Green Brain: Computational Modelling of the Honeybee Brain

建造能够像人脑一样或比人脑更好地执行复杂认知任务的智能机器，是现代科学的长期挑战。IBM的深蓝国际象棋计算机在1997年击败了世界冠军卡斯帕罗夫，这是这一追求的亮点之一。尽管它在国际象棋中表现出色，但这个系统与大脑完全不同，也不像大脑那样强大和多功能。最近，同样由IBM提供部分资金的蓝脑计划开始建立一个人类大脑皮质柱的真实比例模型，使我们更接近最终建立一个与生物大脑功能相似的人工大脑的目标。在绿色大脑项目中，我们建议建造这样一个人工大脑，但蜜蜂的大脑更小。我们将与图卢兹Martin Giurfa教授的世界领先的研究小组合作，他们是蜜蜂大脑解剖、生理学以及蜜蜂认知和行为的方方面面的专家。蜜蜂具有惊人的巨大认知能力，包括从气味到颜色等感官通道之间的习得联系的转移，以及学习抽象概念，如“相同”和“不同”的范畴。同时，他们的大脑比复杂的人脑要小得多，结构也好得多，(从比例上讲)也要比复杂的人脑研究得更好。在绿色大脑项目中，我们将为蜜蜂最重要的两个感觉系统--嗅觉系统和视觉系统--建立详细的计算机模型。在这样做的过程中，我们将纳入现有的数据、模型和原理，并进一步确定它们是如何产生所观察到的令人印象深刻的认知能力的。然后，我们将与Giurfa实验室的专家密切合作，将感觉系统与学习模型和感觉整合模型结合起来，最终建立蜜蜂大脑的全尺寸模型。该模型将在最先进的基于大规模并行图形处理单元(GPU)的超级计算机上实现，这是一项由NVIDIA Corporation带头的新技术，NVIDIA Corporation通过捐赠GPU硬件来支持这一项目。使用GPU计算将使我们能够实时模拟我们的绿色大脑模型，这对于项目的最后阶段至关重要，届时我们将使绿色大脑作为自主飞行机器人的大脑工作。这比目前的大脑模型工作取得了重要的进一步进展，因为越来越清楚的是，大脑功能的一个基本方面是大脑不是孤立地活动，而是与身体和环境不断地相互作用。“体现”的概念及其对认知的影响是现代认知科学的重要见解，对现代神经科学也将变得同样重要。绿色大脑项目的成果将在几个学术领域产生影响。在神经科学方面，它将推动大规模大脑模型领域的发展，以及我们对蜜蜂感觉系统中信息是如何处理的理解。我们还将为将现代GPU技术用于人工脑并将其应用于仿生机器人提供新的工具。对于认知科学，我们将有助于理解生物现实模型系统中的体现。此外，在学术界，自主飞行机器人的发展可能在环境探测、搜救和人工授粉方面有应用。更好地了解认知的潜在机制，最终可能会转化为对人类认知和认知障碍的更深入了解。最后，更好地了解蜜蜂本身可能被证明是重要的，因为在大多数生态系统中，蜜蜂是关键的传粉者，因此是“关键物种”，对粮食安全至关重要。

Building intelligent machines that can perform complex cognitive tasks as well as or better than the human brain is a long-standing challenge of modern science. This quest has seen one of its highlights when IBM's Deep Blue chess computer beat the world champion Kasparov in 1997. Despite its superior performance in chess, this system was however in no way similar to or as powerful and versatile as a brain. More recently the Blue Brain initiative, also partially funded by IBM, set out to build a real-scale model of a cortical column of the human brain, moving us closer to the goal of eventually building an artificial brain that works like its biological counterpart. In the Green Brain project we propose to build such an artificial brain, but of the smaller brain of the honeybee. We will work with the world-leading research group of Prof. Martin Giurfa in Toulouse, who are experts in all aspects of bee brain anatomy, physiology and bee cognition and behavior. Bees have a surprisingly large cognitive capacity including transfer of learned associations across sensory modalities, e.g. from smells to colors, and learning abstract concepts such as the categories of "the same" and "different". At the same time their brains are much smaller, structured and (proportionally) much better researched than the complex human brain. It is also much easier to perform invasive manipulations to dissect how different parts of the bee brain function.In the Green Brain project we will build detailed computer models of the two most important sensory systems of the bee, the senses of smell (olfactory system) and of sight (visual system). In doing so, we will incorporate existing data, models and principles and identify further how they give rise to the observed impressive cognitive abilities. We will then combine the sensory systems with learning models and models of sensory integration in close collaboration with the experts in the Giurfa lab to eventually build a full-scale model of the bee brain. This model will be implemented on state-of-the-art massively parallel graphical processing unit (GPU) based super-computers, a new technology spearheaded by NVIDIA Corporation who is supporting this project with GPU hardware donations. Using GPU computing will allow us to simulate our Green Brain model in real time, which will be essential for the final phase of the project when we will put the Green Brain to work as the brain of an autonomous flying robot. This is an important further advance over current work on brain models because it is becoming more and more clear that an essential aspect of brain function is that the brain is not acting in isolation but in constant interaction with the body and the environment. This concept of "embodiment" and its consequences for cognition are important insights of modern cognitive science and will become equally important for modern neuroscience. The outputs from the Green Brain project will have impacts in several academic areas. In the neurosciences it will advance the field of large scale brain models and our understanding of how information is processed in the sensory systems of bees. We will also contribute new tools for the use of modern GPU technology for artificial brains and employing them in bio-mimetic robotics. For the cognitive sciences we will contribute to the understanding of embodiment in biologically realistic model systems.Beyond academia, the development of autonomous flying robots may have applications in environmental exploration, search and rescue and artificial pollination. Developing a better understanding of the mechanisms underlying cognition may ultimately translate into greater insights into human cognition and cognitive disorders. Finally, developing a better understanding of the honeybee may prove to be important in its own right as bees are a key pollinator in most ecologies and hence a 'keystone species' and vital for food security.