Sztuczna inteligencja w quizie dotyczącym reagowania na katastrofy

Odpowiedzi i wyjaśnienia pojawiają się w miarę rozwiązywania quizu.

1. What does Google's Flood Hub provide to its users?
   • A guarantee that every flood is predicted perfectly
   • Model-based flood forecasts and related location-specific information
   • Physical flood barriers constructed by software
   • Automatic confirmation of every social-media report
2. W jaki sposób systemy AI zazwyczaj mapują zniszczenia budynków po katastrofie?
   • Przeprowadzając wywiady z każdym mieszkańcem
   • Domyślając się z wiadomości
   • Licząc rozmowy telefoniczne
   • Poprzez wykrywanie zmian poprzez porównywanie obrazów przed i po
3. What modeling approach is central to the GraphCast weather-forecasting study?
   • A graph neural network trained for global weather forecasting
   • A rule that no weather observations are needed
   • A speech-only transcription model
   • A guarantee of identical performance on every hazard and machine
4. Dlaczego modele reagowania na katastrofy mogą borykać się z nowymi zdarzeniami?
   • Katastrofy są zbyt częste
   • Działają tylko latem
   • Modele przeszkolone na podstawie wydarzeń z przeszłości nie mogą uogólniać na nowe
   • Nie wymagają danych szkoleniowych
5. What disaster-information function did the AIDR project demonstrate for microblog messages?
   • Guaranteed automatic verification of every message's location and truth
   • Prediction of all future earthquakes from punctuation
   • Replacement of every responder's decision with a text label
   • Supervised classification of crisis-related messages using labeled examples
6. Why should a disaster-information workflow account for communication and power dependencies?
   • Every network is guaranteed to remain available during an event
   • Infrastructure disruptions can prevent the workflow from receiving or sharing needed data
   • Power supply never affects communication equipment
   • A model's confidence restores damaged infrastructure automatically
7. What should a crisis-message classification label establish by itself?
   • That the claimed need is definitely current and verified
   • That every reported location has been measured precisely
   • Only the model's classification, with the post's time, location, and factual support still needing assessment
   • That the message must be genuine because it contains a disaster keyword
8. Why is a difference between pre- and post-event satellite images not automatically proof of building damage?
   • Image timing, alignment, visibility, and other scene changes can also affect the comparison
   • Every changed pixel has exactly one possible cause
   • The image file extension establishes a physical damage class
   • No post-event image can contain any useful evidence