Liquid AI Unveils Next-Gen Foundation Models at MIT Event
Liquid AI based in Boston, Massachusetts, held a public event today at MIT to showcase their latest developments. The company, spun out of MIT, is focused on creating foundation models with the aim of building versatile and efficient general-purpose AI systems that can scale across various applications.
Founded by four core members with strong ties to MIT and CSAIL—Ramin Hasani (CEO), Mathias Lechner (CTO), Alexander Amini (Co-founder & Chief Science Officer), and Daniela Rus, Co-Founder and Director, MIT CSAIL—their mission is to develop highly capable AI systems that address problems of any scale, giving users the tools to build, access, and fully control their AI solutions.
Positioning itself as the Boston-based / East Coast rival to foundation model companies like OpenAI, Google, and Anthropic, which are primarily based in Silicon Valley. Liquid AI has directly challenged ChatGPT, claiming that its innovative architecture and AI performance and efficiencies are superior.
Liquid AI has built a suite of Liquid Foundation Models (LFMs) boasting several key features:
Enhanced quality for reliable decision-making, driven by advanced knowledge capacity.
Sustainability, thanks to efficient memory usage and near-constant inference speeds.
Reduced reliance on cloud services, which lowers costs and energy consumption.
Greater explainability, as LFMs are built from first principles, offering more transparency than transformer-based architectures.
According to Liquid AI, their LFMs are versatile and high-performing, with applications spanning various industries, including natural language processing, audio analysis, video recognition, and sequential multimodal data analysis. The vertical applications are obvious and important. Their Bio LFM, Transaction LFM, Time (as in time series) LFM and Liquid DevKit show great promise.
The Event
Kresge Auditorium at MIT, with its 1,200-seat capacity, appeared to be nearly full. The crowd was lively, likely due to the strong turnout from Liquid AI employees, as the company currently has nearly 50 employees and full-time equivalents (FTEs). This contributed to the energetic atmosphere.
One humorous highlight of the event was the repeated use of the acronym "LFMs," which, due to Kresge’s aging sound system, many attendees misheard as "elephants."
Funding
Liquid AI has completed two funding rounds, supported by three lead investors, with a total of 25 investors participating to date. Interestingly, many of these investors lack deep technical expertise in AI, as they are not exclusively focused on the field. In total, Liquid AI has raised $37.6 million in seed funding.
Conclusion
Liquid AI’s suite of products and devkit exemplify the concept of "AI building AI," where AI systems are designed and optimized by other AI systems. This approach automates key stages of development, enabling faster, more efficient, and sophisticated AI model creation with minimal manual intervention. A key takeaway from today’s event is that Liquid AI is a trend leader, driving AI's ability to accelerate its own evolution and unlock new capabilities that were previously out of reach using traditional methods.