A number of major companies including IBM, Microsoft and Amazon have launched new machine learning platforms in the past year.
But data enrichment company CrowdFlower believes that to be effective and commercially viable machine learning needs more training and better data.
It's using the Rich Data Summit to launch its new AI product that combines machine learning and human-labeled training data sets to create predictive models which can be applied against new data. With this capability data scientists will be able to reduce the cost and increase the speed with which they enrich their data, without sacrificing quality.
"The industry has been having the wrong debate about human-versus-machine intelligence," says Lukas Biewald, chief executive officer and founder at CrowdFlower. "Human intelligence and machine intelligence aren't in competition; they're natural complements that reinforce each other. Humans' great strength is the cognitive ability of the brain, which brings into play context, meaning and judgment, and machines' great strengths are consistency and speed. By combining the best of human and machine intelligence into a single platform like CrowdFlower, the result is more data and higher quality data delivered faster and at lower cost. Data scientists who feed their machine learning platforms with high quality large scale human-labeled data sets can make the transition from interesting science experiment to a commercially viable business process generating millions of dollars of value for their company".
With AI, CrowdFlower customers will be able to apply a predictive model against new data sets. For rows of data that fall below a customer-defined confidence level, units can be routed to human contributors to complete enrichment tasks such as sentiment analysis or data categorization. By combining machine and human intelligence in a single platform, CrowdFlower can intelligently assign data enrichment tasks to either humans or a machine based on the customer requirements for scope, quality and cost.
The product is built to be interoperable with both open source and commercial machine learning solutions such as scikit-learn, Google Prediction, IBM Watson, and Metamind. The company plans to partner with other machine learning solutions so data science teams can make their own choice and avoid vendor lock-in.
AI will enter private beta this year with general availability expected in the first quarter of 2016. You can find out more on the CrowdFlower website.
Image Credit: art4all / Shutterstock