Kuis Kecerdasan Umum Buatan

Jawaban dan penjelasan muncul saat Anda mengikuti kuis.

1. What distinguishes artificial general intelligence from narrow AI?
   • Running on more expensive and powerful hardware
   • Broad ability to learn and perform across many domains
   • Having a physical robot body that moves in the world
   • Being trained exclusively on large image datasets
2. What does Moravec's paradox observe?
   • AI progress stops once it matches average human intelligence
   • Larger models always become less accurate over time
   • Perception and motor skills, easy for us, are hard for AI
   • Computers can never beat humans at strategy games
3. Which benchmark did François Chollet propose as a concrete way to measure skill-acquisition efficiency on abstract puzzles?
   • ARC (Abstraction and Reasoning Corpus)
   • MMLU (Massive Multitask Language Understanding)
   • GLUE (General Language Understanding Evaluation)
   • SQuAD (Stanford Question Answering Dataset)
4. What is the 'intelligence explosion' hypothesis associated with I. J. Good?
   • Standard intelligence tests will soon become obsolete
   • AI hardware will physically overheat and fail at scale
   • The number of AI startups will keep growing exponentially every year
   • A self-improving machine could trigger runaway capability gains
5. Rich Sutton's essay 'The Bitter Lesson' argues what about AI progress?
   • Symbolic rules crafted by experts will always outperform learning
   • General methods that scale with compute beat hand-built knowledge
   • Small curated datasets are always better than large ones
   • AI research should stop relying on GPUs and big clusters
6. Why is the Turing test considered an incomplete criterion for AGI?
   • No machine has ever actually attempted to pass it
   • It only tests a machine's mathematical ability
   • It tests conversational imitation, not broad competence
   • It requires a physical robot body in order to pass
7. Why does the guide say AGI discussions need an explicit definition?
   • Different proposed thresholds produce different claims about timing and capability
   • There is a single legal definition used worldwide
   • AGI means any system that generates text
   • A definition proves a system is safe
8. A model performs well on several benchmarks. What conclusion does the guide support?
   • It proves the system can perform every cognitive task
   • It settles the AGI definition debate
   • It shows the model has human experience
   • Benchmark results are evidence about tested tasks, not proof of universally agreed AGI