Imbue, the AI research lab formerly known as Generally Intelligent, announced that it has raised $200 million in a Series B funding round, bringing the company’s valuation to over $1 billion. The funding round saw participation from notable investors, including the Astera Institute, Nvidia, Cruise CEO Kyle Vogt, and Notion co-founder Simon Last.
Key Takeaway
Imbue has raised $200 million in a Series B funding round, bringing its total raised capital to $220 million. The AI research lab aims to develop AI models that can robustly reason and code, with a focus on addressing the primary blocker to effective AI agents: reasoning. Imbue believes that code plays a crucial role in improving reasoning and action, making it an essential component in AI models. The funding will be used to accelerate the development of AI systems that can help accomplish larger goals in the real world.
This latest funding round brings Imbue’s total raised capital to $220 million, solidifying its position as one of the well-funded AI startups in recent months. Although slightly behind AI21 Labs, Cohere, and Adept in terms of funding, Imbue is confident in its ability to make a significant impact in the AI industry.
In a blog post, Imbue emphasized that the funding will be used to further develop AI systems capable of robust reasoning and coding. The company is dedicated to building practical AI agents that can effectively accomplish larger goals and safely operate in the real world.
Shifting Approach: From 3D Worlds to Internal Usefulness
Imbue initially launched with a focus on researching the fundamentals of human intelligence lacking in machines. Their strategy involved turning these fundamentals into various tasks to be solved, with AI models being designed and tested in complex 3D virtual worlds. However, the company’s approach has since shifted.
Imbue is now concentrating on developing internally useful models, including those capable of coding. What sets Imbue’s coding models apart is their ability to robustly reason. The company believes that reasoning is a significant obstacle in creating effective AI agents. Robust reasoning involves addressing uncertainty, adapting approaches, seeking new information, playing out scenarios, making decisions, hypothesizing, and dealing with the complexities of the real world.
The Importance of Code: Improving Reasoning and Action
Imbue strongly advocates for the importance of code as it enhances reasoning and enables models to take actions more effectively. The company argues that an AI agent that writes a SQL query to extract information from a table has a higher chance of satisfying a user request than an agent that attempts to assemble information without using any code. Training on code also helps models improve their reasoning abilities, while training without code often results in poor reasoning.
This philosophy aligns with other companies, like Adept and Google DeepMind, which also prioritize creating AI capable of automating software processes and controlling computers, respectively.
Training Large Models and Investing in AI Tooling
Imbue’s approach involves training very large models, specifically those with over 100 billion parameters. These models are optimized to excel on the company’s internal benchmarks for reasoning. To facilitate this training, Imbue has partnered with Nvidia and utilizes a compute cluster comprising 10,000 GPUs from Nvidia’s H100 series.
Additionally, Imbue is investing in the development of its own AI and machine learning tooling, including AI prototypes for debugging and visual interfaces on top of AI models. The company is also conducting research to better understand the learning process in large language models.
Driving Future AI Innovation
While Imbue does not currently intend to commercialize its ongoing work, the tools and models being developed serve as a foundation for future, more general-purpose AI solutions. Imbue aims to create a platform that empowers users to build robust and custom AI agents, putting the power of AI at everyone’s fingertips.
