据报道, 谷歌在2019年后离开项目 Maven 军事 AI 计划
The company’s contract is set to expire next year, and it involves helping the government use machine learning to analyze drone footage
以下内容由机器翻译生成。如果您觉得可读性不好, 请阅读原文或 点击这里.
Google has decided not to renew its contact for a controversial Department of Defense program known as Project Maven, which involved Google helping the US government analyze drone footage using artificial intelligence.
The news, reported today by gizmodo, was announced internally today by Google Cloud chief Diane Greene at a meeting with employees. The contract allegedly expires in 2019. Greene’s announcement comes just days after news that Google will be announcing a new set of ethical principles to guide the development of military AI, which is set to be revealed next week. Google was not immediately available for comment.
The decision should calm unrest at Google that has been growing for months since the company’s involvement with Project Maven was revealed in March in an investigation by gizmodo. The report noted at the time how employees were “outraged” the company was using them and the software they helped develop to aid a government program that involves the killing of human beings overseas using unmanned aerial vehicles.
Google apparently viewed the contract, which basically amounted to providing the government with open-source software like Google’s TensorFlow platform, as a way to secure more lucrative Pentagon contracts, gizmodo 报告。 Leaked emails obtained recently by 拦截 show that although Google downplayed the importance of its involvement in Project Maven, it expected revenue from similar military AI projects to eventually grow from $15 million per year to $250 million.
Although Project Maven did not involve using Google software to pilot or actively participate in any form of drone warfare, Google was nonetheless providing tools and expertise in analyzing footage. Ostensibly, Project Maven may have been aimed at making drone strikes and other forms of drone warfare more civilian-friendly. Computer vision algorithms trained using AI techniques, like machine learning, could theoretically better identify targets and reduce collateral damage, though the full extent of Google’s involvement in Project Maven was never made clear. “The technology is used to flag images for human review and is intended to save lives and save people from having to do highly tedious work,” Google told 《纽约时报》 in a statement back in April.
However, any use — no matter what the ultimate aim — of AI to aid or otherwise augment warfare or government weaponry was met with fierce backlash from both Google employees and technology critics who see automated warfare as a dangerous slippery slope with deeply complex ethical implications. In early April, more than 3,000 Google employees signed an open letter to Google CEO Sundar Pichai, advising the company to pull out of the program, writing, “Google should not be in the business of war.”
“This contract puts Google’s reputation at risk and stands in direct opposition to our core values,” the letter read. “Building this technology to assist the US government in military surveillance — and potentially lethal outcomes — is not acceptable.” Earlier this month, gizmodo 报告 that around a dozen employees resigned over Google’s inaction on the subject and its continued involvement in Project Maven.
Now, it appears the pressure put on Google from both within its own ranks and outside has ultimately forced top leadership to take a stand. According to gizmodo, Greene told employees that the backlash was the primary reason the company has decided not to renew its contract. Coincidentally, on May 18th, Google updated its code of conduct to remove prominent references to the company’s old motto, “Don’t be evil,” which has since been replaced with, “Do the right thing.”
Update 6/1, 3:35PM ET: Added additional information about Google’s involvement with Project Maven and its plans for additional Pentagon contracts related to AI expertise.