Defining fairness: How IBM is tackling AI governance

Key Takeaways:

– Enterprises are hesitant to adopt AI solutions due to the difficulty in balancing the cost of governance with the behaviors of large language models (LLM)
– Challenges in specifying what a harmful answer is for LLMs
– IBM is looking to develop AI that developers can trust
– IBM aims to use the law, corporate standards, and internal governance to control LLMs
– Contextual documents can be used to fine-tune LLMs and detect harmful content
– IBM develops LLMs on trustworthy data and implements detection mechanisms for biases
– The proposed EU AI Act will link AI governance with user intentions
– Usage is a fundamental part of IBM’s model of governance.

TechRadar:

Enterprises are hesitant to adopt AI solutions due to the difficulty in balancing the cost of governance with the behaviours of large language models (LLM), such as hallucinations, data privacy violations, and the potential for the models to output harmful content.

One of the most difficult challenges facing the adoption of LLM is in specifying to the model what a harmful answer is, but IBM believes it can help improve the situation for firms everywhere.

Source link

AI Eclipse TLDR:

Enterprises are hesitant to adopt AI solutions due to concerns about the behavior of large language models (LLMs), such as generating harmful content and violating data privacy. IBM aims to address these concerns by developing AI that developers can trust. One of the challenges is specifying what constitutes harmful content to the model. IBM proposes using the law, corporate standards, and internal governance as control mechanisms for LLMs, allowing governance to incorporate the ethics and social norms of different regions and industries. This approach enables fine-tuning and detection of harmful content. IBM also emphasizes the importance of developing LLMs on trustworthy data and implementing detection mechanisms to address biases. The proposed EU AI Act will link AI governance with user intentions, and IBM considers usage as a fundamental aspect of its governance model. Overall, IBM aims to improve trust and governance in AI adoption for enterprises.