The Center for AI Safety and Scale AI are crowdsourcing submissions for 'Humanity's Last Exam,' the hardest test designed to challenge advanced AI systems.
CAIS executive director Dan Hendrycks stated that the o1 model has 'destroyed the most popular reasoning benchmarks,' prompting the need for this challenging exam.
The new exam, 'Humanity's Last Exam,' aims to include abstract reasoning questions, furthering the challenge for AI systems which have surpassed traditional tests.
Experts submitting questions for the exam, due November 1, will have a chance to co-author a related paper and win prizes up to $5,000.
Collection
[
|
...
]