使用预测神经网络生成的运动错觉也能欺骗人类

📄 中文摘要

研究提出了一种新的假设,认为视觉运动错觉是大脑预测能力的副作用。通过构建一种名为进化错觉生成器(EIGen)的生成模型,该模型基于视频预测神经网络生成新的视觉运动错觉。研究通过心理测量调查确认了所构建的错觉对人类参与者的有效性。结果支持了这样一种观点:运动错觉可能源于对大脑自身预测的感知,而非对原始视觉信息的感知。这一发现为理解视觉错觉的机制提供了新的视角。

📄 English Summary

Motion Illusions Generated Using Predictive Neural Networks Also Fool Humans

This research proposes a new hypothesis that visual motion illusions are a side effect of the brain's predictive abilities. A generative model called the Evolutionary Illusion GENerator (EIGen) is constructed, which creates new visual motion illusions based on a video predictive neural network. The effectiveness of the constructed illusions on human participants is confirmed through a psychometric survey. Results support the notion that illusory motion may arise from perceiving the brain's own predictions rather than raw visual input. This finding offers a new perspective on understanding the mechanisms behind visual illusions.

Powered by Cloudflare Workers + Payload CMS + Claude 3.5

数据源: OpenAI, Google AI, DeepMind, AWS ML Blog, HuggingFace 等