We’re excited to announce today that Labelbox has received a strategic investment from In-Q-Tel, the not-for-profit strategic investor that accelerates the development and delivery of cutting-edge technologies to U.S. government agencies that keep the nation safe.
One of the biggest challenges for enterprises including the National Security and Intelligence community to pursue AI applications is the need for large, high-quality labeled datasets to train their Machine Learning (ML) models. Creating these datasets requires significant time and expense—in many cases, a combination of expert human annotators using in-house labeling tools and/or a large third-party workforce that can manually label data at scale. Outsourced labeling services can come with a substantial quality trade-off and often end up being slower and more expensive in the long run. As a result, many businesses have no choice but to maintain some degree of internal labeling efforts, which are expensive in terms of infrastructure, human capital, and time. These enterprises are actively looking for software tools that facilitate high-quality, efficient, and cost-effective labeling workflows. That is where Labelbox comes in and why IQT, consummated a strategic investment in Labelbox.
Just like many large corporations, the U.S. Government, IC and National Security community have vast data holdings, but the majority of that data is not labeled for AI/ML applications. And like commercial companies, the IC is in need of tooling that enables in-house human labelers to be more efficient with data annotation while also making it easier to manage labeled training data.
Why a training data platform is critical
Our company has already been highly focused on building out our training data platform to better suit government requirements and we have been expanding our federal footprint these last 12 months across federal agencies in the defense, energy, intelligence and space sectors, along with their contracting partners. The platform supports all types of data across images (including, crucially, geospatial images), video, and text. The platform is already available via a range of secure deployments including Amazon C2S as well as in air-gapped environments. By engaging with IQT and their partners, we are accelerating this push into government applications and excited to support more advanced use cases for the IC and National Security markets.
George Hoyem, Managing Director at In-Q-Tel, summarized for us why this is an important investment for them:
“What the US federal government needs in order to realize the potential of AI applications is a better workflow so that it can aggregate and enrich data for AI to benefit diverse mission areas, from defense to healthcare. Our mission is to identify challenges and seek out commercial innovations that help federal agencies overcome those challenges and remove roadblocks to our country’s AI competitiveness. Labelbox provides a training data platform approach to enable teams across the government to collaborate on their training data and drive faster iteration cycles.”
Building a better workflow for labeling AI data is our core focus at Labelbox. It is why we formed the company and it has driven our growth, including our most recent series C round. Enabling collaboration, transparency, and the ability to identify errors and areas of low confidence are especially important for producing accurate training data quickly across all industries. Our training data platform enables far faster iteration cycles -- a vital asset for all organizations, but especially for government agencies building machine learning applications that must keep up with the rest of the world. Labelbox is also SOC 2 Type 2 certified.
The future ahead
AI will drive much of the future of defense – from identifying adversary activity to understanding satellite imagery. I saw this firsthand during my time at Planet Labs, where geospatial intelligence was a primary use case. But AI is only as good as the labeled data that trains those AI models. Traditional AI labeling approaches like brute force professional services labor can’t deliver the required results of delivering mission-critical AI into production and keeping it there. We are honored to be working with In-Q-Tel to support these mission-critical use cases.
We expect this relationship with IQT to help the IC and National Security markets enhance their ability to accurately and efficiently label and manage large volumes of training data. We are targeting an on-premises version of Labelbox’s software for these markets that is private and secure to support these efforts. In addition, we hope to decrease the amount of time it takes for these users to label data while simultaneously lowering costs by reducing the number of expert labelers needed for a particular labeling task.