Memories AI, an AI research laboratory founded by a former Meta researcher and a Cambridge University PhD in Computer Science, has officially launched, introducing the world's first Large Visual Memory Model (LVMM). This breakthrough technology aims to give AI human-like visual memory capabilities, enabling machines to "see, understand, and remember" visual information. At the same time, Memories AI announced an $8 million seed round led by Susa Ventures, marking its ambitious aspirations in the field of AI visual memory.

Money Capital Funding

 World First: Large Visual Memory Model (LVMM)

Memories AI's core technology is its proprietary Large Visual Memory Model (LVMM), the first AI architecture capable of continuously capturing, storing, and recalling visual information. Unlike existing AI systems, traditional models can typically only process short video clips (15-60 minutes), losing context in long video analysis and making it difficult to answer questions like "Have you seen this before?" or "What changed yesterday?" LVMM simulates human memory mechanisms, enabling it to process video data lasting millions of hours and build persistent, searchable visual memory libraries.

This technology is implemented through a three-layer architecture: first, videos are denoised and compressed, extracting key information; second, a searchable index layer is created, supporting natural language queries; finally, a consolidation layer structures the visual data, allowing AI to recognize patterns, retain context, and perform cross-time comparisons. This makes Memories AI show unprecedented efficiency and accuracy when handling large-scale video data, claiming to have 100 times the video memory capacity of existing technologies.

 Broad Applications: Cross-industry Innovation from Security to Marketing

Memories AI's LVMM technology has already shown great potential in multiple fields, covering the following scenarios:

- Physical Security: Providing anomaly detection for security companies by analyzing long-term surveillance videos to quickly identify potential threats.

- Media and Marketing: Helping marketing teams analyze massive video content on social media, identifying brand mentions, consumer trends, and sentiment. For example, a social media platform has used Memories AI's technology to gain insights into long-term trends on platforms like TikTok, maintaining a competitive edge.

- Robotics and Autonomous Driving: By giving AI long-term visual memory, it supports robots in performing complex tasks or helps autonomous cars remember visual information about different routes.

Memories AI's platform can be accessed via API or a chatbot web application, allowing users to upload videos or connect their own video libraries and query video content through natural language. This flexible interaction method makes it suitable for a wide range of applications, from enterprise-level solutions to personalized uses.

 An $8 Million Seed Round, Accelerating Technology Implementation

Memories AI's seed round was led by Susa Ventures, with participation from well-known investment institutions such as Samsung Next, Crane Venture Partners, Fusion Fund, Seedcamp, and Creator Ventures. The funding amount increased from an initial target of $4 million to $8 million, showing investors' strong confidence in Memories AI's market potential. The funds will be used to expand the engineering team, deepen R&D on privacy and compliance frameworks, and accelerate the acquisition of enterprise clients.

One investor stated that Memories AI's long-term video intelligence technology will provide critical infrastructure for robotics, enterprise software, consumer electronics, and even general artificial intelligence (AGI), with market potential covering multiple industries worth trillions of dollars.

 Team Background: Top Talent from Meta Reality Labs

Memories AI was co-founded by two former researchers from Meta Reality Labs, who have deep expertise in visual AI and device-side learning. The CEO has extensively researched multimodal AI, exploring human visual memory mechanisms, while the CTO developed several production-grade AI systems at Meta. Their technical insights and deep understanding of AI memory bottlenecks laid the foundation for Memories AI's innovations.

Visual Memory Opens a New Chapter for AI

As the editorial department of AIbase, we believe that Memories AI's LVMM technology not only fills a gap in AI's long-term video understanding but also paves the way for AI's shift from "instant response" to "dynamic learning." By granting AI human-like visual memory capabilities, Memories AI is expected to trigger a technological revolution in areas such as security, marketing, consumer electronics, and robotics. Although facing challenges from competitors like Google and TwelveLabs, its "horizontal" technology architecture allows it to be compatible with various video models, demonstrating strong flexibility.

Conclusion

The emergence of Memories AI marks a major breakthrough in the field of AI visual memory. From processing millions of hours of video to empowering next-generation smart devices, its large visual memory model is redefining the possibilities of AI. AIbase will continue to track Memories AI's latest developments and provide readers with more in-depth analysis and reports. Do you look forward to AI having "never forget" memory capabilities? Please share your thoughts in the comments section!