ACM Publishes Principles of Generative AI Development

The Association for Computing Machinery today announced to providers that Generation AI The system and the developers who use it employ technology safeguards and statement It outlines a list of principles for its use and development.

According to the ACM, generative AI poses serious risks to society and current guidelines for safe use of technology are insufficient to ensure that generative AI does not cause significant harm. said the group.

Today’s statement is based on the following longer document: Principles of development, deployment and use of generative AI technology ACM issued in June. Founded in 1947, ACM claims to be the world’s largest scientific and educational computing association with approximately 110,000 student and professional members.

In its statement on artificial intelligence, the group said that the legal and regulatory schemes should be reviewed in light of the particular dangers posed by the abuse of generative AI, and that without substantial human oversight and careful deliberation He argued that “high-risk” AI should not be operated. As to whether the benefits of such a system outweigh the potential negative impacts.

ACM argues that intellectual property rules should be reviewed

Intellectual property ownership rules also need to be reviewed to resolve key issues regarding ownership of works created by AI and the legal status of data used to train AI models, according to the ACM. It is said that there is Personal data used to generate information or train models should be subject to an opt-out policy, and AI creators should ensure that the data generated by the system is as transparent as possible about its accuracy and correctability. A log of errors should be maintained.

However, today’s declaration did not lead to the adoption of any specific legislation. ACM US Technology Policy Committee Chairman Jeremy Epstein said the group was bipartisan and not a lobbying group.

“We respond directly to legislative requests for comment, but we do not take public positions on the bill or its components,” he said.

ACM also takes no special stance on copyright law issues that apply to AI training data. Several class action lawsuits have been filed in recent weeks For AI creators like Meta and OpenAI. Epstein called on the public to consider the matter carefully, and said the matter would likely be resolved through litigation.

“Creating changes to Section 107 of the Copyright Act has historically been and is likely to continue to be very difficult, so ultimately the courts will have to decide this,” he said. rice field.

The ACM statement also goes on to provide more general guidelines for technology development, including auditability and transparency, minimizing environmental impact, and strict attention to privacy and security issues that may arise. I reaffirmed the principles.