Manu Sharma and Brian Rieger•January 6, 2022
We're thrilled to announce that Labelbox has raised $110 Million in Series D financing to accelerate the development and distribution of our AI training data platform.
Softbank’s Vision Fund II led the round. Databricks Ventures and Snowpoint Ventures (a new fund started by a team from Palantir) also participated along with previous investors B Capital Group, Andreessen Horowitz and Catherine Wood, CEO and founder of ARK Invest.
The world has made tremendous progress since we first worked with neural networks in 2011. Now, nearly every few months, we are experiencing AI breakthroughs that push the boundaries of what's possible. Underlying these outstanding achievements is significant innovation in computing and data. AI model training costs are reducing by a factor of two every 16 months, making AI development economical and accessible to all. With state-of-the-art neural network architectures now widely available, it is clear that data is the most critical piece in developing performant AI systems.
Working with data to develop production AI applications requires a new set of tools and technologies. Data must flow and transform through numerous steps in a machine learning system. There needs to be an adequate level of human supervision throughout this process to ensure that the correct data is being used to train the neural networks. This supervision can be in any form: labeling data, debugging models, finding edge cases, or prioritizing the right data to fix the problems in a model. In other words, companies developing AI have to essentially engineer a new kind of system that interfaces humans and neural nets with data as the medium of information exchange. Some AI practitioners in the industry call this a data engine.
Today, hundreds of companies use Labelbox as their primary data engine including Genentech, Warner Bros, Black & Decker and Stryker. In 2021, 80% of Labelbox business came from the enterprise. Enterprises choose Labelbox because it gives them the technology needed for their development teams to rapidly iterate with data, debug their AI models, and curate unstructured data. All in one place. There is no other enterprise solution that offers complete transparency throughout the entire workflow and incorporates labeling automation to minimize human labeling costs.
Labelbox has been at the forefront of bringing innovative ideas into its products. Our company was one of the first to bring model-assisted labeling automation to the market, enabling AI teams to cut their data labeling costs by over 50%. Now, we have added active learning, a state-of-the-art technique in which algorithms help target the areas of your ML model that need the most improvement, enabling AI teams to become far more efficient. By using active learning, AI development teams can target and address specific issues in their AI models and gain a unit increase in model performance in less time and with less data. Our customers rapidly gain a competitive edge with AI because they can iterate faster and focus on model performance.
Just recently, during a webinar, Allstate shared that their iteration cycle time (from an idea to an ML model) has now shrunk to just two weeks. Think about that: a company born during the Great Depression (90 years ago) is now iterating on par with AI teams at Tesla or Google. AI development velocity is critical for startups, too. Labelbox has been the platform of choice for startups such as Overjet.ai, which has grown from a seed company to a leading dental AI technology company valued at $425M. Labelbox is enabling and accelerating AI breakthroughs among its customers that span all major industries, every day.
Progress is happening fast and the practical applications of AI now touch every aspect of life. No matter how powerful these AI systems may become, there will remain a consistent need to ensure these models are making the right decisions. We believe that the data used to train these models will be the single most important factor to ensure that happens. With powerful models fueled by accurate data, an entire frontier of AI-driven breakthroughs is on the horizon. We are excited to be building the company that helps the world’s leading AI companies achieve these breakthroughs as quickly as possible.
Watch this demo to get a closer look at what we've been building.
Labelbox•May 25, 2021
Announcing the Labelbox connector on Databricks: Productionizing unstructured data for AI and analytics at scale
Labelbox has recently launched a connector between Databricks and Labelbox — the LabelSpark library — so teams can connect an unstructured dataset to Labelbox. With LabelSpark, teams can programmatically set up an ontology for labeling and return the labeled dataset in a Spark DataFrame.