We believe that what holds machine learning back in the real world today isn’t actually the algorithms. Rather, it’s the lack of quality training data and the means to make imperfect algorithms useful.
CrowdFlower makes machine learning work by combining the best of human and machine intelligence in a single platform. We call this human-in-the-loop.
CrowdFlower has always made it easy to collect and label training data. Now we make it easy to connect that training data directly to state-of-the-art machine learning algorithms.
There are always places where algorithms struggle and 80% accurate algorithms are tough to use in the real world. CrowdFlower has filled in these gaps for years. Now, we’ve made it easy to automatically identify where machine learning is struggling and send them back to CrowdFlower jobs for humans to label.
Most importantly, our AI product makes it easy to take the new labels and continuously retrain algorithms in a process known as active learning. Active learning has been regarded for years as an incredibly powerful approach and one that’s essential for making machine learning deployments work in the real world.