- Published on
用人工智能推测大脑如何学习(Neuroscience: Great minds)
- Authors
用人工智能推测大脑如何学习
他希望通过开发能够学习解决复杂问题的人工神经网络，去揭示大脑神经网络如何做到同样的事情。 
双语精读
Neuroscience: Great minds
神经科学：了不起的大脑
Five decades of research into artificial neural networks have earned Geoffrey Hinton the moniker of the Godfather of artificial intelligence (AI). Work by his group at the University of Toronto laid the foundations for today's headline-grabbing AI models, including ChatGPT and LaMDA.
五十年来对人工神经网络的研究为杰弗里·辛顿赢得了"人工智能教父"的称号。他在多伦多大学的研究小组的工作为ChatGPT和LaMDA等引人注目的人工智能模型奠定了基础。
These can write coherent (if uninspiring) prose, diagnose illnesses from medical scans and navigate self-driving cars. But for Dr Hinton, creating better models was never the end goal.
这些模型可以写出连贯（虽然平淡无奇）的文章，通过医学扫描诊断疾病，并驾驶自动驾驶汽车。但对于辛顿博士来说，创建更好的模型从来都不是最终目标。
His hope was that by developing artificial neural networks that could learn to solve complex problems, light might be shed on how the brain's neural networks do the same.
他希望通过开发能够学习解决复杂问题的人工神经网络，去揭示大脑神经网络如何做到同样的事情。
Brains learn by being subtly rewired: some connections between neurons, known as synapses, are strengthened, while others must be weakened.
大脑通过巧妙地重新连接来学习：一些神经元之间的连接（也叫突触）得到加强，而其他连接则必须削弱。
But because the brain has billions of neurons, of which millions could be involved in any single task, scientists have puzzled over how it knows which synapses to tweak and by how much.
但因为大脑有数十亿个神经元，其中的数百万个可能参与任何一项任务，所以科学家们一直在思索大脑如何知道要调整哪些突触以及调整到何种程度。
Dr Hinton popularised a clever mathematical algorithm known as backpropagation to solve this problem in artificial neural networks. But it was long thought to be too unwieldy to have evolved in the human brain.
辛顿博士推广了一种叫做反向传播的巧妙数学算法，来解决人工神经网络中的这个问题。但长期以来人们认为这种方法太过复杂笨拙，不可能在人脑中进化出来。
Now, as AI models are beginning to look increasingly human-like in their abilities, scientists are questioning whether the brain might do something similar after all.
现在，随着人工智能模型的能力开始越来越像人类，科学家们开始怀疑大脑是否还是能做类似反向传播的事情。
Working out how the brain does what it does is no easy feat. Much of what neuroscientists understand about human learning comes from experiments on small slices of brain tissue, or handfuls of neurons in a Petri dish.
弄清楚大脑的工作原理绝非易事。神经科学家对人类学习过程的大部分理解都来自对小片脑组织或培养皿中少量神经元的实验。
It's often not clear whether living, learning brains work by scaled-up versions of these same rules, or if something more sophisticated is taking place.
人们通常不清楚活的、学习中的大脑是否按照这些相同规则的放大版本工作，还是发生了更复杂的事情。
Even with modern experimental techniques, wherein neuroscientists track hundreds of neurons at a time in live animals, it is hard to reverse-engineer what is really going on.
即使运用现代实验技术，让神经科学家可以同时追踪活体动物的数百个神经元，也很难逆推真正发生了什么。
词汇预习
increasingly [高考]
美[ɪnˈkriːsɪŋli] | 英[ɪnˈkriːsɪŋli]
adv. 逐渐地；越来越多地
mathematical [高考]
美[ˌmæθəˈmætɪkl] | 英[ˌmæθəˈmætɪkl]
adj. 数学的；精确的；可能性极小的
diagnose [高考]
美[ˌdaɪəɡˈnoʊs] | 英[ˈdaɪəɡnəʊz]
v. 判断；诊断(疾病)
coherent [高考]
美[koʊˈhɪrənt] | 英[kəʊˈhɪərənt]
adj. 连贯的；有条理的；一致的；粘着的
experimental [高考]
美[ɪkˌsperɪˈmentl] | 英[ɪkˌsperɪˈmentl]
adj. 实验的；实验性的；试验性的
reverse [高考]
美[rɪˈvɜːrs] | 英[rɪˈvɜːs]
v. 逆转；倒退；互换；改变 adj. 相反的；反面的；颠倒的 n. 相反；背面；失败；倒档
navigate [高考]
美[ˈnævɪɡeɪt] | 英[ˈnævɪɡeɪt]
vi. 航行；驾驶；操纵 vt. 航行；驾驶；横渡；[口]设法穿过，走过，经过
complex [高考]
美[kəmˈpleks , ˈkɑːmpleks] | 英[ˈkɒmpleks]
adj. 复杂的；合成的；复合的 n. 综合体；复合体；[医]综合症状；[心]情结
intelligence [高考]
美[ɪnˈtelɪdʒəns] | 英[ɪnˈtelɪdʒəns]
n. 智力；智慧；才智；（尤指关于敌国的）情报
artificial [高考]
美[ˌɑːrtɪˈfɪʃl] | 英[ˌɑːtɪˈfɪʃl]
adj. 人造的；虚伪的；武断的
tissue [高考]
美[ˈtɪʃuː] | 英[ˈtɪʃuː]
n. 纸巾；薄绢；(生物的)组织；织物
self [高考]
美[self] | 英[self]
n. 自己；本性；自我 pron. 我自己；你自己；他或她自己 adj.同一的；纯净的；单一的 v. 近亲繁殖；自花授粉
be involved in [高考]
美[bi ɪnˈvɑlvd ɪn] | 英[bi: ɪnˈvɔlvd in]
涉及到； 与 ... 有关联
research into [高考]
美[rɪˈsɚtʃ ˈɪntu] | 英[riˈsə:tʃ ˈɪntuː]
就…研究
living [高考]
美[ˈlɪvɪŋ] | 英[ˈlɪvɪŋ]
adj. 活着的；生动的 n. 生活；生计
sophisticated [高考]
美[səˈfɪstɪkeɪtɪd] | 英[səˈfɪstɪkeɪtɪd]
adj. 老练的；精密的；复杂的；久经世故的
shed [四级]
美[ʃed] | 英[ʃed]
vt. 除掉；使脱落；使流出；使清晰 n. 小屋；车棚
neural [四级]
美[ˈnʊrəl] | 英[ˈnjʊərəl]
adj. 神经的
involved in [四级]
美 | 英
涉及； 包含在 ... 中
involved [四级]
美[ɪnˈvɑlvd] | 英[ɪnˈvɒlvd]
adj. 涉及的；牵连的；复杂的；感情投入的；有密切关系的
prose [六级]
美[proʊz] | 英[prəʊz]
adj. 散文的 n. 散文
feat [六级]
美[fiːt] | 英[fiːt]
n. 技艺；功绩；技艺表演
learning [专八]
美[ˈlɜːrnɪŋ] | 英[ˈlɜːnɪŋ]
n. 学问；学识；学习
subtly [考研]
美['sʌtlɪ] | 英['sʌtlɪ]
adv. 敏锐地；巧妙地；精细地
unwieldy [托福]
美[ʌnˈwildi] | 英[ʌnˈwi:ldi]
adj. 笨重的；笨拙的；庞大的
developing [雅思]
美[dɪˈvɛləpɪŋ] | 英[dɪˈveləpɪŋ]
adj. 发展中的
重点讲解
Neuroscience: Great minds
神经科学：了不起的大脑
earn v. （通过努力）博得
【例】Glory is not given to you. You have to earn it.
光荣不是白给你的，你必须靠自己去赢得光荣。
【搭配】earn sb sth 为某人赢得某物
【拓展】a well-earned rest 值得好好休息一下
lay the foundation 奠定基础
【例】The builders are now beginning to lay the foundations of the new school.
建筑工人正开始给新校舍打地基。
【反义词】shake/rock the foundation 动摇基础
Five decades of research into artificial neural networks have earned Geoffrey Hinton the moniker of the Godfather of artificial intelligence (AI). Work by his group at the University of Toronto laid the foundations for today's headline-grabbing AI models, including ChatGPT and LaMDA.
五十年来对人工神经网络的研究为杰弗里·辛顿赢得了"人工智能教父"的称号。他在多伦多大学的研究小组的工作为ChatGPT和LaMDA等引人注目的人工智能模型奠定了基础。
earn v. （通过努力）博得
【例】Glory is not given to you. You have to earn it.
光荣不是白给你的，你必须靠自己去赢得光荣。
【搭配】earn sb sth 为某人赢得某物
【拓展】a well-earned rest 值得好好休息一下
lay the foundation 奠定基础
【例】The builders are now beginning to lay the foundations of the new school.
建筑工人正开始给新校舍打地基。
【反义词】shake/rock the foundation 动摇基础
These can write coherent (if uninspiring) prose, diagnose illnesses from medical scans and navigate self-driving cars. But for Dr Hinton, creating better models was never the end goal.
这些模型可以写出连贯（虽然平淡无奇）的文章，通过医学扫描诊断疾病，并驾驶自动驾驶汽车。但对于辛顿博士来说，创建更好的模型从来都不是最终目标。
coherent adj. 符合逻辑的；前后一致的；连贯的
【例】Her argument was clear and coherent, making it easy to follow.
她的论证清晰、连贯，很容易理解。
The excited woman was not very coherent.
那个女人情绪激动，语无伦次/前言不搭后语。
【辨析】cohesive adj. 衔接的；结成一个整体的
shed light on sth 阐明某事
【例】The new evidence sheds light on the cause of the accident.
新的证据说明了事故发生的原因是什么。
【近义词】illuminate/enlighten v. 阐明；使变得清楚
【反义词】be in the dark about sth 对某事一无所知
His hope was that by developing artificial neural networks that could learn to solve complex problems, light might be shed on how the brain's neural networks do the same.
他希望通过开发能够学习解决复杂问题的人工神经网络，去揭示大脑神经网络如何做到同样的事情。
coherent adj. 符合逻辑的；前后一致的；连贯的
【例】Her argument was clear and coherent, making it easy to follow.
她的论证清晰、连贯，很容易理解。
The excited woman was not very coherent.
那个女人情绪激动，语无伦次/前言不搭后语。
【辨析】cohesive adj. 衔接的；结成一个整体的
shed light on sth 阐明某事
【例】The new evidence sheds light on the cause of the accident.
新的证据说明了事故发生的原因是什么。
【近义词】illuminate/enlighten v. 阐明；使变得清楚
【反义词】be in the dark about sth 对某事一无所知
Brains learn by being subtly rewired: some connections between neurons, known as synapses, are strengthened, while others must be weakened.
大脑通过巧妙地重新连接来学习：一些神经元之间的连接（也叫突触）得到加强，而其他连接则必须削弱。
rewire v. 重新连接；重新布线
【例】The company rewired its network to improve internet speed and reliability.
公司重新布置了网络线路，从而提高网速、加强网络稳定性。
【拓展】wire n. 铁丝；电线
hard-wired adj. 与生俱来的
puzzle over 苦苦思考；琢磨
【例】She puzzled over the complex math problem for hours.
这道数学难题她思考了好几个小时。
【拓展】puzzle out sth 琢磨出了某事的答案
【例】She finally puzzled out the answer to the math problem.
她终于思考出了这道数学题的答案。
tweak v. 拧；轻微调整
【例】You just need to tweak the last paragraph and then it's done.
你只需要把最后一段稍微修改一下，然后就完工了。
But because the brain has billions of neurons, of which millions could be involved in any single task, scientists have puzzled over how it knows which synapses to tweak and by how much.
但因为大脑有数十亿个神经元，其中的数百万个可能参与任何一项任务，所以科学家们一直在思索大脑如何知道要调整哪些突触以及调整到何种程度。
rewire v. 重新连接；重新布线
【例】The company rewired its network to improve internet speed and reliability.
公司重新布置了网络线路，从而提高网速、加强网络稳定性。
【拓展】wire n. 铁丝；电线
hard-wired adj. 与生俱来的
puzzle over 苦苦思考；琢磨
【例】She puzzled over the complex math problem for hours.
这道数学难题她思考了好几个小时。
【拓展】puzzle out sth 琢磨出了某事的答案
【例】She finally puzzled out the answer to the math problem.
她终于思考出了这道数学题的答案。
tweak v. 拧；轻微调整
【例】You just need to tweak the last paragraph and then it's done.
你只需要把最后一段稍微修改一下，然后就完工了。
Dr Hinton popularised a clever mathematical algorithm known as backpropagation to solve this problem in artificial neural networks. But it was long thought to be too unwieldy to have evolved in the human brain.
辛顿博士推广了一种叫做反向传播的巧妙数学算法，来解决人工神经网络中的这个问题。但长期以来人们认为这种方法太过复杂笨拙，不可能在人脑中进化出来。
unwieldy adj. 不灵巧的；尾大不掉的
【例】A piano is a very unwieldy item to get down a flight of stairs.
钢琴很笨重，不容易从楼梯上搬下来。
The organization is unwieldy and slow to change.
这个组织尾大不掉，做出改变很慢。
【拓展】wield v. 挥；使用（武器、工具等）
Now, as AI models are beginning to look increasingly human-like in their abilities, scientists are questioning whether the brain might do something similar after all.
现在，随着人工智能模型的能力开始越来越像人类，科学家们开始怀疑大脑是否还是能做类似反向传播的事情。
unwieldy adj. 不灵巧的；尾大不掉的
【例】A piano is a very unwieldy item to get down a flight of stairs.
钢琴很笨重，不容易从楼梯上搬下来。
The organization is unwieldy and slow to change.
这个组织尾大不掉，做出改变很慢。
【拓展】wield v. 挥；使用（武器、工具等）
Working out how the brain does what it does is no easy feat. Much of what neuroscientists understand about human learning comes from experiments on small slices of brain tissue, or handfuls of neurons in a Petri dish.
弄清楚大脑的工作原理绝非易事。神经科学家对人类学习过程的大部分理解都来自对小片脑组织或培养皿中少量神经元的实验。
feat n. 伟绩；壮举；英勇事迹
【例】Climbing Mount Everest is a remarkable feat of endurance.
攀登珠峰是考验耐力的壮举。
【近义词】enterprise/undertaking n. 艰巨而重大的事业
It's often not clear whether living, learning brains work by scaled-up versions of these same rules, or if something more sophisticated is taking place.
人们通常不清楚活的、学习中的大脑是否按照这些相同规则的放大版本工作，还是发生了更复杂的事情。
scale sth up 扩大；增大
【例】to scale up production 扩大生产规模
【反义词】scale sth down/back 减少；缩小
【拓展】scale n. 规模；比例尺
reverse-engineer v. 逆向工程
【例】By reverse-engineering the competitor's product, they were able to create a similar one at a lower cost.
他们把竞争对手的产品拆开并仿制，从而用低成本造出了类似的产品。
【拓展】reverse adj. 逆向的；相反的
Even with modern experimental techniques, wherein neuroscientists track hundreds of neurons at a time in live animals, it is hard to reverse-engineer what is really going on.
即使运用现代实验技术，让神经科学家可以同时追踪活体动物的数百个神经元，也很难逆推真正发生了什么。
scale sth up 扩大；增大
【例】to scale up production 扩大生产规模
【反义词】scale sth down/back 减少；缩小
【拓展】scale n. 规模；比例尺
reverse-engineer v. 逆向工程
【例】By reverse-engineering the competitor's product, they were able to create a similar one at a lower cost.
他们把竞争对手的产品拆开并仿制，从而用低成本造出了类似的产品。
【拓展】reverse adj. 逆向的；相反的
课后练习
What was Geoffrey Hinton's ultimate goal in developing artificial neural networks?
A. To figure out how to advance the AI technology.
B. To define the essence of human consciousness.
C. To create AI models that can do everything for humans.
D. To understand how the human brain solves complex problems.
What has puzzled scientists about how the brain learns?
A. How synapses are initially formed.
B. How neurons communicate with each other.
C. How the brain decides which synapses to adjust.
D. How many neurons are involved in a single task.
What are scientists suspecting about the human brain?
A. It uses forward propagation to learn.
B. It might solve similar problems as AI models do.
C. It does not involve synaptic changes for learning.
D. It might use a mechanism similar to backpropagation.
Why is it difficult for neuroscientists to understand how the brain learns?
A. Because they can only study small samples and cannot get the big picture.
B. Because they cannot always apply modern experimental techniques.
C. Because they can easily reverse-engineer the brain's processes.
D. Because they cannot obtain useful data from live animals.
