A group of experts on artificial intelligence (AI) has published a framework aimed at encouraging the release of what it describes as “more ethical” models of the technology.

The World Ethical Data Foundation (WEDF) is a non-profit group that includes people from academia, as well as experts working in the technology sector.

Its ‘open suggestion’ sets out lists of questions that individual developers and groups of developers should ask themselves while working on AI projects.

It has also set up a free online forum to allow members of the public to make their own suggestions.

Categories

“We are here to set an actionable standard every team can use immediately that also clarifies how AI is built, so we can reduce silos and discuss as a group,” the group says.

The questions are divided into three categories, covering the training, building and testing stages of AI systems.

They are further divided into questions for three different sets of individuals or groups:

• ‘Me’ questions that each individual working on AI should ask themselves before they stary working on a project,
• ‘We’ questions that groups should ask themselves and “define the diversity required to reduce as much human bias as possible”, and
• ‘It’ questions that should be asked about the model being created and the impact it could have.

Risk

The WEDF describes itself as “an organisation devoted to the impartial and balanced exploration of the urgent questions around the use and future of data and information”.

It says that, while AI has the potential to bring about significant benefits, it can also force human moderators to witness horrible content, disadvantage the disenfranchised, amplify data privacy and safety concerns, steal intellectual property not created for the process, and perpetuate biases that cause harm.

“Without careful consideration and planning, we risk creating AI systems that worsen the very problems we seek to solve,” its open letter states.