A global consortium of AI experts and data scientists has recently unveiled a comprehensive and voluntary framework to promote the safe development of artificial intelligence (AI) products.
The World Ethical Data Foundation, boasting a membership of 25,000 individuals, includes professionals from major technology companies such as Meta, Google, and Samsung.
The framework comprises an extensive checklist of 84 crucial questions developers should consider at the outset of any AI project.
The Foundation has also invited the general public to submit their questions for review. These questions will be considered carefully during the Foundation’s upcoming annual conference.
In alignment with the AI community’s preferences, the framework has been introduced through an open letter format, which has garnered the support and endorsement of hundreds of industry experts.
Understanding AI and Its Implications
AI is a field that empowers computers to mimic human actions and responses with remarkable precision. By processing vast amounts of data, computers can be trained to identify patterns, make predictions, solve problems, and learn from mistakes.
In addition to data, AI systems rely on algorithms, which are rules that guide the computer in performing tasks in the correct order.
The Emergence of the World Ethical Data Foundation
Established in 2018, the World Ethical Data Foundation is a non-profit global consortium that brings together professionals from the technology and academic sectors to explore and evaluate the development of new technologies.
The Foundation’s comprehensive questions for developers encompass crucial aspects such as preventing bias in AI products and addressing situations where a tool’s outcome violates the law.
Highlighting the importance of ethical AI usage, Yvette Cooper, the shadow home secretary, recently announced that the Labour Party intends to criminalise the deliberate use of AI tools for terrorist activities.
Recognising the significance of AI advancements, Prime Minister Rishi Sunak has appointed Ian Hogarth, a tech entrepreneur and AI investor, to spearhead an AI task force. Hogarth expressed his commitment to understanding the risks of frontier AI systems and holding responsible companies accountable.
Critical Considerations in the Framework
The framework developed by the World Ethical Data Foundation encompasses various essential aspects, including compliance with data protection laws in different jurisdictions, transparent AI interaction disclosure to users, and fair treatment of human workers involved in data input and tagging.
The comprehensive list of questions is divided into three main chapters: questions for individual developers, questions for collaborative team consideration, and questions for product testing.
Some notable questions from the 84-point checklist include:
- Do I feel compelled or pressured to incorporate data from unreliable sources?
- Is the team responsible for selecting the training data composed of individuals from diverse backgrounds and experiences to minimise bias?
- What are the intended applications of the trained model?
Addressing Challenges in AI Development
Vince Lynch, the founder of IV.AI and an advisor to the World Ethical Data Foundation board, emphasised the current “wild west stage” of AI development. He stated, “We’re in this Wild West stage, where it’s just kind of: ‘Chuck it out in the open and see how it goes’.” He believes that the emerging cracks in the AI landscape need to be addressed, particularly in areas such as intellectual property rights and the consideration of human rights.
Resolving issues related to copyright protection is another significant challenge. If a model has been trained using copyright-protected data, it may require retraining the entire model rather than simply removing the protected data. This process can be immensely costly, amounting to hundreds of millions of dollars in some cases.
Safe AI Development Other Voluntary Frameworks and Collaborative Efforts
Several other voluntary frameworks for the safe development of AI have been proposed globally. Margarethe Vestager, the EU’s Competition Commissioner, spearheads efforts to establish a voluntary code of conduct in collaboration with the US government. This code would require companies involved in AI development or utilisation to adhere to non-legally binding standards.
The World Ethical Data Foundation’s framework took three years to develop, ensuring sufficient data collection for its construction. The organisation temporarily paused its development in response to ethical concerns raised by its customers, emphasising that its AI capabilities are not utilised for decision-making purposes. All final decisions are left to the employers, demonstrating their commitment to upholding human judgment.
Transparency to users emerged as a critical focus in the Foundation’s framework. Co-founder Euan Cameron emphasised the importance of indicating when AI technology is involved in content creation, stating, “If anyone’s using AI, you can’t sneak it through the backdoor and pretend it was a human who created that content.” Transparent attribution of AI-generated content is essential to ensure trust and accuracy.
Conclusion
With the release of this voluntary framework, the World Ethical Data Foundation aims to promote the responsible and safe development of AI products. By raising pertinent questions and addressing critical considerations, this framework fosters awareness of potential challenges and encourages ethical practices within the AI community. As AI continues to advance and permeate various industries, ensuring the alignment of technological innovation with ethical principles remains paramount for a sustainable and inclusive future.
