Data and AI Learning Group

 View Only

How Is IBM Leading To Transparent Artificial Intelligence?

By Silver Hawak posted Fri April 09, 2021 12:48 PM


The humans are biased, such as anchoring bias with which the decisions rely on the first piece of information; confirmation bias which focuses on information that ensures the preconceptions about specific topics; and gender bias which means women are associated with men in terms of traits and professions. These biases can impact decision-making capacity. 

Similarly, these different types of bias can be found in artificial intelligence when machine learning techniques are implemented for programming AI systems. Supervised machine learning is one machine such technique which means the AI systems must be trained with specific problems and solutions. That being said, AI systems can identify the correlations in examples and make decisions. 

There are algorithms for outlining such biases, detecting them, and mitigating them. Still, the AI biases are complicated and have different types of data, and all of them require different technologies for outlining and mitigating the bias. This is an ever-growing challenge, and IBM has played an influential role in making the AI transparent, fair, and accurate, such as;

AI Ethics Board 

The ethical considerations regarding different technologies have become a priority. To ensure lasting and epic changes on the issues, the organizations must support the cultural change. IBM has the AI governance frameworks with a centralized approach and focuses on the AI ethics board. It can support the non-technical and technical initiatives for ensuring transparency and trust (the IBM principles). 

Defining The Policies Around AI 

IBM has launched the Principles for Trust & Transparency, which guides the policy approach to artificial intelligence in ways that encourage responsibility. These principles can identify the commitment to using AI for augmenting human intelligence. In addition, the data policy protects the insights and data of the clients. The precision regulation policy can regulate the AI apps after in-depth analysis. 

Working With Trusted Partners

IBM has been working on creating multi-stakeholder relationships with external partners for advancing the ethics in artificial intelligence. This initiative was launched in collaboration with the Vatican, which focuses on creating the human-focused AI. It aligns with human values, such as directing attention to vulnerable parts. For the most part, it can regulate the AI future.

Integrating Open-Source Toolkits 

In addition to defining the policies, principles, and collaborations, IBM has to prioritize the release and research regarding tangible tools. The research part can move the trust in AI. That being said, they have launched an open-source toolkit, known as AI Fairness 360, through which developers can share and receive the datasets and codes regarding bias. Also, the toolkit allows collaboration among developers. 

In addition, IBM research has led to the release of tools for defining, measuring, and advancing artificial intelligence. For instance, the Adversarial Robustness Toolbox offers the tools for ensuring robustness in AI. Secondly, the tool named AI FactSheets is directed towards increasing the transparency levels when it comes down to the end-to-end development of the lifecycle (AI, to be precise).