Generative artificial intelligence (AI) chatbots are not free of inherent biases, and several projects such as Amazon and Microsoft are working to reduce these biases associated with large language models (LLM).

Rajeev Rastogi, head of machine learning at Amazon, leads such projects that several Amazon India employees are working on.

The international nonprofit Bill and Melinda Gates Foundation is funding a three-year project by Bengaluru-based artificial intelligence startup Karya to develop a method to tell with some degree of certainty whether a dataset is gender-biased or not, and they have a pipeline to delete data sets.

In response to questions sent to the Gates Foundation and Karya, Vivek Seshadri, Karya's chief technology officer, says that the project is currently in the early stages of research and the company hopes to get feedback on mitigating biases language technology by mid-2024.

Hanna Wallach, head of partner research at Microsoft Research says that the company believes AI technologies should work fairly. He said the company will continue to invest in research to identify measure and mitigate a range of justice-related harms, and create new ways to proactively test its AI systems.

To that end, the company employs a wide range of experts, including anthropologists, linguists, and social scientists, who all bring valuable perspectives to the table that push and challenge the thinking of engineers and programmers.

Monojit Choudhury, professor of artificial intelligence at MBZ University in Abu Dhabi says, “Until last month he was working on responsible AI aspects of GPT4-based Bing Chat at Microsoft Turing, etc.”