Home / Technology / AI Learns From Millions via Mobile Games
AI Learns From Millions via Mobile Games
19 Feb
Summary
- Rapidata gamifies AI training using 20 million mobile app users.
- Feedback is integrated into AI training loops in near real-time.
- The company raised $8.5 million to scale its unique approach.
Despite AI's advancements, training models still heavily depends on human feedback through a process known as RLHF. Traditionally, AI labs relied on fragmented networks of foreign contractors for this crucial step, leading to lengthy delays in model development.
Rapidata has emerged with an innovative solution, gamifying RLHF by integrating opt-in feedback tasks into popular mobile apps. This platform connects AI labs with approximately 20 million users, enabling them to provide feedback instantly by choosing tasks over watching ads.
This method allows AI labs to iterate on models in near-real-time, drastically reducing development cycles. Rapidata aims to provide human judgment at a global scale, facilitating constant feedback loops for evolving AI systems.
The company announced an $8.5 million seed round co-led by Canaan Partners and IA Ventures, with participation from Acequia Capital and BlueYard, to scale its on-demand human data approach.
Rapidata's core innovation is its distribution method, leveraging the attention economy of mobile apps. Users choose between watching ads or providing AI feedback, with a significant majority opting for the latter. This 'crowd intelligence' approach taps into a diverse global demographic, processing up to 1.5 million annotations per hour.
The platform facilitates 'online RLHF,' moving human judgment directly into the AI training loop via API integration with GPUs. This real-time feedback prevents issues like 'reward model hacking' by grounding training in actual human nuance.
As AI expands into generative media, the need for subjective, taste-based curation grows. Rapidata allows AI teams to test models in real-world contexts across different regions, ensuring outputs feel authentic and human-like.
Operationally, Rapidata acts as an infrastructure layer, eliminating the need for companies to manage their own annotation processes. This lowers the barrier to entry for AI teams struggling with the cost and complexity of traditional feedback loops.
CEO Jason Corkill envisions a future of 'human use,' where AI models programmatically solicit human judgment for design and simulation, ensuring AI remains connected to evolving societal nuances.
