Scale AI vs Turing

Scale AI and Turing are both American startups that provide services for generation, preparing and improving data for training AI models. They also help customers to evaluate and fine-tune trained AI models. Both companies use a combination of automated algorithms and human in the loop to ensure model quality. Their clients include large AI companies and research labs.

However, Scale AI (founded in 2016) primarily specializes in data annotation. It operates a distributed network of annotators and subcontractors via online platforms Remotasks and Outlier. The company also develops the "Scale Data Engine" platform for fine-tuning and reinforcement learning based on human feedback. The company is 49% owned by Meta, that often raises concerns about independence, data leaks and client conflicts of interest.

Turing (2018) focuses specifically on assessing the quality of AI models. It hires experts (PhDs, domain specialists) to create input-output data pairs and provide internal logical context. It operates as a research accelerator for frontier AI labs, offers custom training datasets and organizes research projects from the scratch. Turing positions itself as more independent and neutral toward major players than Scale.