"A perfect score will require a breakthrough in AI, well beyond current capabilities."
AI vs. Animal
To test cutting-edge artificial intelligences, a team of researchers wants to see how well they handle tasks designed to measure the intelligence of animals. They’re calling their experiment the Animal-AI Olympics, and they’ve giving out more than $10,000 in prizes to high-achieving AIs.
This month, the group plans to release a virtual “playground” they’re developing for the competition and a list of the cognitive abilities they plan to test. At that point, researchers can begin training their AIs to navigate the playground.
In June, the event organizers will begin subjecting each competitor AI to 100 playground tasks they’ve never seen before, each with the same goal usually given to animals in similar intelligence-testing scenarios: retrieve food.
While the goal may be the same, the various obstacles the AIs will need to overcome to achieve success will vary — they might need to move an object, for example, or demonstrate an understanding of object permanence.
“We expect this to be a hard challenge,” Matthew Crosby, one of the researchers behind the Animal-AI Olympics, told New Scientist. “A perfect score will require a breakthrough in AI, well beyond current capabilities.”
“However,” he continued, “even small successes will show that it is possible, not just to find useful patterns in data, but to extrapolate from these to an understanding of how the world works.”
READ MORE: AIs go up against animals in an epic competition to test intelligence [New Scientist]