UPDATED 13:17 EDT / FEBRUARY 02 2021

AI

Superb AI raises $9.3M to automate the creation of AI training datasets

Startup Superb AI Inc. announced today it has raised $9.3 million from investors to pursue its vision for artificial intelligence development, which involves using deep learning to automate and thereby speed up key parts of the workflow.

In particular, San Mateo, California-based Superb AI focuses on automating the task of putting together AI training datasets. The process is normally performed mostly by hand and can in some cases take up the majority of development teams’ work week. Superb AI says that it has created a set of technologies that reduces the amount of time involved in the chore by as much as 80%

A typical training dataset includes thousands of individual files and descriptive labels that are manually attached to each file by humans. For example, if a company is developing an autonomous driving AI, its training dataset might include photos of road traffic with annotations describing the vehicles shown in each photo. During the AI training phase, those annotations give the neural network essential context that helps it accurately process the data.

It’s an effective approach to developing neural networks, but it comes with a catch: Manually labeling training datasets can take thousands of person-hours.

Superb AI’s answer to the challenge is to introduce another set of deep learning models into AI projects. The startup provides a service that enables teams building AI software to create a second, auxiliary neural network optimized specifically to label training data for the AI software they’re developing. Superb AI says that this auxiliary neural network can perform labeling several times faster than a human.

Parts of the workflow still require human input: datasets must be manually reviewed to catch any erroneous annotations that the label-generating AI may have created. But though it’s yet not possible to label training data using deep learning alone, Superb AI says it has found a way to partially automate the manual review process as well. The startup’s service assigns a confidence score to each AI-generated annotation so human reviewers can more quickly find files likely to contain errors.

Superb AI’s approach has helped it win deals with Samsung Electronics Co. Ltd., Qualcomm Inc. and LG Electronics Inc.

The startup’s value proposition has also won over a half-dozen tech investors. The $9.3 million round of funding it announced today was led by Atinum Investment with participation from Premier Partners, Stonebridge Ventures, Murex Partners, KT Investments and Duke University’s Angel Network. Overall, Superb AI has raised $11.3 million to date

The startup plans to use the funding to hire more workers, invest in product development and gear up customer acquisition activities. 

Superb AI competes with several other venture-backed startups that are working to reduce the amount of manual work involved in training neural networks. Research groups such as OpenAI are also tackling the challenge. At the start of the year, OpenAI demonstrated an AI called Clip that it had developed using a new, more efficient approach to creating training datasets. 

Image: Superb AI

A message from John Furrier, co-founder of SiliconANGLE:

Your vote of support is important to us and it helps us keep the content FREE.

One click below supports our mission to provide free, deep, and relevant content.  

Join our community on YouTube

Join the community that includes more than 15,000 #CubeAlumni experts, including Amazon.com CEO Andy Jassy, Dell Technologies founder and CEO Michael Dell, Intel CEO Pat Gelsinger, and many more luminaries and experts.

“TheCUBE is an important partner to the industry. You guys really are a part of our events and we really appreciate you coming and I know people appreciate the content you create as well” – Andy Jassy

THANK YOU