小程序
传感搜
传感圈

7 Biggest Barriers to AI Adoption & Their Solutions

2022-07-13
关注



We have seen how COVID-19 mounted pressure on businesses to fast-track their digital transformation journeys by months and, in several cases, by years. The arrival of the pandemic made them reconsider technologies they had at their fingertips – artificial intelligence (AI) in particular – and harness them to boost productivity, address supply chain issues, and seamlessly deliver products and services. Organizations have realized the indispensability of integrating AI into their digital strategy and this article will focus on addressing common AI adoption challenges.


'One of the most critical barriers to profitable AI adoption is the poor quality of data used. Any AI application is only as smart as the information it can access.' -Damco SolutionsClick To Tweet

Artificial Intelligence is a revolutionary technology that saves time, energy, and money. It is no longer confined to science textbooks or science-fiction fantasies; it has countless applications in the real world. Businesses are now acknowledging the importance of implementing this futuristic technology. In fact, a high-level penetration of machine intelligence can resolve fundamental problems.

A McKinsey survey illustrates that AI adoption followed an upward trajectory in the year 2021 and continues to do so. It states, “56 percent of all respondents report AI adoption in at least one function, up from 50 percent in 2020.”

While businesses have realized that AI adoption is the way forward, it’s not always easy. So, what are the critical barriers that prevent businesses from realizing the tremendous potential of this next-gen technology? Let’s discuss these AI adoption challenges one by one. 

Ethical Considerations 

The first challenge to AI adoption is how ethics becomes a pressing concern as organizations integrate AI with more processes. AI brings seemingly scientific credence to human biases and tends to amplify them, putting its potential for decision-making doubtful. Fortunately, we have a solution.

A promising sign is the growing awareness of this issue and acknowledging AI’s potential for bias is the first step. As businesses train their AI/ML models, they must actively counter prejudiced data and specifically program their AI to be unbiased. Also, annotators must carefully analyze training data before it is fed into the algorithm. This way, it won’t lead to biased conclusions. 

Poor Data Quality

One of the most critical barriers to profitable AI adoption is the poor quality of data used. Any AI application is only as smart as the information it can access. Irrelevant or inaccurately labeled datasets may prevent the application from working as effectively as it should. 

Many organizations collect far too much data. It may be full of inconsistencies and redundancies, resulting in data decay. Data quality can be improved by streamlining the collection process. The stakeholders must pay more attention to data cleansing, labeling, and warehousing. These workflow changes can provide businesses with high-quality data. 

Data Governance 

In the face of rising cybercrime, responsible data governance is more vital than ever. People are concerned about how companies access and use their confidential information, so it’s important that organizations leveraging customer-facing AI hold themselves accountable when deploying applications.

The key here is segmentation and visibility. Organizations have to ensure that they can monitor and restrict how their AI algorithms use data at all stages. Segmentation mitigates the impact of a breach and keeps user information as safe as possible. Likewise, transparent data collection policies can also help alleviate concerns related to AI. 

Process Deficiencies

Companies often use in-house tools and pipelines for AI deployment and monitoring. Constructing an efficient AI model from scratch demands a significant expenditure of time and money. So, if you have just started, AI adoption can cost you exorbitantly. Also, your tool may include inappropriate algorithms and biased data. In such a scenario, adopting third-party tools for AI integration or using a market-tested tool is a comparatively sensible option. 

Cybersecurity

AI implementation introduces cybersecurity risks. Many data breaches have already occurred in an effort to collect data for AI initiatives. Protecting stored data from malware and hackers should, therefore, be a top priority for companies. A robust cybersecurity defense approach can help prevent such attacks. Besides, AI adoption leaders need to acknowledge the growing menace of sophisticated threats and evolve from a reactive to a proactive strategy. 

Storage Limitations

Training the AI/ML models demands constant amounts of high-quality labeled datasets. Organizations, therefore, need to feed massive volumes of data into the machine learning algorithms, so that they can perform the desired activities and deliver reliable results.

This has become challenging as traditional storage technologies are quite expensive and come with space limitations. However, recent technological breakthroughs like flash storage seem to provide a solution. Unlike traditional hard disks that are expensive, flash storage is more reliable and affordable. 

Regulatory Compliance

AI and other data-centric operations increasingly face legal regulations with their rising prominence. Organizations have to abide by these restrictions, especially if they operate in highly regulated industries such as finance and healthcare. 

Taking a flexible approach toward maintaining high privacy and governance standards can help these companies be more legally compliant. Third-party auditors are more likely to be in demand owing to increased regulations.

The Way Forward

AI is gradually becoming a game-changer and its potential is up for grabs. A PwC study states “AI could contribute up to $15.7 trillion to the global economy in 2030, more than the current output of China and India combined. Of this, $6.6 trillion is likely to come from increased productivity and $9.1 trillion is likely to come from consumption-side effects.”

But what can make AI work for companies? Anticipating the barriers to AI adoption and taking a strategic approach to its implementation can help organizations achieve transformational growth and maximize returns.



Tweet

Share

Share

Email

  • Artificial Intelligence

  • AIoT

  • Cybersecurity

  • Data Analytics

  • Machine Learning


  • Artificial Intelligence

  • AIoT

  • Cybersecurity

  • Data Analytics

  • Machine Learning

参考译文
人工智能采用的7大障碍及其解决方案

我们看到,2019冠状病毒病给企业带来了压力,迫使它们数月甚至数年地加快数字化转型进程。大流行的到来让他们重新考虑他们触手可及的技术,特别是人工智能(AI),并利用它们来提高生产力,解决供应链问题,无缝交付产品和服务。组织已经意识到将AI整合到他们的数字战略中是不可缺少的,这篇文章将专注于解决常见的AI采用挑战。人工智能是一项革命性的技术,可以节省时间、精力和金钱。它不再局限于科学教科书或科幻小说;它在现实世界中有无数的应用。企业现在已经意识到实施这种未来技术的重要性。事实上,机器智能的高层次渗透可以解决基本问题。麦肯锡(McKinsey)的一项调查显示,人工智能的应用在2021年呈上升趋势,并将继续上升。报告称,“56%的受访者表示,人工智能在至少一种功能上的应用,高于2020年的50%。”虽然企业已经意识到采用人工智能是前进的方向,但这并不总是容易的。那么,阻碍企业实现这种新一代技术的巨大潜力的关键障碍是什么?让我们来逐一讨论这些人工智能应用面临的挑战。采用人工智能的第一个挑战是,当组织将人工智能与更多流程整合在一起时,伦理如何成为一个紧迫的问题。人工智能给人类的偏见带来了看似科学的可信性,并倾向于放大它们,使其做出决策的潜力令人怀疑。幸运的是,我们有一个解决方案。一个有希望的迹象是,人们对这一问题的认识越来越多,承认人工智能可能存在偏见是第一步。当企业训练他们的AI/ML模型时,他们必须积极地对抗有偏见的数据,并专门编程使其AI不带偏见。此外,注释器在将训练数据输入算法之前必须仔细分析这些数据。这样,就不会得出有偏见的结论。人工智能采用的最关键的盈利障碍之一是使用的数据质量较差。任何人工智能应用程序的智能程度取决于它能够访问的信息。不相关或标记不准确的数据集可能会阻止应用程序有效地工作。许多组织收集了太多的数据。它可能充满了不一致和冗余,导致数据衰减。可以通过简化收集过程来提高数据质量。涉众必须更加注意数据清理、标记和仓库。这些工作流更改可以为业务提供高质量的数据。面对不断上升的网络犯罪,负责任的数据治理比以往任何时候都更加重要。人们关心公司如何访问和使用他们的机密信息,所以利用面向客户的人工智能的组织在部署应用程序时对自己负责是很重要的。这里的关键是细分和可见性。组织必须确保他们能够监控和限制他们的AI算法在所有阶段如何使用数据。细分可以减轻信息泄露的影响,并尽可能保证用户信息的安全。同样,透明的数据收集政策也可以帮助缓解与人工智能相关的担忧。公司经常使用内部工具和管道来部署和监控AI。从头开始构建一个高效的人工智能模型需要大量的时间和金钱支出。所以,如果你刚刚开始,人工智能可能会让你付出高昂的代价。此外,您的工具可能包括不适当的算法和有偏见的数据。在这种情况下,采用第三方的人工智能集成工具或使用经过市场测试的工具是比较明智的选择。 实施人工智能会带来网络安全风险。为了为人工智能计划收集数据,已经发生了许多数据泄露事件。因此,保护存储的数据不受恶意软件和黑客的侵害应该是公司的首要任务。一个强大的网络安全防御方法可以帮助防止此类攻击。此外,人工智能应用的领导者需要认识到复杂威胁的日益增长的威胁,并从被动的策略演变为主动的策略。训练AI/ML模型需要恒定数量的高质量标记数据集。因此,组织需要向机器学习算法输入大量数据,以便它们能够执行所需的活动,并提供可靠的结果。这已经成为挑战,因为传统存储技术相当昂贵,并伴有空间限制。然而,最近的技术突破,如闪存,似乎提供了一个解决方案。与昂贵的传统硬盘不同,闪存更可靠,更实惠。随着人工智能和其他以数据为中心的业务日益突出,它们越来越多地面临法律监管。组织必须遵守这些限制,特别是如果他们在金融和医疗等高度监管的行业中运营。采取灵活的方法来保持高度的隐私和治理标准可以帮助这些公司在法律上更合规。由于监管的加强,第三方审计师的需求可能会更大。人工智能正逐渐成为游戏规则的改变者,它的潜力有待开发。普华永道的一项研究指出:“到2030年,人工智能对全球经济的贡献可能高达15.7万亿美元,超过中国和印度目前的产出总和。”其中,6.6万亿美元可能来自生产率的提高,9.1万亿美元可能来自消费方面的影响。“但什么能让人工智能为企业服务呢?预见到人工智能采用的障碍,并采取战略方法来实施它,可以帮助组织实现转型增长和最大化回报。

您觉得本篇内容如何
评分

评论

您需要登录才可以回复|注册

提交评论

iotforall

这家伙很懒,什么描述也没留下

关注

点击进入下一篇

2022广东佛山国际工业互联网及工业通讯展览会

提取码
复制提取码
点击跳转至百度网盘