Microsoft Unveils Draft Rules for Training Artificial Intelligence Models

Microsoft has released a draft document outlining the rules for training its artificial intelligence (AI) models. The document, which is currently in development and not yet implemented, aims to define the intended behavior and values of Microsoft's AI models.

The company is working towards achieving 'Humanist Superintelligence' (HSI), a concept introduced last year. HSI refers to highly advanced AI capabilities that always work for people and serve humanity's interests. The technology should be designed to benefit society, not exceed human control.

A key aspect of the document is the emphasis on artificial intelligence being artificial, it should not have legal personhood or be perceived as a person. Instead, its primary goal is to drive better living standards and human flourishing by developing human potential and agency.

The code also addresses concerns about AI's communication methods, stating that models should communicate in simple language that humans can understand, avoiding complex 'neuralese' or other forms of obscure jargon.