but honing in on something that solved a problem instead of just using a cool AI feature was hard.
  • It was initially difficult for the programmers (Cindy and Reo) to train their models to be accurate.

  • Audience: Producers, Mission Control: AKA those who want to analyze the behavior of astronauts on their mission.

  • Problem space: Astronauts inevitably deal with consistent stressors on their missions that last hundreds of days, which leads to burnout. Mission Control and the general public often do not understand the full extent of their daily trials and novelties.

How might we use AI to enable Mission Control and the general public to empathize and understand the reality of astronauts working on extensive missions?

Concept

  • Anti-gravity bot that follows the astronauts around to:
    • Track astronauts’ emotions visually with tensorflow AI computer vision to track significant high-emotion moments:
      • Sad
      • Angry
      • Disgust
      • Fear
      • Surprise
    • Use AI to transcribe conversations into text Whisper
    • Send this text to Mission Control to:
      • Help NASA recognize patterns/solutions to different scenarios in astronauts’ day-to-day life
      • Give producers real-world content for reality television

Technical Stuff/Building the Model

Built With

  • kaggle
  • openai
  • python
  • tensorflow
  • whisper
Share this project:

Updates