AI Ethics Goes Beyond Data Privacy & Bias

2022-10-16
关注

Why AI Ethics go Beyond Data Privacy and Bias
Illustration: © IoT For All

Artificial Intelligence (AI) and ethics are a strange combination to imagine together. Yet, here we are with cohorts of academic researchers and AI labs across the globe dedicated to the topic of emerging technologies. Or, more specifically, the real and negative impact that AI technology can have if left unchecked and created in the absence of ethics.

'It is necessary to truly demonstrate a commitment to AI ethics and AI alignment that is more than lip service but is truly building helpful and non-harmful AI tools and systems.' -Claire CarrollClick To Tweet

What is AI Alignment?

The trajectory of what is known as responsible AI is a nascent space. Ten years ago there was little or no voice representing this idea of responsibility in technology design and development. With some experiences of crises that had far-reaching implications, we are now seeing the area of “AI ethics” grow, however still not as quickly as the technology itself.

There are also teams of academics and professionals working on something called AI alignment. While in the same ballpark, AI ethics and AI alignment are not strictly the same thing. AI alignment describes the effort to ensure machine learning and advanced analytics tools and solutions do not get appropriated, leading to unintended consequences.

A reality that’s already here. 

Data Privacy & Bias Issues

As the idea of ethics and AI have moved into the mainstream, the concerns have mostly centered around data privacy and bias. Cambridge Analytica is the obvious example here for its unprecedented breach of personal data. Undoubtedly, these are two critical areas that need relentless focus. However, relinquishing critical decision-making to algorithms, particularly within public services, could have serious consequences. Essentially, AI ethicists are working on understanding where AI systems are being discriminatory and arriving at misleading decisions on individuals.

The AI alignment professionals are focusing on how societies are outsourcing more decision-making, especially in how resources are allocated, from school places to social welfare and AI systems. It should be noted that our policymakers and lawmakers do not understand these systems anywhere near well enough to legislate and regulate.

AI Ethics as a PR Strategy?

While AI ethicists and alignment researchers don’t see eye to eye on everything, they both share a disdain for “ethics designed by PR committees.” Jack Clark, co-founder of Anthropic, an AI safety and research company, shed some light on how moral concerns around AI are being treated in many cases in a recent tweet:

“A surprisingly large fraction of AI policy work at large technology companies is about doing ‘follow the birdie’ with government – getting them to look in one direction, and away from another area of tech progress,” said Clark.

This is not encouraging.

Yet, many AI ethicists and alignment professionals have stated how tech companies are treating the growing concerns around AI as a box-ticking exercise. The number one priority is to eliminate as far as possible the possibility of a PR disaster, rather than build AI capabilities founded on the principles of open collaboration with instruments for challenging the system. This disillusionment is so widespread that in 2021, Pew Research Center published a survey entitled, “Experts Doubt Ethical AI Design Will Be Broadly Adopted as the Norm Within The Next Decade.”

Our Limited Understanding of AI

Rather than being cynical about PR overall, there is a role this business practice can, and should be playing when it comes to AI ethics. There is a reason why Big Tech is not being held accountable for the AI tools being created that affect millions of people. And that is because there is limited public understanding of what is truly happening with AI.

Limited understanding comes from poor digital literacy. Irrespective of your socio-economic background, the majority of people don’t have a mental model to understand algorithms or where their data travels and resides. But when the citizenry gets behind things, big businesses sit up and take notice.

Think Apple and its App Tracking Transparency feature. Being able to afford Apple means you can afford to buy a more robust contract for data privacy. No one is claiming Apple did this exclusively out of the goodness of their hearts. There is a strong brand and revenue strategy behind this move. Furthermore, Apple was able to make this move due to their walled garden operating system. But Apple has caught the public’s attention. Not to mention, the public is increasingly interested in learning about how they can have better transparency about where their data is shared, who has access to it, and where it’s stored.

We are also seeing the European Union (EU) rigorously tackle AI ethics issues, including transparency, bias, and data protection. The EU introduced the Artificial Intelligence Act in April 2021. It is currently going through an evaluation and rewrite process. The act is expected to go into effect in early 2023. It is expected to have a “broad impact on the use of AI and machine learning for citizens and companies around the world.”

According to a presentation on the law by the European Commission, the EU’s governing body, the act seeks to codify the EU’s regulatory and legal framework for how AI is used. This framework “requires AI to be legally, ethically and technically robust, while respecting democratic values, human rights and the rule of law.”

A Commitment to Ethics

Technology companies that truly do want to build AI solutions that improve people’s lives will enhance their own sustainability by communicating with all stakeholders transparently. There is too much at stake for this to be relegated to a box-ticking exercise. It is necessary to truly demonstrate a commitment to AI ethics and AI alignment that is more than lip service but is truly building helpful and non-harmful AI tools and systems.

Tweet

Share

Share

Email

  • Artificial Intelligence
  • Connectivity
  • Data Analytics
  • Privacy
  • Security

  • Artificial Intelligence
  • Connectivity
  • Data Analytics
  • Privacy
  • Security

参考译文
人工智能伦理超越数据隐私和偏见
图示:© IoT For All --> 人工智能(AI)与伦理似乎很难想象被放在一起。然而,我们确实看到全球成群的学术研究者和AI实验室致力于探讨新兴技术。更具体地说,他们关注的是,如果AI技术在缺乏伦理约束的情况下发展,可能会带来哪些真实且负面的影响。“我们有必要真正体现出对AI伦理和AI对齐的承诺,这种承诺不能只是嘴上说说,而是要切实构建出有益且无害的AI工具和系统。” ——克莱尔·卡罗尔 点击推文分享 什么是AI对齐?目前所知的“负责任AI”轨迹仍处于起步阶段。十年前,在技术设计和开发中,几乎没有人提出这种责任意识。随着一些具有深远影响的危机事件的发生,如今“AI伦理”领域正在逐渐兴起,尽管其发展速度仍然赶不上技术本身的进步。此外,还有一批学术研究者和专业人士正在研究所谓的“AI对齐”。尽管AI伦理和AI对齐属于同一领域,但它们并不是完全相同的概念。AI对齐描述的是努力确保机器学习和高级分析工具不会被滥用,从而导致意外的后果。而这种现实已经存在。 数据隐私与偏见问题 随着AI伦理理念逐渐进入主流,大多数关注点集中在数据隐私和偏见问题上。剑桥分析公司(Cambridge Analytica)就是显而易见的例子,因为它造成了前所未有的个人数据泄露。毫无疑问,这两个领域是必须持续关注的重点。然而,将关键决策权交由算法决定,尤其是在公共服务领域,可能会带来严重后果。本质上,AI伦理研究者正在努力识别AI系统在何处产生了歧视行为,并导致对个体产生误导性判断。而AI对齐专业人员则专注于研究社会如何将越来越多的决策权外包给AI系统,尤其是在资源分配方面,比如学校录取和社保福利等。值得注意的是,我们的政策制定者和立法者对这些系统几乎一无所知,因此难以制定和实施有效的立法和监管。 AI伦理是否只是公关策略? 尽管AI伦理研究者和对齐研究者并非在所有方面都持相同观点,但他们共同对“由公关委员会制定的伦理”表示反感。Anthropic公司联合创始人杰克·克拉克(Jack Clark)最近在推文中分享了他对AI道德关切是如何被处理的看法:“令人惊讶的是,大型科技公司中相当大比例的AI政策工作,其实只是在做‘转移视线’的游戏——引导政府把注意力引向一个方向,而忽略另一个技术进展的领域。” 这种现象并不令人鼓舞。然而,许多AI伦理和对齐专业人士都指出,科技公司是如何将AI相关的担忧视为一种“完成任务”的形式。优先考虑的是尽可能减少公关灾难的可能性,而不是构建以开放协作原则为基础的AI能力,并配备挑战系统的方法。这种失望情绪如此普遍,以至于2021年皮尤研究中心发布了一项调查报告,标题为:“专家怀疑在未来十年内,伦理AI设计能够广泛被采纳成为常态。” 我们对AI的理解有限 与其对公关活动本身持怀疑态度,不如承认在AI伦理方面,这种商业实践确实可以而且应该发挥一定作用。大型科技公司之所以没有因所开发的AI工具——这些工具影响了数百万人的生活——而受到应有的问责,是有原因的。那就是公众对AI技术的真实运作缺乏理解。有限的理解源于数字素养的缺乏。无论你的社会经济背景如何,大多数人缺乏理解算法的思维模型,也不清楚自己的数据流向和存储位置。但一旦公众开始关注这些问题,大型企业就会高度重视。想想苹果公司推出的“App跟踪透明度”功能。 能买得起苹果产品的人,也就买得起更强的数据隐私保护合同。没有人会认为苹果公司是出于纯粹的善意才这么做。苹果背后有明确的品牌和收入战略。此外,苹果之所以能推出此功能,还得益于其封闭的生态系统。但苹果确实吸引了公众的注意。更不用说公众对了解数据透明度的兴趣日益增加,包括想知道自己的数据被共享到哪里、谁能够访问这些数据以及它们被存储在何处。 我们也看到欧盟(EU)正在积极应对AI伦理问题,包括透明度、偏见和数据保护等问题。欧盟于2021年4月推出了《人工智能法案》。目前该法案正在进行评估和修订。预计该法案将在2023年初生效,并对全球范围内的公民和企业使用AI和机器学习产生“广泛影响”。根据欧洲委员会——欧盟的管理机构——发布的一份关于该法律的介绍材料,该法案旨在将欧盟对AI使用的监管与法律框架制度化。这一框架要求AI在法律、伦理和技术上都具备稳健性,同时尊重民主价值观、人权和法治精神。 对伦理的承诺 如果科技公司真的希望构建能够改善人们生活的AI解决方案,它们就必须通过与所有利益相关者进行透明的沟通来提高自身的可持续性。鉴于此事关系重大,我们不能将其仅仅视为一种形式主义的任务。我们有必要真正体现出对AI伦理和AI对齐的承诺,这种承诺不应只是嘴上说说,而是要真正构建出有益且无害的AI工具和系统。推文 分享 分享 邮件 人工智能 连通性 数据分析 隐私 安全 --> 人工智能 连通性 数据分析 隐私 安全
您觉得本篇内容如何
评分

相关产品

Honeywell 霍尼韦尔 传感器评估套件 传感器评估套件

霍尼韦尔传感器评估套件(SEK001/SEK002)是一款专为加速传感器集成与验证而设计的开发工具。该套件将霍尼韦尔多种高性能传感器连接到符合行业标准的 Arduino™ Uno Rev3 微控制器面板,并配备专用的霍尼韦尔软件。用户无需编写任何代码,即可通过个人电脑实时查看传感器测量数据(如压力、湿度、颗粒物浓度等),并将数据记录为 .csv 文件以便进一步分析。该套件旨在简化传感器评估流程,帮助客户在更短的时间内做出明智的部件选型决策,从而加快产品开发进程。

Yanmade 燕麦科技 MEMS传感器测试设备 MEMS传感器测试

适用于用于温度传感器、湿度传感器和压力计的产品测样、小批量生产及性能验证等应用场景; 设备内置智能数据分析模块,可对测试数据进行自动处理、分析,生成测试报告。

微著科技 高性能传感器ASIC解决方案 MEMS传感器

微著科技是国内为数不多能够给传感器厂商提供定制高性能传感器解决方案的团队,目前已为国内众多院所及知名传感器公司提供了十余个传感器解决方案并已经实现量产。微著传感器ASIC方案的特点:成熟的仪表信号模块IP易于快速搭建;系统方案超低噪声;成熟的24ADC可同时实现模拟数字传感器方案设计;高效率及丰富的方案设计经验。

HOTO 汇投智控 HT6204 多合一传感器

多合一传感器可以同时测量多种气体指标,包括激光粉尘传感器、CO2传感器、TVOC传感器和温湿度传感器,多种数据以数字接口的形式统一输出。

EC Sense 爱氪森科技 MP6 四气体传感器模组 MP6 四气体传感器模组

• 灵活的传感器适配性• 模块化的结构设计• 小体积• 传感器工作状态自动判断• 5-24V DC宽范围供电• 同时采集气体和温湿度传感器的测量数据• DS4氧气和毒气传感器寿命自动监测

方舟 位移传感器 传感器产品

公司坐拥 10000 平方米研发制造基地,现有员工 400 余人,长期专注于柔性集成电路及各类柔性传感器的研发、生产与系统化解决方案。

Honeywell 霍尼韦尔智能工业 在线/便携烟气分析仪专用传感器 气体传感器

霍尼韦尔智能工业在线/便携烟气分析仪专用气体传感器,采用高精度传感技术,专为烟气成分检测与燃烧过程分析设计。该传感器支持在线连续监测与便携式设备快速部署,具备响应迅速、长期稳定性高、抗干扰能力强等核心特性,广泛应用于工业排放监测与燃烧优化控制场景。

Huba Control 富巴 680系列 压力传感器

680系列压力传感器采用压电电阻测量芯片,产生经过补偿校准和放大的传感器信号,适用于标准化电压或电流输出。多种压力和电气接头可供选择。此传感器采用不锈钢材质。

CNSCAN 天策电子 足底压力分布检测系统 柔性压力传感器 分布式薄膜传感器 分布式薄膜压力传感器

CNSCAN天策电子足底压力分布检测系统是一款基于柔性薄膜压力传感器的无线步态分析设备。系统采用自主研发的足底专用采集器(FPD)与上位机软件配合,通过放置在鞋内或地面的柔性传感器实时感知足底动态压力。

评论

您需要登录才可以回复|注册

提交评论

广告
广告

iotforall

这家伙很懒,什么描述也没留下

关注

点击进入下一篇

XDR级MiniLED电视海信E8H 6999元起开启预售

提取码
复制提取码
点击跳转至百度网盘