Samsung invests in AI startup Memories.ai for advanced video analysis
While many existing AI tools can summarize short video clips, their efficacy diminishes significantly when tasked with analyzing multiple videos or extensive, hours-long footage.
ADVERTISEMENT
Memories.ai, a burgeoning AI startup focused on advanced video understanding, has successfully closed an oversubscribed $8 million seed funding round. The investment, led by Susa Ventures with significant participation from Samsung Next, Fusion Fund, Crane Ventures, Seedcamp, and Creator Ventures, will fuel the company's ambitious plans to process and contextualize thousands of hours of video footage, addressing a critical gap in current AI capabilities.
While many existing AI tools can summarize short video clips, their efficacy diminishes significantly when tasked with analyzing multiple videos or extensive, hours-long footage. This limitation poses a substantial challenge for industries like security, which needs to scrub through vast amounts of camera footage, and marketing, which seeks to derive insights from extensive campaign and product shoot videos.
Memories.ai aims to overcome this hurdle with its proprietary AI platform, designed to process an astonishing 10 million hours of video. The startup's core offering provides a "contextual layer" for companies with large video libraries, enabling searchable indexing, intelligent tagging, precise segmentation, and robust aggregation of video data.
The company's co-founder, Dr. Shawn Shen, brings a strong research background from Meta’s Reality Labs, where he was a research scientist while pursuing his PhD. His co-founder, Enmin (Ben) Zhou, also has significant experience as a machine learning engineer at Meta.
"All top AI companies, such as Google, OpenAI, and Meta, are focused on producing end-to-end models. Those capabilities are good, but these models often have limitations around understanding video context beyond one or two hours," Shen explained to TechCrunch. "But when humans use visual memory, we sift through a large context of data. We were inspired by this and wanted to build a solution to understand video across many hours better."
The strong investor interest, which saw the round grow from an initial target of $4 million to $8 million, underscores the perceived market need for Memories.ai's solution. Misha Gordon-Rowe, a partner at Susa Ventures, lauded Shen as "a highly technical founder" with a deep commitment to "pushing boundaries of video understanding and intelligence." Gordon-Rowe emphasized that "Memories.ai can unlock a lot of first-party visual intelligence data with its solution. We felt that there was a gap in the market for long context visual intelligence, which attracted us to invest in the company."
Samsung Next's investment thesis offered a distinct perspective, highlighting the potential for Memories.ai's technology in consumer applications. Sam Campbell, a partner at Samsung Next, noted, "One thing we liked about Memories.ai is that it could do a lot of on-device computing. That means you don’t necessarily need to store video data in the cloud. This can unlock better security applications for people who are apprehensive of putting security cameras in their house because of privacy concerns." This on-device processing capability could significantly enhance privacy and data security for end-users.
Memories.ai employs a sophisticated, in-house tech stack and models for its analysis. The process begins by removing noise from videos, followed by a compression layer that retains only critical data. An indexing layer then makes this video data searchable using natural-language queries, complete with segmentation and tags. Finally, an aggregation layer summarizes the indexed data, facilitating the creation of comprehensive reports.
Currently, Memories.ai serves two primary client segments: marketing and security. Marketing firms leverage the platform to identify brand trends across social media videos and inform the creation of new video content. For security companies, the startup provides tools to analyze surveillance footage, reasoning through patterns to detect potentially dangerous actions.
While clients currently upload their video libraries for analysis, Shen envisions future integrations where customers can easily sync content from shared drives. He foresees a future where users can ask an AI assistant contextual questions like: "Tell me all about people I interviewed in the last week." Looking further ahead, Shen believes the technology could play a transformative role in training humanoid robots for complex tasks and assisting self-driving cars in remembering diverse routes.
Memories.ai currently has a team of 15 employees and plans to utilize the new funding to expand its team and enhance its search capabilities. The startup is navigating a competitive landscape, facing other "memory layer" AI startups like mem0 and Letta, which currently offer limited video support. It also competes with established players such as TwelveLabs and Google, both of whom are actively developing AI models for video understanding.
However, Shen remains confident in Memories.ai's position, asserting that its solution is more "horizontal," enabling seamless integration with various existing video models.
