When the world is panning for machine learning gold, it pays to help them dig through the data.
What’s new: Machine learning entrepreneurs can make their mark (and their fortune) building services that help other companies develop, deploy, and monitor AI, venture capitalist Rob Toews argues in Forbes.
How it works: Toews points to Scale.AI, a startup that labels data, as one of a new generation of companies capitalizing on the AI industry’s demand for ancillary services. In August, the four-year-old company raised $100 million at a valuation of more than $1 billion. And labeling isn’t the only area of machine learning ripe for entrepreneurship.
- Synthetic data: Applied Intuition, Parallel Domain, and Cognate specialize in making synthetic data for autonomous driving and other applications where real-world training data is often scarce.
- Optimization: Gradio and Alectio help AI developers curate data to improve training efficiency. SigOpt offers a platform that guides companies through model specification from choosing an architecture to determining the number of training epochs.
- End-to-end management: Amazon’s SageMaker offers tools that help manage custom models throughout their lifecycle. Microsoft Azure Machine Learning Studio is geared toward data analysis. Google recently released Cloud AI Platform to get in on the action.
Behind the news: Companies like Adobe and Capital One are spending hundreds of millions on cloud computing. This is driving demand for services that help them handle their cloud resources more efficiently. Among the beneficiaries are companies like Alation, Collibra, and Starburst Data that help catalog, query, and manage machine learning data, writes investor Matt Turck.
Why it matters: Toews believes there are billions of dollars to be made by companies that provide machine learning services. Such services will also nurture new AI applications and accelerate their adoption across a variety of industries.
We’re thinking: These companies aren’t only promising businesses. By taking on tasks like data procurement, model optimization, and lifecycle management, they could free engineers to focus on building products that fulfill deep learning’s potential.