Wirestock raises $23M Series A to supply creative multi-modal data to AI labs

Wirestock raises $23M to supply creative multi-modal data to AI labs | TechCrunch

In the past few years, creative marketplaces and platforms have realized that they are sitting on a data goldmine, and they can either use that data to develop AI models themselves or turn it into a source of revenue by licensing it to other AI labs. Wirestock, which previously helped photographers distribute and sell their work on stock photography services like Shutterstock, has taken the latter path. The company pivoted to being a data provider in 2023, and now supplies datasets of images, videos, design assets, and gaming and 3D content to AI labs. Wirestock said its platform has signed up more than 700,000 artists and designers who complete different tasks for data collection, similar to freelancers on platforms like Fiverr. Mikayel Khachatryan, the company's co-founder and CEO, said Wirestock was transparent about its shift, and allowed artists to opt out of its data supply business (in 2022, the platform had over 100,000 photographers on its platform). Khachatryan didn't specify how many people switched over as data providers for AI, only saying "the majority" did. "Initially, a lot of our deals were just selling what we had off the shelf, like our existing library. But then it turned into a lot of custom requests for content and data, and that created new opportunities for creators, and the platform just took off," he said. The startup said on Thursday it has raised $23 million in Series A funding to build out the new data supply business. The round was led by Nava Ventures, and saw participation from SBVP (co-founded by Sheryl Sandberg), Formula VC, and I2BF Ventures. Khachatryan says Wirestock currently provides multi-modal data to six of the largest foundation model makers, but he wouldn't name them. He noted the company currently has an annual run-rate revenue of $40 million and has so far paid out $15 million to its contributors. As part of the transition, the startup had to retrain some of its teams to annotate and label data in detail to make it useful for AI labs. It also had to build sales and enterprise teams to be able to pitch to hyperscalers, and find ways to get more creative assets in areas like 3D modeling. Wirestock currently uses email marketing and referral programs to bring in new contributors. Photographers, videographers, and illustrators can apply to provide data on its website, but they have to complete an unpaid task as a quality check before they're accepted. The company said it uses a mix of AI and human reviews to evaluate all work on the platform. Demand for data supply services is sky high right now as AI labs race to continue improving their models. Companies like Surge, Scale AI, and Mercor have built businesses worth tens of billions seemingly overnight on the back of demand for different kinds of datasets, and a slate of new startups such as Micro1, Human Archive, and Human Native AI are working with top AI model makers, too. Wirestock wants to focus on providing data for models that aid creative use cases, such as image and video generation. The company is also exploring other modalities like audio and music. Freddie Martignetti, founder of Nava Ventures, said his fund was looking for a startup that was innovating on data procurement and refinement even before it learned about Wirestock. "I think Wirestock has a deep understanding of what foundational models and hyperscalers need in terms of multi-modal data to start creating more human-like systems. The cornerstone of our thesis was that multi-modal data will be increasingly important, not just to create images or videos, but for models to complete real-world tasks," Martignetti told TechCrunch. Wirestock currently employs 60 people, and will use the new funding to hire for research, engineering, and product roles. It is also building enterprise software for AI labs to collaborate on datasets. The funding round brings the startup's total capital raised to about $26 million.

AI training data provider Wirestock raises $23M in funding - SiliconANGLE

Wirestock Inc., a startup that provides multimodal training datasets for artificial intelligence developers, has raised $23 million in funding. TechCrunch reported today that early-stage fund Nava Ventures led the investment. It was joined by SBVP, Formula VC and I2BF Ventures. Wirestock has raised a total of $26 million in external funding to date. Wirestock launched in 2018 with a platform that helped photographers and other creative professionals distribute their images through stock image services. Earlier this year, the company announced a pivot to providing training datasets for AI projects. Wirestock offers model developers access to a catalog of more than 50 million images and videos created by about 700,000 users. The company also sells custom training datasets. According to Wirestock's website, one of its specialties is recording footage that demonstrates how users interact with applications. A startup with an AI-powered website design tool could commission images that show how to edit interface elements in Figma. According to TechCrunch, Wirestock's annualized revenue topped $40 million prior to its funding round. The company's growth may partly be driven by the surging industry interest in world models. Startups developing world models, which are trained on visual datasets, have raised billions of dollars in funding over the past year. Much of the capital has gone to two companies. World Labs Inc., which is led by pioneering AI researcher Fei-Fei Li, raised $1 billion in February from a consortium that included Nvidia Corp. It's developing world models that can be used for tasks such as training robots. Rival AMI Labs Inc. closed a $1.03 billion round around the same time to build world models that can automate hardware engineering tasks. Visual datasets of the kind sold by Wirestock can also be used to train computer use models. Standard Intelligence Inc., a startup that raised $75 million last month, used footage of users performing tasks in applications to build an algorithm called FDM-1. It says that the AI can automate multi-step business tasks in a more hardware-efficient manner than earlier models. Wirestock will reportedly use its capital to recruit more AI researchers, engineers and other technical professionals. The new hires will help the company develop a software product that will enable "enterprise AI labs to collaborate on datasets." Wirestock is also considering expanding its focus beyond images and videos to other types of training data. The company's funding round comes a few weeks after another AI data provider closed a $30 million investment of its own. San Francisco-based AfterQuery sells datasets that comprise AI prompts, user-created replies to the prompts and a step-by-step overview of the thought process behind each response. Those explanations enable AI models to hone their reasoning capabilities.