AI systems should be transparent and explainable: IBM

Multinational information technology company IBM has released a new regulatory framework for AI (artificial intelligence). 

IBM policy lab’s new ‘precision regulation’ guidelines include hiring an AI ethical lead and providing transparency in how AI systems functions.

The guidelines come at a time when the new European Commission said it will legislate on AI within the first 100 days of 2020. The United States has also released new guidelines for the regulation of AI on January 8. 

According to the company, the guidelines bring more accountability, fairness and security into the AI systems. 

The framework for AI systems come months after the Big Blue released precision regulation guidelines for facial recognition and illegal online content in November last year.    

The basic notion of precision regulation is to hold companies more accountable but not hinder innovation and ethics in the larger digital economy. 

IBM's policy paper recommends a designated AI ethics lead to ensure responsibility for trustworthy AI implementations.

 Secondly, not all AI systems can be treated equally, considering different use cases carry different risks. The paper also stresses on an initial high-level assessment regarding the technology’s potential for harm.

The guidelines also prioritise transparency. According to the company, the best way to promote transparency is through disclosure of the purpose of the AI system to consumers and businesses.

Any AI system in the market should be required to provide a context and an explanation for its decision, IBM said. 

AI systems need to be tested for bias, according to the recommendations. This includes all individuals in the AI development lifecycle having a level of shared responsibility in ensuring the systems are deployed in a fair manner.

A previous study by the Morning Consult study also found that 70% of Europeans and 60% of Americans supported AI regulation in some form. Additionally, 74% of American and 85% of EU respondents also insisted that artificial intelligence systems required to be transparent and explainable.

Read: Not deploying AI can put firms out of business by 2025, say 80% C-level execs: Accenture