忘了人工智能吧，真正的威胁是“人工智障”

思科公司安全业务主管大卫•尤勒维什不喜欢对人工智能的炒作，这个领域涉及一门蒸蒸日上的计算机科学，那就是机器学习。

突然之间，所有科技界人士都在大谈这个话题。营销人员，特别是来自网络安全行业的那些，已经开始把它比作灵丹妙药；按他们的话说，没有什么网络是AI保护不了的（另外，也有人警告说超级智能可能毁了这个世界，就像终结者那样）。

不久前，尤勒维什在RSA安全会议上发表主题演讲前对《财富》杂志表示：“安全领域的真正威胁并非是否存在过于聪明的东西。我更担心‘人工智障’。”尤勒维什计划在他的演讲中谈到这个话题。

无论什么时候，别人抛出算法、分析论、AI这些时髦词语的时候尤勒维什都会选择无视。他说：“那就跟说自己有Wi-Fi一样。这并不表示你的安全措施很好。”

用尤勒维什的话来说，网络安全产品最有吸引力的特点是“盒子外的整合”。太多的网络安全公司都只提供API，即应用程序接口，却把将API和现有防御机制衔接在一起的重担留给了消费者。

在尤勒维什看来，这样做还不够好。新工具应立即兼容它们要保护的网络。他说：“API只代表可能性。只有很少一部分公司会用到它们。”

尤勒维什认为解决之道在于整合。这样做可以减轻IT经理和安全专业人员的负担，而且这些人也没有协调一切的资源和时间。

尤勒维什介绍说，对整合型产品的偏好一直是思科安全策略背后的一个影响因素，这也解释了思科拥有众多合作伙伴以及频繁收购的原因（实际上，两年前思科斥资6.35亿美元收购了他的初创公司OpenDNS后，尤勒维什才进入了思科）。

尤勒维什用一件日常事物来解释他对数字威胁的看法。他说，大家可能会用漂亮的尖头栅栏来保护自己的院子，但除非你加固它，否则栏杆之间还是会有很大的洞。

他指出：“大家需要的东西要非常牢固，难以对付、形成整体而且紧密相连。”

这才是聪明的做法，而且没有那么多的人工智能色彩。（财富中文网）

译者：Charlie

David Ulevitch, head of Cisco's security business, has a problem with the hype around artificial intelligence, the field that encompasses the booming computer science discipline of machine learning.

Suddenly, everyone in the tech world is buzzing about the subject. Marketers, especially ones in the cybersecurity industry, have started plugging it as something of a panacea; there's no network AI can't protect, if the salespeople are to be believed. (Alternatively, others warn that a super-intelligence could destroy the world, Terminator-style.)

"The real threat in security is not whether we have things that are too smart," Ulevitch told Fortune in advance of his Wednesday RSA security conference keynote, during which he planned to touch on the topic. "I'm more concerned about 'Artificial Stupidity.'"

Whenever someone starts dropping the latest buzzwords—algorithms, analytics, AI—Ulevitch tunes them out, he said. "That's like saying you have Wi-Fi," he said. "That doesn't tell me that your security is good."

For Ulevitch, one of the most compelling features for a cybersecurity product is what he calls "integration out of the box." Too many cybersecurity companies simply offer APIs, or application programming interfaces, which puts the burden on customers to connect them to their existing defenses.

That's not good enough, in Ulevitch's view. New tools should be instantly compatible with the networks they're designed to protect. "APIs just represent potential," he said. "Only a small fraction of companies deal with them."

Integrations are the solution, according to Ulevitch. They take the load off the IT managers and security pros, who don't have the resources and time to sync everything up themselves.

This preference for integrated products has been a driving factor behind Cisco's (csco) security strategy, helping to explain its vast array of partnerships and frequent acquisitions, per Ulevitch. (In fact, he landed at Cisco after the company purchased his startup OpenDNS for $635 million two years ago.)

Ulevitch used a domestic analogy to explain his view of the digital threat landscape. You might have a nice picket fence protecting your yard, he said, but unless you reinforce the barrier, sizable holes still exist between the pickets.

"You need something much more firm, formidable, holistic, and tied together," he said.

That's the smart approach—and no artificially so.