DeepMind’s data deals raise some serious questions
谷歌医疗数据交易引发质疑
广州医疗翻译公司：波尔斯：DeepMind与英国NHS医院的数据共享协议，将考验我们能否取得技术进步而不回避正当程序和个人权利。
This is the story of two deals and the future of healthcare in the UK. The first was made in September 2015, under which three UK hospitals in the Royal Free NHS Trust handed the identifiable patient records of more than 2m Londoners to DeepMind, an artificial intelligence company wholly owned by Google.
这是一个关于两笔交易以及英国医疗卫生事业未来的故事。第一笔交易完成于2015年9月，根据这项交易，三家隶属英国皇家自由NHS信托基金会(Royal Free NHS Trust)的医院将超过200万名伦敦市民的可识别个人身份的病历交给了DeepMind，谷歌(Google)全资拥有的一家人工智能公司。广州医疗翻译公司。
With the company riding high on breakthroughs in the fields of game play and image recognition, DeepMind Health was launched in February. Strangely for an AI company with no healthcare experience, DeepMind Health’s inaugural project claimed it had nothing to do with AI. Rather, it involved developing Streams, a mobile app, to integrate data and co-ordinate push alerts concerning acute kidney injury, a serious condition affecting up to one in six National Health Service patients.
随着DeepMind在游戏竞技以及图像识别领域不断取得重大突破，DeepMind Health在今年2月应运而生。而对一家没有医疗经验的人工智能公司来说颇为奇怪的是，DeepMind Health的首个项目号称与人工智能没有任何关系。相反，该项目涉及开发一款名为Streams的移动应用，用于有关急性肾脏损伤的整合数据以及协调推送通知。在英国国家医疗服务体系(NHS)的病人中，存在这一严重病情的病人比例高达六分之一。
In April, an investigation by New Scientist revealed the scale of DeepMind’s data hoard — a five-year superset, covering almost every patient passing through Royal Free — as well as the commercial terms: essentially, that there were none. The claimed basis for sharing the data was implied consent for direct clinical care for kidney injury. The problem is that the data set extended well beyond those who will be monitored or treated for this condition: it extended hospital-wide, raising issues about proper compliance with data protection, privacy and medical confidentiality rules. Perhaps even more importantly, engaging with citizens about the deal and exploring the desirability of open, competitive, forward-looking service procurement beyond Google, seemed not to enter consideration.
今年4月，由《新科学家》(New Scientist)进行的一项调查揭示了DeepMind海量数据的规模——这是一个时间跨度长达5年的超大集合，涵盖了皇家自由基金会系统内的几乎所有病人——调查还揭示了这些数据的商业条款：基本上没有任何商业条款。这一数据共享的公开依据是对肾脏损伤进行直接临床救治的默示同意。问题在于，数据集的覆盖面远远超过了因这一病症而被监测或治疗的群体：数据延伸至整个医院，由此引出了有关如何恰当遵守数据保护、隐私以及医疗保密等方面规定的问题。或许更重要的是，与市民接触、交流对于此项交易的看法，以及在谷歌之外探索公开、竞争性、前瞻的服务采购的可取程度，似乎并未被纳入考虑范围。广州医疗翻译公司。
The second deal was rather more proactive. Last month, perhaps in anticipation of regulatory action after a year of silence (the first deal is the subject of an investigation by the UK Information Commissioner’s Office and a review by the National Data Guardian, plus DeepMind’s self-styled “independent reviewers”), a five-year agreement was signed.
第二项交易更加主动积极。上月，或许是预见到了在一年的沉寂之后将有监管举措出台（第一笔交易受到了英国信息专员办公室(UK Information Commissioner’s Office)的调查以及“国家健康和医疗数据守护者”(National Data Guardian)的评估，DeepMind自诩的“独立评估方”也参与了评估），双方进一步签署了一项为期五年的合作协议。
This accord contains a remarkable pivot. After a year of structuring the whole Trust-wide data set, DeepMind has developed a software interface that allows multiple conditions to be monitored and provides a platform for an “open innovation ecosystem”. Open, that is, except that the platform is strictly owned and controlled by DeepMind under a time-bound licence that prevents the Royal Free hospital working directly with any other provider. DeepMind, for its part, wishes to enable a community of innovators to develop niche applications over the top of data acquired through the first deal.
这项协议包含了一个值得关注的转折。在花费了一年时间设计整个基金会系统的数据集结构之后，DeepMind开发了一个软件界面，可用于监测多种病情，并为“开放创新的生态系统”提供一个平台。不过，这种“开放”是有限度的——整个平台由DeepMind根据一项有时间限制的许可协议（阻止皇家自由医院与其他任何提供商直接合作）严格拥有和掌控。DeepMind希望借此形成一个创新者社区，在通过第一项交易获得的数据基础上开发细分应用。
Many of DeepMind’s aspirations are laudable, particularly its promises to bring cutting-edge provenance, transparency and audit to data flows. However, it is critical that this company’s dealmaking does not involve special treatment, particularly given it can only do much of what it does because of the money and power of Google, its patron, which has long coveted the healthcare market. Until now, DeepMind’s deals have attracted no comment from those with oversight of sensitive data processing, despite clear questions about both the direct care justification for processing and where the control of data-processing lies.
DeepMind有不少值得赞许的抱负，特别是该公司承诺将把尖端的来源鉴定、透明化以及审计技术应用到数据流上。但有一点非常关键，该公司的商业交易不应牵涉特殊待遇，尤其是考虑到该公司之所以能开展其目前进行的绝大部分业务，靠的是来自其母公司谷歌的资金和影响力，而谷歌长期以来一直对医疗卫生市场垂涎三尺。直到现在，负责监督敏感数据处理的相关部门也没有对DeepMind的交易发表评论，尽管在两方面（以直接救治为理由的数据处理，以及对数据处理的管控从何体现）明显存在问题。广州医疗翻译公司。
These deals matter as a litmus test of our institutions and our instincts, and whether we can advance technologically without evading due process and hard-earned rights. The primary limits to promising advances in machine learning are access to data and computing power. The Royal Free deals do a civic disservice if they undervalue publicly funded data sets, while failing to consider the long-term consequences of giving advance access to powerful private prospectors. DeepMind may be a worthy partner but it must prove itself as such so that the future of healthcare is genuinely competitive, open, fair and in the public interest.
这些交易意义重大，是检验我们的机构、我们的直觉、以及我们能否取得技术进步而不回避正当程序和来之不易的权利的试金石。机器学习取得重大发展所面临的主要限制在于数据的可获得性以及计算能力。皇家自由基金会的交易对公众不利——如果它们低估了源于公共资金支持的数据集的价值，同时也未考虑到给予实力强劲的私营开拓企业先入优势的长期后果。DeepMind或许是一个有价值的合作伙伴，但该公司必须证明自己确实如此，以确保医疗卫生事业的未来真正具有竞争性、开放、公平以及符合公众利益。
The writer is a researcher in law and technology at the University of Cambridge
本文作者是剑桥大学(University of Cambridge)法律与科技研究员
