Just ahead of an industry meeting at the White House, major generative artificial intelligence (AI) vendors made proposals for labeling content from their systems, according to the US government. Content generated by AI should be clearly recognizable as such so that it is no longer confused with human content. In addition, companies such as the ChatGPT maker OpenAI have committed to carefully testing their AI software before publication, shared the US Presidential Administration. Generative AI can generate text and images that look like they were made by humans.
US President Joe Biden had invited representatives of important AI providers to a meeting on Friday. He is working on a decree and a bipartisan bill to regulate this technology.
In this context, in addition to OpenAI, the rivals Anthropic and Inflection as well as the Alphabet subsidiary Google and the Facebook parent Meta want to develop a system with which any content created by an AI can be marked with a watermark. The companies also pledged to protect their users’ personal data, keep the technology free from prejudice and bias, and prevent it from being used to discriminate against minorities. Because if the data with which the AI systems have learned reflects prejudices against certain groups of people, the AI adopts these prejudices – because it is based on the data, not on criteria such as ethics or fairness.
Other commitments include developing AI programs for medical research or mitigating climate change. Due to the discussion about the risks, the AI development departments responsible for ensuring the safety and trustworthiness of this technology have come under the spotlight in recent months.