Ideas of India 2025: 'AI Issues with Bias Are Not Fully Solved Yet'-Manish Gupta

At the Ideas of India Summit 2025, Manish Gupta, Director of Google DeepMind, delved into the critical challenges and ethical considerations surrounding Artificial Intelligence (AI). Moderated by author and columnist Chetan Bhagat, Gupta explored the moral implications of AI and how it learns from human biases. He pointed out that AI itself does not have a sense of morality, but rather mirrors the data it is trained on, which often includes human biases.Gupta acknowledged the importance of responsible AI development, stressing that companies like Google are working to mitigate biases, especially in non-western contexts such as caste and religion. He emphasized, "AI's issues with bias are not fully solved yet," highlighting that AI models may inadvertently carry biases present in society, such as gender or racial discrimination, which have been observed in hiring algorithms. He explained that while AI can be incredibly powerful, its application must be guided by strict ethical frameworks and continuous audits to ensure fairness.Gupta also touched on the potential for AI to be misused, raising concerns about how AI could be used for harmful purposes, such as designing dangerous weapons. He concluded by acknowledging the ongoing efforts to address these ethical dilemmas, but emphasized that solving these issues remains an open problem.