Scientists Create ‘Biocomputer’ with Lab-Grown Brain Tissue

Scientists Create ‘Biocomputer’ with Lab-Grown Brain Tissue

科学家用实验室培养的脑组织创建了“生物计算机”

【听力音频】

-----------------------------------------------------------------------------------

　　By Bryan Lynn

　　17 December 2023

　　American researchers have combined lab-grown human brain tissue with computer hardware to create a working biocomputer.

　　The scientists say brain cells used in the experiment were able to recognize speech and complete simple math problems.

　　The team made brain-like tissue that took the form of what they called a "brain organoid." Harvard University's Stem Cell Institute explains that an organoid is a collection of individualized, complex cells that can be grown from stem cells in a lab.

　　Under the right laboratory conditions, organoids can be made to look and even work similarly to real human tissue and organs. In this process, stem cells "can follow their own genetic instructions to self-organize," the Stem Cell Institute says.

　　So far, scientists have been able to produce organoids that look like, or resemble, some organs. These organs include the brain, kidney, lung, stomach and liver. Such lab-created organoids are generally used to study how organs work without needing to experiment on actual organs.

　　In the biocomputer experiment, the team said stem cells were able to form neurons similar to those found in the human brain. Neurons are electrically charged cells that transport signals to the brain and other parts of the body.

　　Feng Guo led the experiment. He is a bioengineer and professor of Intelligent System Engineering at Indiana University Bloomington. His team recently published their research results in a study in Nature Electronics.

　　The researchers attached the brain organoid to a set of traditional electronic computing circuits. The researchers call this system Brainoware. The system was used to establish communication between the organoid and electronic circuits. An artificial intelligence (AI) tool was used to help read the neural activity of the organoid.

　　The scientists aim to build "a bridge between AI and organoids," Guo explained to Nature. Guo believes that combining organoids and computer circuits could provide additional speed and energy to improve the performance of AI computing systems.

　　The study notes that adding human brain power might be able to help machines with the things they do not do as well as people. For example, the researchers said humans generally have a faster learning ability and use less energy thinking than computers do.

　　During one part of the experiment, the team tested the Brainoware system's voice recognition ability. The team trained the system on 240 recordings of eight different voices. The researchers said the organoid produced different neural signals in reaction to the different voices. The accuracy level of the system reached 78 percent, Guo said.

　　"This is the first demonstration of using brain organoids [for computing]," Guo told MIT Technology Review. He added, "It's exciting to see the possibilities of organoids for biocomputing in the future."

　　Guo said these results persuaded his team that a brain-computer system can work to improve computing performance, especially for some AI jobs. But he noted the best accuracy rates recorded by the Brainoware system were still below the accuracy rates of traditional AI networks. Guo said this is one of the things his team plans to try to improve.

　　Lena Smirnova is a developmental neuroscientist at Johns Hopkins University in Baltimore, Maryland. She told Nature that more research will be needed to improve such systems. But she said, "The study confirms some key theoretical ideas that could eventually make a biological computer possible."

　　Smirnova noted that in earlier experiments, researchers have used other kinds of neuron cells to perform similar computational activities. But the latest study, she said, was the first to demonstrate this kind of performance in a brain organoid.

　　I'm Bryan Lynn.

　　Bryan Lynn wrote this story for VOA Learning English, based on reports from Nature Electronics, The Associated Press, Harvard's Stem Cell Institute and MIT Technology Review.

-----------------------------------------------------------------------------------

　　作者：Bryan Lynn

　　日期：2023年12月17日

　　美国研究人员将实验室培养的人脑组织与计算机硬件结合，创建了一个工作的生物计算机。

　　科学家们表示，实验中使用的脑细胞能够识别语音并完成简单的数学问题。

　　该团队制造了类似大脑的组织，他们称之为"大脑类器官"。哈佛大学干细胞研究所解释说，器官样是一种可以在实验室中从干细胞培养出来的个体化、复杂的细胞集合。

　　在适当的实验室条件下，器官样可以被制造得看起来甚至工作起来像真正的人体组织和器官。在这个过程中，干细胞“可以遵循自己的遗传指令进行自我组织”，干细胞研究所说。

　　到目前为止，科学家们已经能够生产出看起来像，或者类似于一些器官的器官样。这些器官包括大脑、肾脏、肺、胃和肝脏。这种实验室创造的器官样通常用于研究器官的工作原理，而无需在实际的器官上进行实验。

　　在生物计算机实验中，团队表示干细胞能够形成类似于人脑中发现的神经元。神经元是带电的细胞，它们将信号传输到大脑和身体的其他部分。

　　郭峰领导了这次实验。他是印第安纳大学布卢明顿分校的生物工程师和智能系统工程教授。他的团队最近在《自然电子学》上发表了他们的研究结果。

　　研究人员将大脑器官样连接到一套传统的电子计算电路上。研究人员称这个系统为Brainoware。该系统被用来建立器官样和电子电路之间的通信。一个人工智能（AI）工具被用来帮助读取器官样的神经活动。

　　科学家们的目标是建立“一个连接AI和器官样的桥梁”，郭峰向《自然》杂志解释说。郭峰认为，将器官样和计算机电路结合起来，可能提供额外的速度和能量，以提高AI计算系统的性能。

　　研究指出，增加人脑的力量可能能够帮助机器改善它们不如人类做得好的事情。例如，研究人员表示，人类通常具有更快的学习能力，并且在思考时比计算机使用的能量更少。

　　在实验的一部分中，团队测试了Brainoware系统的语音识别能力。团队训练了该系统，使用了8种不同声音的240个录音。研究人员表示，器官样对不同的声音产生了不同的神经信号。郭峰说，系统的准确率达到了78%。

　　"这是首次使用大脑器官样进行计算的演示，"郭峰告诉MIT科技评论。他补充说："看到器官样在未来生物计算中的可能性，真是令人兴奋。"

　　郭峰说，这些结果使他的团队确信，大脑-计算机系统可以用来提高计算性能，特别是对于一些AI工作。但他指出，Brainoware系统记录的最佳准确率仍低于传统AI网络的准确率。郭峰说，这是他的团队计划尝试改进的事情之一。

　　Lena Smirnova是马里兰州巴尔的摩约翰霍普金斯大学的发育神经科学家。她告诉《自然》杂志，需要进行更多的研究来改进这样的系统。但她说："这项研究确认了一些关键的理论观点，这些观点最终可能使生物计算机成为可能。"

　　Smirnova指出，在早期的实验中，研究人员已经使用了其他类型的神经细胞来执行类似的计算活动。但她说，最新的研究是首次在大脑器官样中展示这种性能。

　　我是Bryan Lynn。

　　Bryan Lynn根据《自然电子学》、美联社、哈佛干细胞研究所和MIT科技评论的报道，为VOA学英语写了这个故事。

-----------------------------------------------------------------------------------

　　这个故事中的词汇

　　instruction – n. 指示，步骤，如何做某事的一套步骤

　　circuit – v. 电流周游的完整圈路

　　accurate – adj. 准确的，正确的，精确的

　　theoretical – adj. 基于与某一主题相关的理论，而非该主题的实际应用