Testing Autonomous AI-Based Systems CT-AI Practice Questions (Page 3)

Questions 11–15

1. 11expert · hard

   A team is testing an autonomous drone delivery system. The drone's flight controller uses a neural network for obstacle avoidance. The team has two testing goals: (1) prove that the drone will never enter a designated no-fly zone under any circumstances, and (2) verify that the drone can navigate a variety of real-world environments. The neural network is non-deterministic due to floating-point variations across hardware. Which combination of testing approaches best addresses both goals?

   • AUse extensive simulation testing for both goals, running millions of random scenarios to empirically demonstrate that the no-fly zone is never entered.
   • BUse formal verification to prove the no-fly-zone property on the neural network, and use simulation-based testing to evaluate navigation performance across diverse environments.
   • CUse runtime monitoring during field trials to detect no-fly-zone violations and rely on simulation only for navigation performance.
   • DUse real-world field testing for both goals, flying the drone in many locations and observing whether it ever enters a no-fly zone.
   Reveal Solution Discussion Select an answer first
2. 12foundation · easy

   In the context of autonomous systems, what is the primary difference between verification and validation?

   • AVerification checks that the system meets its specified requirements, while validation checks that the system meets user needs in the real world
   • BVerification is only for software, while validation is for hardware
   • CVerification is performed by developers, while validation is performed by testers
   • DVerification is done after deployment, while validation is done before deployment
   Reveal Solution Discussion Select an answer first
4. 13foundation · easy

   Which ethical consideration is particularly relevant when testing autonomous AI systems?

   • AEnsuring the system does not exhibit bias against certain groups of people
   • BEnsuring the system is fast enough for real-time operation
   • CEnsuring the system is compatible with existing hardware
   • DEnsuring the system is easy to use for operators
   Reveal Solution Discussion Select an answer first
5. 14application · medium

   An autonomous forklift operates in a warehouse at SAE Level 4 autonomy: it can handle all driving tasks within a defined area without human intervention, but it is restricted to that area. The forklift's path-planning algorithm is being updated. The team must perform a risk assessment before deploying the updated algorithm. Which factor is most critical to include in the risk assessment?

   • AThe forklift's autonomy level, since higher autonomy levels inherently have lower risk.
   • BThe cost of the forklift and the potential financial loss if it is damaged during operation.
   • CThe severity of potential harm to human workers if the forklift malfunctions, combined with the likelihood of such malfunctions.
   • DThe probability that the forklift will encounter a human worker in its defined operational area.
   Reveal Solution Discussion Select an answer first
6. 15foundation · easy

   In the context of autonomous AI-based systems, what is the primary characteristic that defines the level of autonomy of a system?

   • AThe degree of human oversight and intervention required during operation
   • BThe speed at which the system can process data
   • CThe sophistication of the machine learning algorithms used
   • DThe number of sensors and actuators used by the system
   Reveal Solution Discussion Select an answer first