Bias in artificial intelligence (AI) is a critical issue with significant ethical, legal, and social implications. Various technical solutions have been developed to address this problem, but they have limitations. This paper explores these limitations across five dimensions: who defines fairness and bias, which fairness metrics to use and prioritize, when these metrics are most effective, and for which populations and contexts the metrics are designed. We argue that technical solutions alone cannot fully address bias and that greater collaboration is needed among AI developers, end-users (especially those most affected by AI), and other stakeholders, including ethicists and policymakers. We discuss how participatory AI can help overcome the limitations of technical solutions and enhance bias mitigation efforts.