OpenAI wants to work with organizations to build new AI training data sets

OpenAI has launched Data Partnerships to address the cultural and linguistic biases inherent in Western-centric training data. By collaborating with third-party institutions, the company aims to create diverse datasets that better reflect global societies, languages, and domains. This initiative seeks to move beyond the limitations of existing internet-based corpora, which often amplify toxic language and narrow perspectives, by incorporating specialized content that is not easily accessible online. The program involves digitizing and curating data through partnerships with organizations like governments and legal entities, ensuring models understand specific industries and cultures. OpenAI plans to release both public open-source datasets and private ones for proprietary models. This dual approach allows partners to retain control over sensitive information while contributing to AI systems that offer broader, more inclusive understanding. The goal is to enhance the utility and safety of AI by exposing it to a wider array of human intentions and experiences. This effort is relevant to open data as it highlights the critical need for inclusive, representative datasets to mitigate bias in artificial intelligence. However, it also raises questions about transparency, compensation, and the commercial motivations behind data collection. As the AI landscape evolves, ensuring that data contributions are ethical and beneficial to all stakeholders remains a significant challenge for the open data community and developers alike.

Source: businessghana.com
Published on 2023-11-13

News about open_data on 2023-11-13 »

Open Data

Related news