Imbue Secures $200 Million Investment to Develop Practical AI Agents

Artificial intelligence (AI) startup Imbue, formerly known as Generally Intelligent, has raised $200 million in a Series B funding round, with a valuation of over $1 billion.

The funding round attracted investment from notable names such as Astera Institute, NVIDIA, Cruise CEO Kyle Vogt and Notion co-founder Simon Last, the company said in a Thursday (Sept. 7) blog post. This significant investment will accelerate the development of AI systems that can reason and code.

Imbue’s primary objective is to imbue computers with intelligence and human values, aiming to revolutionize the concept of personal computers, according to the post. The company envisions computers as intelligent tools that empower individuals. Imbue’s ultimate goal is to build practical AI agents capable of accomplishing larger goals and safely operating in the real world.

To achieve this, Imbue adopts a “full stack” approach, encompassing various aspects of AI development, the post said. The company focuses on training foundation models optimized for reasoning, utilizing large language models and reinforcing good reasoning patterns through pre-training data. Additionally, the company invests in techniques that enhance the robustness and reliability of AI agents during inference time.

Imbue’s work extends beyond training models to prototyping experimental agents and interfaces, per the post. The company believes that AI agents should work seamlessly alongside humans, necessitating the invention of tools and interfaces to facilitate collaboration. Imbue acknowledges the significance of coding in improving the reasoning abilities and action-taking capabilities of AI agents. By training agents to code, Imbue aims to create more effective and successful agents capable of accomplishing complex tasks.

Research and theory are also integral to Imbue’s approach, according to the post. The company emphasizes the need for a theory of deep learning to develop models that provide a solid foundation for AI agents and ensure long-term safety. Imbue’s researchers have published work on the theoretical underpinnings of self-supervised learning and the fundamental laws governing the learning process in neural networks. By understanding the core mechanisms of learning, Imbue can engineer better models and advance the understanding of theory and safety in AI.

While initially focused on developing reasoning models for its use, Imbue envisions a future where anyone can create their own AI agents, the post said.

This news comes on the heels of the Aug. 30 announcement that AI and large language model (LLM) company AI21 Labs raised $155 million in Series C funding to develop further and provide generative AI services to businesses.