Abstract:Recent progress in artificial intelligence has renewed interest in building machines that learn like animals. Almost all of the work comparing learning across biological and artificial systems comes from studies where animals and machines received different training data, obscuring whether differences between animals and machines emerged from differences in learning mechanisms versus training data. We present an experimental approach-a "newborn embodied Turing Test"-that allows newborn animals and machines to be raised in the same environments and tested with the same tasks, permitting direct comparison of their learning abilities. To make this platform, we first collected controlled-rearing data from newborn chicks, then performed "digital twin" experiments in which machines were raised in virtual environments that mimicked the rearing conditions of the chicks. We found that (1) machines (deep reinforcement learning agents with intrinsic motivation) can spontaneously develop visually guided preference behavior, akin to imprinting in newborn chicks, and (2) machines are still far from newborn-level performance on object recognition tasks. Almost all of the chicks developed view-invariant object recognition, whereas the machines tended to develop view-dependent recognition. The learning outcomes were also far more constrained in the chicks versus machines. Ultimately, we anticipate that this approach will help researchers develop embodied AI systems that learn like newborn animals.

What problem does this paper attempt to address?

The paper attempts to address the problem of how to build machines that can learn like newborn animals. Specifically, the researchers designed an experimental platform called the "Newborn Entity Turing Test," allowing newborn animals (chicks in this study) and machines to grow in the same environment and undergo the same task tests. This approach permits a direct comparison of the learning abilities of both. The focus of the paper is to verify whether machines can spontaneously develop visual preference behaviors similar to those of chicks (such as imprinting behavior) and viewpoint-invariant object recognition capabilities. The study found that although machines can mimic these behaviors to some extent, there is still a significant performance gap compared to real newborn animals. Additionally, the research emphasizes that biological systems have much higher constraints on learning outcomes than current artificial systems, suggesting that future research needs to explore more innovative methods to narrow this gap.

A newborn embodied Turing test for view-invariant object recognition