News Summary

AI startups are increasingly taking data collection into their own hands to gain a competitive edge. Companies like Turing and Fyxer are moving away from freely-scraped data or low-paid annotators, opting instead for high-quality, curated datasets. Turing, for example, hires freelancers to collect diverse data through manual collection, aiming to train AI models in sequential problem-solving and visual reasoning. Similarly, Fyxer uses a focused training approach with experienced personnel to enhance email sorting AI models. The emphasis on proprietary data collection serves as a robust competitive moat, ensuring that the quality of the data, rather than quantity, defines AI performance.