MIT has just released an important document for unions, government, public and student institutions on the risks associated with the use of AI and categorization of potential hazards.
The AI Risk Repository is a comprehensive database of hundreds of documented risks posed by AI systems. The repository is designed to help decision makers in government, research, and industry assess the evolving risks posed by AI.
As the guys from MIT put it: “A comprehensive, live database of over 700 AI risks, categorized by their causes and risk domain.”
The AI Risk Repository consists of three parts:
- AI Risk Database Covers over 700 risks extracted from 43 existing systems, with citations and page numbers.
- Causal Taxonomy of AI Risks classifies how, when and why these risks arise.
- Domain Taxonomy Risks of AI classifies these risks into seven domains (e.g., “Disinformation”) and 23 subdomains (e.g., “False or Misleading Information”).
The AI Risk Repository provides:
- An accessible overview of the AI risk landscape.
- A regularly updated source of information on new risks and research.
- A common framework for researchers, developers, companies, evaluators, auditors, policymakers and regulators.
- A resource to help develop research, curricula, audits and policy.
- An easy way to find relevant risks and research.
“We started our project with the goal of understanding how organizations respond to AI risks,” Peter Slattery, a new postdoc at MIT FutureTech and the project’s lead, told VentureBeat. “We wanted to get a comprehensive view of AI risks that could be used as a checklist, but when we reviewed the literature, we found that existing risk classifications were like puzzle pieces: individually interesting and useful, but incomplete.”
The repository uses a two-dimensional classification system. First, risks are classified based on their causes, taking into account the responsible party (human or AI), intent (intentional or unintentional), and when the risk occurs (before or after deployment). This causal taxonomy helps to understand the circumstances and mechanisms through which AI risks may arise.
Second, risks are divided into seven distinct areas, including discrimination and toxicity, privacy and security, disinformation and bad actors, and misuse.
The AI risk repository is designed as a dynamic database. It is publicly available and organizations can download it for personal use. The research team plans to regularly update the database with new risks, research discoveries, and evolving trends.
Source: Digital Trends
I am Garth Carter and I work at Gadget Onus. I have specialized in writing for the Hot News section, focusing on topics that are trending and highly relevant to readers. My passion is to present news stories accurately, in an engaging manner that captures the attention of my audience.
