TELUS Digital Launches Expert-Curated STEM Datasets to Enhance AI Training and Mitigate Bias

TELUS Digital, a leading global technology company specializing in digital customer experiences, has unveiled 13 off-the-shelf STEM datasets designed to enhance the training and performance of AI models. These datasets, which cover areas such as coding and reasoning, are crucial for advancing large language models (LLMs) and are curated by a diverse group of experts, including Ph.D. researchers, professors, and industry professionals. A recent survey by TELUS Digital highlights the growing concern over data sourcing in AI. Out of 1,000 U.S. adults surveyed, 87% (up from 75% in 2023) believe companies should be transparent about how they source data for generative AI (GenAI) models. Furthermore, 65% of respondents think that excluding high-quality, verified content from trusted media sources can lead to inaccuracies and biases in LLMs. Amith Nair, Global VP and General Manager of Data & AI Solutions at TELUS Digital, emphasizes the importance of expert-curated datasets in high-stakes domains. "As AI systems become more specialized and embedded in critical use cases, the quality of the data used to optimize outputs is a key differentiator between average performance and real-world impact," he stated. "General crowdsourced or internet data no longer suffice for today’s enterprises. Instead, 'wisdom of the experts' is crucial." TELUS Digital's new STEM datasets are designed to provide enterprises with high-quality, pre-validated data that can be immediately integrated into AI training workflows. The datasets are curated by a global network of experts, ensuring they are accurate, contextually relevant, and responsibly built. This approach not only enhances model performance but also reduces the risk of errors and biases in complex fields like healthcare and finance. Several data annotators highlight the significance of their work. Dancan, an AI scientist with a background in organic chemistry, explains that proper data annotation enables AI to collaborate more effectively with scientists, speeding up the discovery of life-saving therapies. Sourabh, a software engineer from India, notes that his background helps him tackle practical annotation projects, such as coding challenges, making AI models more effective and efficient. Justin, a Ph.D. candidate in Chemistry at the University of Vermont, adds that high-quality data annotation, combined with well-trained LLMs, can significantly lower the barrier to entry for scientists, fostering more profound and efficient innovation. TELUS Digital's expertise in data annotation spans over two decades, with a comprehensive portfolio of end-to-end capabilities. Beyond its off-the-shelf datasets, the company offers fully custom AI data projects tailored to specific use cases, including multilingual and multimodal annotation. This ensures that datasets align with industry requirements and regulatory standards, particularly in sectors where accuracy and compliance are paramount. The company's data annotation services are supported by its advanced proprietary platform, Ground Truth (GT) Studio, which combines global human expertise with intelligent automation. GT Studio ensures exceptional data quality, accuracy, and efficiency, enabling enterprises to scale their AI initiatives effectively. TELUS Digital's commitment to responsible AI is encapsulated in its Humanity-in-the-Loop principles, which guide the development and deployment of transformative technologies. This responsible approach considers and addresses the broader impacts of AI, fostering trust and reliability in AI-driven solutions. Industry Insights and Company Profile Industry insiders and experts agree that the quality and sourcing of training data are critical factors in the success of AI models. High-quality, expert-curated datasets can significantly reduce the time and resources needed to develop and refine LLMs, making them indispensable for enterprises aiming to stay competitive in the AI landscape. TELUS Digital, with its robust platform and extensive network of experts, positions itself as a leader in providing high-context, high-quality data for AI training. The company's deep experience and commitment to responsible AI development align well with the increasing demands of businesses and the expectations of consumers for transparency and accuracy in AI systems. For more information on TELUS Digital's off-the-shelf datasets and custom AI data solutions, visit their Data & AI Solutions page at telusdigital.com.