How Anthropic uses Surge AI to Train and Evaluate Claude
Blog post from Surge AI
Anthropic, a leading AI company, focuses on building safe, state-of-the-art large language models (LLMs) like their AI assistant, Claude, which surpasses OpenAI's ChatGPT in various domains. Faced with challenges in gathering high-quality human feedback vital for training their models, Anthropic partnered with Surge AI to leverage their human data labeling platform. Surge AI provides proprietary quality control technology, domain expert labelers, and rapid experimentation interfaces, which have proven crucial in delivering the sophisticated human feedback needed to refine LLMs. This partnership has enabled Anthropic to advance their reinforcement learning from human feedback (RLHF) research, ensuring their models remain helpful and harmless while pushing the boundaries of AI safety and capability.