Sixteen AI firms have unveiled fresh commitments to ensure the safe development of Artificial Intelligence, according to the UK government.
These commitments seek to ensure that the world’s top AI companies demonstrate transparency and accountability in their plans towards a safe AI, British Prime Minister, Rishi Sunak said in a statement released on Tuesday by the UK’s Department for Science, Innovation and Technology.
The agreement was signed by companies such as OpenAI (ChatGPT), Google DeepMind, and Anthropic. It was based on a consensus reached at the first Global AI Safety Summit, held at Bletchley Park, UK, last year.
At the second summit, in Seoul - co-hosted by the South Korean and British governments - companies yet to publicise how they appraise the safety of their technology have committed to doing so.
They have agreed to identify “intolerable” risks and what they would do to ensure that such thresholds are not breached.
In the most extreme circumstances, the companies are to refrain from developing or deploying any models or systems if mitigation measures do not enable them to keep the risks below set thresholds.
These thresholds are to be defined before the next AI summit, to be held in 2025 in France.
Tech giants Microsoft, Amazon, IBM and Meta from the USA, Mistral AI from France, and Zhipu.ai from China are among those that have accepted the new safety rules.
The unprecedented success of ChatGPT shortly after its 2022 launch triggered a rush into the field of generative AI. Tech companies worldwide are investing billions to develop their own models.
Generative AI models can produce text, images, sound, and even videos from simple prompts. Supporters see them as breakthroughs that will enhance the lives of citizens and businesses globally.
However, human rights activists and governments fear their misuse in many situations, including manipulating voters via fake news or “deepfake” images and videos of political leaders.
Many are calling for international standards to govern the development and use of AI.
