Autoscience Raises $14M to Advance Autonomous AI Research Lab

Autoscience, an applied research laboratory based in San Mateo, California, has raised $14 million in seed funding to further the automation of artificial intelligence (AI) research and development. The investment, led by General Catalyst with participation from Toyota Ventures, Perplexity Fund, S32, and MaC Ventures, will support the expansion of Autoscience’s virtual AI lab comprised entirely of non-human AI scientists and engineers.

The company claims its AI-powered laboratory can invent, validate, and deploy state-of-the-art machine learning models without direct human intervention. Its approach aims to address what Autoscience identifies as the key bottleneck in the current AI landscape: the human limitation in generating and testing new ideas at scale. By relying on AI systems to perform much of the experimentation and iteration, the company hopes to accelerate the typically lengthy process of model discovery and validation.

Autoscience’s CEO, Eliot Cowan, stated, “We’ve reached a point where human intuition is no longer enough to navigate the complexity of algorithmic discovery. We’ve built a research organization where the researchers are AI systems. We aim to compress a decade of machine learning research into months, unlocking new AI capabilities for scientists and forming a competitive edge for our customers.”

This funding round follows Autoscience’s recent milestone: the publication of a full-length research paper generated end-to-end by its AI agent, Carl, which successfully passed peer review last year. The achievement demonstrates the growing feasibility of AI systems handling increasingly sophisticated tasks, including academic research and development.

The initiative reflects broader trends in the field, as both private industry and public institutions experiment with automating parts of the AI development pipeline. Companies and research organizations have been exploring how AI and automation could help manage the increasing complexity of machine learning, which often involves vast numbers of experiments, simulations, and iterative testing.

Commenting on the investment, Yuri Sagalov, Managing Director at General Catalyst, said, “We believe Autoscience is tackling an increasingly important challenge in machine learning: the pace and scalability of experimentation. As research output continues to grow, teams are looking for ways to more efficiently test, validate, and translate new ideas into production systems.”

Industry leaders have also highlighted the significance of autonomous R&D. In a speech at the 2024 World Economic Forum in Davos, Anthropic CEO Dario Amodei noted the transformative potential of AI systems that build other AI systems, suggesting this self-improving loop may shape the future of the field.

However, automating research with AI presents new challenges. Ensuring the reliability, safety, and reproducibility of models generated autonomously remains an open problem. There are also resource concerns, as large-scale automated experimentation requires significant computing power, potentially limiting access to well-funded organizations.

Despite these hurdles, developments like Autoscience’s virtual lab signal the rising influence of AI-driven automation within science and technology. Rather than replacing human expertise, these systems are positioned as tools to help researchers innovate faster and tackle more complex questions.

For further details, visit hpcwire.com.