"Zurich-based startup Rapidata has just taken a significant step toward fixing that with a €7.2 million seed round aimed at scaling a global network for human feedback, the essential ingredient for aligning and refining AI models. The round was disclosed today by EU-Startups, positioning Rapidata as an early player in what could become a major layer of AI infrastructure: a system that compresses feedback loops from humans into AI training systems from weeks or months down to hours or days."
"Modern AI systems have become astonishingly capable at generating text and images, but they still struggle with nuance, judgement and context, the very aspects humans excel at. That's why many advanced models are trained with reinforcement learning from human feedback (RLHF), where people judge outputs and rate responses to shape the model's behaviour. Rapidata claims to change tha t by tapping a continuously available, worldwide network of real people, enabling AI teams to gather large volumes of high-quality human judgments far more quickly and flexibly"
Rapidata, founded in 2023 and based in Zurich, secured a €7.2 million seed round to build a global, on-demand network for human feedback. The platform offers continuous, worldwide access to real people to produce large volumes of high-quality judgments for model training. The service aims to compress human-in-the-loop feedback cycles from weeks or months down to hours or days. The approach contrasts with static annotator pools or traditional labeling services by scaling with demand and removing geographic segmentation. Investors and developers are increasingly treating human judgment as a core layer of AI infrastructure for safety and alignment.
Read at TNW | Startups-Technology
Unable to calculate read time
Collection
[
|
...
]