Photo: Eugene Mymrin/Getty Images
Autoscience, which automates the generation of new machine learning models, has raised $14 million in seed funding.
General Catalyst led the round, with participation from Toyota Ventures, MaC Ventures, Perplexity Fund and S32.
WHAT IT DOES
The California-based company has built an AI system that autonomously conducts research. Autoscience's virtual laboratory uses non-human AI scientists to invent, validate and deploy machine learning models for research and development, with its first deployments focused on financial applications, manufacturing and fraud detection.
The company says it uses "two core AI systems: automated scientists that ideate and test new algorithmic hypotheses and automated engineers that optimize and deploy those validated inventions into the real world."
Autoscience will use the funds to scale its engineering workforce and to offer Fortune 500 and large private companies that plan to train machine learning models in high-stakes environments.
"We’ve reached a point where human intuition is no longer enough to navigate the complexity of algorithmic discovery," Eliot Cowan, CEO of Autoscience, said in a statement. "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."
MARKET SNAPSHOT
Autoscience claims to have published the first academically peer-reviewed paper authored largely by its AI agent, Carl, with only minor human edits for formatting and citations.
According to R&D World, the company reports that Carl also authored a full-length paper, "Investigating Alignment Signals in Initial Token Representations," which was accepted at an International Conference on Learning Representations (ICLR) workshop.
Another company, Sakana AI in Tokyo, also claims to have passed peer review, with a paper focused on AI entitled, "Compositional Regularization: Unexpected Obstacles in Enhancing Neural Network Generalization," Sakana also submitted its paper to a workshop at ICLR.
