用AI拯救大象(下)

And then they hit record.

然后他们按下了录音键。

Three months later, they would return to the forest, locate the recorders, change the batteries, put in new audio cards, and start all over again.

三个月后，他们会回到森林，找到录音机，更换电池，然后放入新的声音存储卡，重新开始录音。

As the months wore on, the recorders were collecting hundreds of thousands of hours of jungle sounds, more than any team of graduate students could realistically listen to — which meant Wrege had another problem: How could he sort through all these recordings to find the elephant voices he wanted?

几个月过去了，录音机手机了成千上万小时的来自丛林的声音，团队里的研究生们都听不过来了——这也意味着雷吉面临另一个问题：他怎样从这些录音里找到他想找的大象的声音呢？

"In AI circles this is known as the 'cocktail party problem,' " said computer scientist Josephine Wolff, who is now a professor at the Fletcher School at Tufts University.

“在AI圈里，这个问题叫做“鸡尾酒派对问题”，”电脑科学家乔瑟芬·沃尔夫说道，他现在是塔夫斯大学弗莱彻学院的一名教授。

"At a party with a lot of background noise, the human brain can focus on a specific person's voice and amplify that above all the other voices. AI can do the same thing."

“在有很多背景杂音的派对上，人脑却可以集中关注某一个特定人的声音，将其放大压过别的声音。AI也可以做到同样的”

In fact, there's a subset of AI — something called a neural network — that is very good at this.

实际上，有一种AI设备—类似神经网络系统—非常擅长此事。

A neural network is essentially a group of algorithms, or mathematical equations, working together to cluster and classify information and find patterns humans wouldn't necessarily see.

这种神经网络是由一组算法或者数学方程组成的，它们一起可以聚集和分类信息然后找到人类不容易看到的规律。

It is particularly good at working with images, so Wrege ran the audio through a software program that turned the recordings he had collected into spectrograms — ghostly little pictures of sound waves.

它尤其擅长处理图像，所以雷吉把这些音频通过一个软件程序转换成了光谱图——这是幽灵般的声波的图片。

He then had a company in Santa Cruz, Calif., called Conservation Metrics build him a neural network that could sort through the cacophony of jungle sounds and find elephants.

后来，他在加州圣克鲁斯开了一家名为“环保指标”的公司，这家公司为他建立了一个神经网络，可以从丛林里的各种杂音中找出大象。

"Basically each of these 'neurons' in the network is determining how likely one piece of the spectrogram is to belong to an elephant call," Wolff explains.

“基本上网络里的每一条“神经”都可以辨别一张声谱图中的一条属于大象叫声的可能性，”沃尔夫解释道。

"Neurons in the first layer have just the spectrogram to consider, and so will likely recognize things like pitch and other things it sees as defining characteristics of elephant calls."

“第一层的神经元只需要考虑声谱图，因此很可能会识别音调和其他它认为定义了大象叫声的特征的东西。”

Then the next neuron recognizes something else relatively simple, building on what has come before it, and so on.

然后建立在之前的基础上，下一个神经元识别其他相对简单的东西，然后继续下去。

When the network thinks it has what it needs, it makes a statistical calculation; essentially asking itself, what's the likelihood that the pattern I see is an elephant? 50%? 80%?

当网络认为它有它所需要的东西时，它就会进行统计计算;本质上是问自己，我看到的图案是大象的可能性是多少?50% ?80% ?

Until it finally gets what it wants, that elephant.

直到它最终得到它想要的，那头大象。

The neural network started producing files with lots and lots of isolated elephant sounds.

这个神经网络就开始生成大量的大象声音文件。

But then an interesting thing happened — Wrege and his team at the Elephant Listening Project heard something else they weren't expecting in all those recordings: gunshots.

但后来发生了一件有趣的事情——雷吉和他的大象听力项目团队听到了他们在所有录音中没有预料到的另一种声音:枪声。

Assuming the gunshots were a pretty good proxy for poaching attempts, Wrege decided to ask Conservation Metrics to build another neural network, this time to look for the sounds of gunshots.

假设枪声是偷猎行为的一个很好的标志，雷吉决定要求环保指标建立另一个神经网络，这一次是寻找枪声。

He hoped it would provide additional information about where the forest elephants were being killed and perhaps even stop poachers before they fired.

他希望它能提供更多关于森林象在哪里被杀的信息，甚至可能在偷猎者开枪之前阻止他们。

Wrege wasn't alone in trying to find ways to marry cutting-edge technology with conservation and poaching prevention.

雷吉并不是唯一一个试图将尖端技术与保护和防止偷猎相结合的人。

Four years ago, a park manager named Craig Reid was navigating a slow-motion crisis at Liwonde National Park in Malawi.

四年前，一位名叫克雷格·里德的公园经理开始在马拉维利翁代国家公园处理一场长期的危机。

The wildlife reserve was on the verge of collapse: Infrastructure was crumbling, roads were washed out, people came in and out of the park to hunt, poaching was endemic, and elephants were terrorizing nearby villagers.

野生动物保护区濒临崩溃:基础设施摇摇欲坠，道路被冲毁，人们进进出出公园打猎，偷猎猖獗，大象也威胁着附近的村民。

"I would describe Liwonde when we found it as being in a state of terminal decline," he told us during a recent visit to the park.

“当我们发现利翁代处于最终衰退状态时，我会这样描述它，”他在最近一次去公园的时候告诉我们说。

"Effectively what would have happened had we not intervened would be a total elimination of all wildlife over the 10-year period following."

“实际上，如果我们不干预的话，接下来的10年里，所有野生动物都会被彻底猎杀。”

So Reid stole a page from big-city policing and decided to use artificial intelligence and predictive analytics to see whether it could help him manage the park.

因此，里德从大城市的警察那里偷了一页纸，决定使用人工智能和预测分析来看看它是否能帮助他管理公园。

He thought technology could help him uncover the secret rhythms of the place, anticipate poaching, and create an environment where the animals could safely be animals.

他认为技术可以帮助他发现这个地方的秘密节奏，预测偷猎，并创造一个动物可以安全生活的环境。

"When we were rugged rangers in the bush down in South Africa we would talk about the day when we'd sit behind our desks and manage the park from behind a computer screen," Reid told us.

里德告诉我们:“当我们还是南非丛林中的粗壮护林员时，我们会谈论也许有一天我们会坐在办公桌后面，在电脑屏幕后面管理公园。”

"And that really is exactly what has happened."

“如今这已经成为了现实。”

As fate would have it, the resurrection of Liwonde was happening right around the time that Allen, who died in 2018, was wrapping up that Great Elephant Census.

正如命运所料，利翁代的复苏发生在艾伦(死于2018年)完成大象普查的时候。

If Allen learned anything from that project it was that actionable information was the key to saving the elephant and better managing parks.

如果说艾伦从这个项目中学到了什么，那就是可操作的信息是拯救大象和更好地管理公园的关键。

So his company, Vulcan Inc., created EarthRanger: an analytics program turbocharged with artificial intelligence and predictive analytics.

因此，他的公司火神集团创建了地球骑警:一个充满人工智能和预测分析的分析程序

"EarthRanger was entirely customer-driven, and when I say customers here, it's essentially the park rangers," said Pawan Nrisimha, the director of product management at Vulcan, which Allen founded to tackle a host of these kinds of problems.

“地球骑警完全是客户驱动的，当我说这里的客户时，基本上是公园管理员，”火神的产品管理总监帕万·内西姆哈说，这个程序是艾伦建立的用来解决这一类的问题。

"They are the ones who said, 'OK, we are logging all our reports on paper. We need a technology solution.' And that's where we started."

“他们说，'好吧，我们把所有的报告都记录在纸上。我们需要一个技术解决方案。’这就是我们的出发点。”

The idea was to take all the information park managers like Reid had both in their heads and in their daily field reports and then make it smarter.

这个想法是把所有像里德这样的公园管理人员的头脑和他们的日常里的信息实地报告，然后使它更聪明。

They started by creating a real-time visualization program that would allow managers to see all the park assets on one screen: rangers, animals, helicopters and information from sensors and security cameras were all brought together in one place, just a mouse click away.

他们首先创建了一个实时可视化程序，可以让管理员在一个屏幕上看到所有的公园资产:管理员、动物、直升机、传感器和安全摄像头的信息都集中在一个地方，只需点击鼠标即可。

The second part was providing park managers with the analytics tools they needed to manage their patrols better.

第二部分是为公园管理者提供他们需要的分析工具来更好地管理他们的巡逻。

"We have things like heat maps that tell about problem areas, where there are snares and animal traps happening, where there are fence breaks," said Nrisimha.

瑞辛哈说:“我们有像热成像图这样的东西，它能告诉问题地区，哪里有陷阱和动物陷阱，哪里有栅栏断裂。”

"That allows them to respond to security problems in the park much more quickly."

“这让他们能更快地应对公园的安全问题。”

The program actually produces an animation that allows operators to run incidents back, like a real-life interactive video game.

这个程序实际上会生成一个动画，允许操作员回放事故，就像一个真实的交互式视频游戏一样。

Finally, there is an artificial intelligence component.

最后，还有一个人工智能组件。

"We are trying to see how we can build artificial intelligence or predictive analytics to proactively tell the park management how to do the patrols and manage the security effort better," Nrisimha said.

瑞辛哈说:“我们正在研究如何建立人工智能或预测分析，主动告诉公园管理人员如何更好地进行巡逻和管理安全工作。”

The program ingests all the information and then essentially learns the rhythms of the park to offer suggestions for how to run it most efficiently.

这个程序吸收所有的信息，然后从本质上学习公园的节奏，为如何最有效地管理它提供建议。

At Liwonde, the EarthRanger program is housed in an innocuous-looking brick building behind the main ranger station.

在利翁代，“地球骑警”项目被安置在主游骑兵站后面的一幢不起眼的砖房里。

There are rangers in camouflage uniforms sitting at computer screens and manning the radios.

穿着迷彩服的护林员坐在电脑屏幕前，操纵着收音机。

It looks like the command center of a medium-size-city police department.

它看起来像一个中型城市警察局的指挥中心。

There are flat screens on the wall, closed circuit television monitors, and two long tables with a series of computers analyzing and categorizing information coming into headquarters and running it through the EarthRanger program.

墙上有平板显示器，闭路电视监视器，还有两张长桌，上面有一系列的计算机，它们分析和分类信息，然后进入总部，并通过地球骑警程序运行这些信息。

Lawrence Munro is the park's operations manager.

劳伦斯·门罗是公园的运营经理。

A key part of his job is to plan, schedule and deploy ranger patrols.

他工作的一个关键部分是计划、安排和部署护林员巡逻。

EarthRanger is helping him work out the best way to do that.

地球骑警可以帮他更好地做好工作。

To get an idea of what it looks like, the screens on the wall are showing a real-time satellite image of the park.

为了了解它是什么样子，墙上的屏幕显示了公园的实时卫星图像。

There are little elephant icons tracking GPS location signals from collared elephants and little rhino icons that track the rare black rhinos kept in a special sanctuary deep inside the park.

公园里有小象的图标，它们在追踪有领大象发出的GPS定位信号;还有小犀牛的图标，它们在追踪公园深处一个特殊保护区里的稀有黑犀牛。

雷吉说:“当你看这些大象发出隆隆声的图片时，你会发现它们发出隆隆声的方式与人类发出元音的方式非常相似。

"When you look at the spectrogram, you can see that the energy is changing from one call to the next and those very well could be different words."

“当你看声谱图时，你可以看到能量从一个呼叫到下一个的变化，很可能是不同的词。”

Translating those words is another natural job for a neural network.

翻译这些词是神经网络的另一项自然工作。

Wrege envisions a time when it will be able to distinguish the sounds of distress or danger in the calls recorded in the forest.

雷吉设想有一天它将能够分辨出森林中记录的呼叫声中悲伤或危险的声音。

Eventually maybe they'd be able to send out authorities in real time as soon as they hear from the elephants themselves.

最终，一旦他们听到大象自己的声音，他们可能会立即派出官方人员。

That may be some years away, but if you ask Wrege whether he thinks AI will save the elephant, he is unequivocal.

这可能还需要几年的时间，但如果你问雷吉他是否认为人工智能会拯救大象，他是毫不含糊的

"I actually do," he said.

“我真的相信可以，”他说。

"It is definitely going to be our salvation."

“这肯定会是我们的救星。”

问题

文中提到了哪一个国家公园？

留言回复正确答案，前十名朋友可以获得红包奖励哦，赶快来试试吧！

感谢关注

跟amber一起看世界