2021年负责任的人工智能报告（英文）

The State of Responsible AI: 2021How 100 C-level AI Leaders are Ensuring AI is Used Ethically, Transparently, Securely and in Their Customers’ Best InterestsContents3Executive Summary9What Executives Believe About AI Ethics1418How Executives are Trying to Ensure AI is Used ResponsiblyEnterprises Must Establish Practices for Responsible AIClick below to navigate5Methodology10Combating AI Model Bias1519The Ethical Implications of Adversarial AI Attacks6The True Extent of the AI Ethics ProblemPlotting a Path to Responsible AI 2The State of Responsible AI: 2021Executive Summaryhe age of AI-enabled business is upon us. COVID-19 accelerated many companies’ need to rapidly digitize and enhance their business processes with AI. However, with new technology, comes new responsibilities.This global survey of 100 C-level data and analytics leaders reveals the complex issues AI-focused executives are considering and tackling to prepare their organizations to be AI-enabled in an ethical way.Our findings highlight the true extent of the AI ethics problem. AI-focused executives may be paying more attention to questions around AI ethics. But there’s still no consensus about what responsibilities enterprises have around AI use. This research also explores how effectively enterprises are combatting AI model bias during the development process. Few enterprises have the kind of ‘ethics by design’ approach in place that would ensure they routinely test for and correct AI bias. But our findings paint a picture of what current industry best practices are.We also explore the ethical implications for enterprises with poor AI security measures in place. While there’s still a lack of awareness around this topic, this study shows that AI-focused leaders are taking steps to ensure their models are safe from malicious attacks.Crucially, our findings suggest a path forward for executives who want to ensure their enterprises use AI responsibly. Enterprises can start to address the challenges they are facing today by 1) understanding the risks associated with AI, 2) developing better practices to ensure regulatory compliance and 3) placing greater emphasis on ongoing model monitoring and maintenance. T3The State of Responsible AI: 2021say their enterprise has an AI ethics board to consider questions on AI ethics and fairness. The 78% that don’t are poorly equipped to ensure the ethical implications of using new AI systems are considered properly say the processes they have to ensure AI projects comply with regulations are ‘ineffective’, with 46% describing them as ‘very ineffective’. This points to a need for greater investment in this area in most enterprisesagree that inefficient processes for model monitoring represent a barrier to AI adoption in their enterprises. This suggests that almost all enterprises could benefit from improving their model operationalization processes68%90%Source: Corinium Intelligence, 202178%are finding it hard to secure executive support