如何规范人工智能

图片来源: The Economist

导读

随着人工智能(AI)在越来越多的领域得到应用，人们有理由担心可能会出现意想不到的后果。

最直接的担忧就是，为了积累训练人工智能系统所需的数据而展开的争夺，会侵犯个人隐私。监控人们在网上做的每一件事，从购物到阅读，再到在社交媒体上发帖，让互联网巨头可以建立详细的个人资料，用于投放广告或推荐感兴趣的商品。最好的回应不是直接监管人工智能的使用，而是专注于有关如何收集、处理和储存个人资料的规则。

第二个担忧的领域是“算法偏差” —— 当系统被训练使用历史数据，它们将学习并延续现有的偏差。为了确保公平，人工智能系统需要更好地解释它们如何做出决策 (这是一个需要大量研究的领域); 人工智能应该帮助人类做出更好的决定，而不是替人类做决定。

人工智能引起担忧的第三个领域是自动驾驶汽车。许多公司现在正在测试自动驾驶汽车，并在公共道路上试行“机器人出租车”服务。但是这样的系统并不完美，在2018年3月，亚利桑那州坦佩的一名行人被一辆自动驾驶汽车撞死，这是同类事故中的首例。正确的做法是，要求自动驾驶汽车的制造商发布定期的安全报告，让安全驾驶员在车内监督测试过程，并安装“黑匣子”数据记录器，以便调查人员能够在问题出现时查明情况。

简而言之，现有的关于隐私、歧视、车辆安全等方面的规定必须加以调整，以便将人工智能考虑在内。

Regulatingartificial intelligence

As artificial intelligence (AI) is applied in a growing number of areas, there arelegitimateconcerns about possibleunintendedconsequences.

Theimmediateconcern is that thescrambletoamassthe data needed to train ai systems isinfringingon people’s privacy. Monitoring everything that people do online, from shopping to reading to posting on social media, lets internet giants build detailed personal profiles that can be used totargetadvertisements or recommend items of interest. The best response is not to regulate the use of AI directly, but instead to concentrate on the rules about how personal data can be gathered, processed and stored.

A second area of concern is that of “algorithmic bias”—the worry that when systems are trained using historical data, they will learn andperpetuatethe existing biases. To ensure fairness, AI systems need to be better at explaining how they reach decisions (an area of much research); and they should help humans make better decisions,rather thanmaking decisions for them.

A third area where AI is causing concern is inself-driving cars. Many companies are now testingautonomous vehiclesand running pilot “robotaxi” services on public roads. But such systems are not perfect, and in March 2018 a pedestrian was killed by an autonomous car in Tempe, Arizona—the firstfatalityof its kind. The right response is to require makers of autonomous vehicles to publish regular safety reports, put safety drivers in their cars tooverseethem during testing and install “black box” data recorders so that investigators can work out what happened if something goes wrong.

In short, existing rules on privacy, discrimination, vehicle safety and so on must beadaptedto take AI into account.

英文选自：The Economist

Language Points

regulate v.〔尤指通过规章〕控制，管理

strict rules regulating the use of chemicals in food

限制化学品在食物中使用的严格规定

legitimate a. 合理的，正当的

reasonable a. 合理的

Most scientists believe it is legitimateto use animals in medical research.

大多数科学家认为用动物作医学研究是正当的。

unintended a. 无心的，无意识的

unplanned, unexpected

the unintended consequences of human action人类行为无心的结果

immediate a. 最直接的，紧接着的

The mostimmediate effectof retirement is a dramatic reduction in living standards.

退休的最直接影响是生活水准的大幅度下降。

scramble v. n. 抢夺，争夺

Thousands of people will be scrambling for tickets.

将有数千人抢购门票。

amass v. 积聚，积累

collect v. 收集，聚集

Heamassedafortuneafter the war.

战后，他积累了大量财富。

infringe v.违反〔法律〕；侵犯〔他人权利〕

The film exploited his image and infringed his copyright.

该影片利用了他的肖像，侵犯了他的肖像权。

target v. 以...为目标

These products will target young people. 这些产品是面向年轻人的。

perpetuate v.使持续，使长久〔尤指不好的事物〕

an education system that perpetuates the divisions in our society

使我们的社会持续分化的一种教育制度

rather than 而不是

self-driving cars 自动驾驶汽车

autonomous vehicles 自动驾驶汽车

fatality n.〔事故或暴力袭击中的〕死亡

a 50% increase in the number of traffic fatalities

交通事故死亡人数增长50%

oversee v. 监管，监察，监督

A team leader was appointed to oversee the project.

任命了一位组长来监管该工程。

adapt v. （使）适应，（使）适合

These flowers arewell adaptedto harsh winters.

这些花非常适应严冬。