[Adobe Stock]
A 2016 survey by Nature found that over 70% of researchers have tried and failed to reproduce another scientist’s experiments, and more than half have failed to reproduce their own experiments. A significant challenge that contributes to this crisis is the difficulty scientists face in understanding the experimental parameters across studies. “The fundamental problem was the massive amount of information coming in for any given scientist and the challenge of distilling that information, extracting all the necessary details and actually being able to run those experiments,” said Nick Edwards, CEO of Potato AI.
This challenge is not new to science. Consider Thomas Edison’s 14-month search for the perfect light bulb filament in 1878–1879. What Edison estimated would take three or four months turned into exhaustive testing of thousands of plant materials—from baywood to bamboo—before he found success. Today’s researchers face similar obstacles, but with exponentially more information to process. “We fundamentally believe that by helping to increase the reproducibility of science, we can help it move faster, using AI tools to do that,” Edwards said.
Reading a single dense scientific article can take hours, and comparing multiple studies adds to the difficulty. While pairing large language patterns with web search may seem like an obvious solution, it often introduces new problems. Large general-purpose language models can hallucinate or fabricate references on their own, while Internet searches return a mix of verified and unverified information. “Not everything on the Internet is true,” noted Ryan Kosai, CTO of Potato AI.
Why Wiley and Potato are partners
Potato AI’s new partnership with Wiley, one of the world’s largest scientific publishers, provides access to an essential resource: peer-reviewed scientific content. By implementing Retrieval-Augmented Generation (RAG), Potato AI can base its AI responses on verified scientific literature. This technique addresses the major limitations of traditional large language models by improving factual accuracy, contextual understanding, and information retrieval.
“Researchers and practitioners are looking for more than generic AI tools—they need relevant applications that enhance and support their research efforts,” explained Josh Jarrett, Senior Vice President and General Manager for AI Advancement at Wiley, in a press release from mid-October. “This new partnership program is designed to meet these needs by inviting collaboration with start-ups and scale-ups to deliver specialized AI solutions.”
In essence, RAG offers a promising alternative to traditional genAI techniques by pairing generative capabilities with a retrieval system that retrieves information from curated and trusted databases. Custom building a large-scale RAG implementation tailored for scientific research presents significant technical challenges. The vector databases required for comprehensive science RAG systems can be extremely cumbersome when using performant embeddings, requiring significant computational resources to manage and query effectively. Additionally, accessing comprehensive and authoritative resources often requires navigating complex subscription models and ensuring compliance with licensing agreements.
That said, when it works well, RAG can result in substantial improvements in the accuracy and reliability of AI-generated content. Potato RAG can “create generative content that links individual components within the protocol’s experimental guidelines to literature references,” Edwards noted.
Current capabilities and features
Nick and Ryan from Potato at TechCrunch Disrupt.
One of the current capabilities of the potato is automated paper review. The platform enables researchers to enter any scientific paper or laboratory document for comprehensive analysis. “It helps distill the methods they used, some of the results, and helps evaluate the things they were controlling for,” Edwards explains. The system goes beyond simple summarization by generating experimental insights and hypotheses for future testing.
Next is protocol generation. Researchers can ask plain-language questions about experimental procedures, and the system responds with detailed, actionable protocols. “You can ask questions like, ‘I want to purify this protein,’ and it will pull relevant research protocols across thousands of open-access documents to build detailed, reproducible methods—step-by-step instructions that can connect them again. references,” Edwards said.
The platform also includes literature-based query capabilities that scan vast bodies of literature and open-access databases to surface existing scientific knowledge. This feature helps researchers quickly find specific information without having to manually review dozens of documents.
Results so far
The impact of these capabilities is already evident in real-world research environments. One example comes from Edwards’ own research experience with brain slice preparation. His lab found that replacing sodium with another chemical could double the lifespan of brain slices. “It’s a very simple solution and it dramatically increases the time you have to do experiments,” explains Edwards. “These small details don’t just shave off days or months—they allow faster iteration cycles between experiments.”
Another example involves a researcher who discovered that potato AI could have saved nine months of Ph.D. the work that had been expended to optimize a specific reagent through repeated testing.
“This is real-time acceleration – it speeds up research, lowers costs and reduces capital expenditure if something goes wrong,” Kosai noted.
Plans to build an AI scientist
The long-term vision for Patato AI extends beyond his current role as a research assistant. “The long-term vision here is to build an AI scientist,” Edwards explained. “For now, we’ve built an AI research assistant that helps with some of these functions, but we very much envision a future where it can help automate the experimental process.”
This evolution would involve multiple aspects of the scientific process. “If you think about what that means – an AI scientist is something that can help with hypothesis generation, detailed experimental planning, actually running experiments. These could be computational experiments or things done in the lab,” Edwards added. An AI scientist’s dreams will take time to come true. “For now, we are focused on enabling scientists to enhance their skills and accelerate their research.”
But in the long run, scientific research is likely to be a collaborative model between human knowledge and AI capabilities. “In a few years, a lot of the best research will probably have a combination of human thinking — the human driving the direction of things — while leveraging AI tools to do some of the work,” Kosai predicted.
With the backing of Axial, Pioneer Square Labs, the Allen Institute for Artificial Intelligence and angel investors, Potato AI is positioned to pursue this vision of accelerated scientific discovery. The goal isn’t just to save time or money—it’s to fundamentally transform the way science is done.