未登录,请登录后再发表信息
最新评论 (0)
播放视频

人工智能可以帮我们分辨动物吗?

Can AI Help Us Identify Animals?

Hi, this is Kate from MinuteEarth.
大家好 我是来自《分钟地球》的凯特
I know from personal experience that
我从我的个人经历中了解到
it can take a ton of time, energy, and resources
仅仅收集一点点关于野生动物的信息
to gater even a teensy bit of information about wild animals
可能就要花费大量的时间 精力和资源
But recently, technology has given wildlife biologists new ways to study critters
但最近 随着科技发展 野生动物学家拥有了研究生物的新方法
– like using undersea microphones to monitor humpback whales
比如 用水下麦克风监测座头鲸
and satellites to study elephants
以及用卫星来研究大象
– and these methods can generate mountains of information really rapidly.
这些方式可以快速生成大量信息
Take the Snapshot Serengeti project,
例如塞伦盖蒂快照项目中
where more than 200 automatic cameras across the savannah
草原上的200多台自动摄像机
have amassed millions of photos of animals
已经拍摄了数百万张动物照片
– predator and prey, day and night,
捕食者和猎物 白天和黑夜
near and far, head-on and…whatever that is.
近照和远照 正面照和其他等等各种照片
That kind of bonanza is bananas!
那种“橡胶”其实是香蕉
But wildlife biologists have had trouble figuring out how to actually deal with all that information.
但是 野生动物学家们的麻烦在于如何处理这些信息
I mean, when the Snapshot Serengeti scientists go to actually use these images,
也就是说 当塞伦盖蒂快照项目的科学家实际使用这些图像时
they can’t just plug them into a spreadsheet…
他们不能仅仅把它们插入电子表格
they need to know who’s doing what with who in each photo
他们需要知道每张照片中谁和谁正在做些什么
– in other words, they need data.
换言之 他们需要数据
And turning all those photos into data has already taken more than 30,000 hours
但是将这些照片转换为数据要花费超过3万小时
– that’s like working 8 hours a day for 14 years.
如果一天工作8小时 那么将需要14年
If only there was some sort of automatic system that could do the job;
要是有某种自动化系统可以做这项工作 问题将迎刃而解
one that can learn that not only is this is a zebra, but so is this, and this, well…used to be one.
这样的系统可以分辨出不仅这是一匹斑马 这个也是 以及这个……曾经也是
Some sort of…artificial intelligence.
也就是某种……人工智能
The scientists built a computer program
科学家开发了 一个计算机程序
that essentially uses trial and error
通过反复实验
to learn to identify dozens of species, count animals, and categorize behaviors.
学习如何识别众多物种 给动物计数 以及分类动物行为
The AI turned out to be able to do this stuff about as accurately as humans
人工智能几乎可以做到和人类一样精确
– and the photos that took 14 years of human labor to classify
同时 人类需要14年才能完成分类的照片
took the program just a single day.
人工智能只需要一天就能完成
Plus, the more opportunities these kinds of algorithms have to practice, the better they get,
另外 这类算法运行的次数越多 效果就越好
so the mountains of information that were such a burden to researchers
因此 对研究者而言是巨大负担的大量信息
are actually awesome for artificial intelligence.
实际上对优化人工智能十分有利
And you know another place we can go to find mountains and mountains of information?
而你知道另一个我们能找到大量信息的地方吗?
The internet.
互联网
One AI-based program spends each night
一个人工智能程序每天晚上
combing through the 80 years of Youtube videos uploaded that day,
都会梳理当天上传到Youtube的时长近80年的视频
looking for videos of whale sharks
寻找有关鲸鲨的视频
like that awesome one someone saw on their vacation.
就像某些人在他们的假期看到的这只一样
The bot IDs each individual based on its unique constellation of spots
人工智能根据每只鲸鲨身上独一无二的斑点来识别不同个体
and records info about the sighting
并记录有关信息
– all while the researchers are catching some zzz’s.
这一切都在研究人员呼呼大睡时进行
That’s a whale of a change from what it has normally taken to study animals in the wild,
这与通常情况下在野外研究动物相比是一个巨大的变化
and AI just might free up researchers’ time and resources to connect the dots
人工智能使研究者有更多的时间和资源来理清头绪
and draw new conclusions about their subjects
并得出新结论
– a task where artificial intelligence still can’t beat the good old-fashioned human kind.
在这方面 人工智能仍然无法胜过传统的人类
This video was sponsored by the university of Minnesota
本视频由明尼苏达大学赞助
where students, faculty and staff across all fields of study are working to solve the grand challenges facing society.
在这儿 各个领域的学生和教职员工致力于解决社会面临的巨大挑战
One of these challenges is assuring clean water and sustainable ecosystems
其中一个任务便是保证水资源清洁和生态系统可持续发展
which includes making sure that we understand ecosystems well enough to
这要求我们足够了解生态系统
help them thrive into the future
来帮助它们繁荣发展
Dr. Criag Packer, the director of the university’s lion center
克雷格·帕克博士 该大学狮子研究中心的管理者
helped establish the Snapshot Serengeti Project
协助建立了塞伦盖提快照项目
to collect information about how animals interact with each other and with the environment,
来收集有关动物与动物 动物与环境之间如何相互影响的信息
so that we can better conserve species and manage ecosystems around the world
以便我们更好的保护物种以及生态环境
Thanks, University of Minnesota!
谢谢 明尼苏达大学

发表评论

译制信息
视频概述

人工智能可以分辨出不同的动物吗?该视频介绍了应用于野生动物研究的人工智能技术。

听录译者

收集自网络

翻译译者

咩咩

审核员

审核员BY

视频来源

https://www.youtube.com/watch?v=tSoqJpisKIg

相关推荐