AI platforms have shown their capability in various areas, yet the actual desired use case leans towards automating monotonous tasks, a common occurrence in research and academic circles. Reliant aims to carve a niche in executing the labor-intensive job of data extraction, traditionally undertaken by weary graduate students and interns.
Karl Moritz, the CEO, believes “The optimal use of AI is to enhance the human condition: minimizing repetitive work and freeing individuals to pursue meaningful activities.” This is particularly true in the research domain, he points out, where he, along with co-founders Marc Bellemare and Richard Schlegel, have extensive experience, especially concerning the exhaustive process of literature review.
Referencing literature in research papers is a rigorous task due to the vastness of scientific works, with some reviews citing or utilizing data from several thousands of works.
In a particular study, Moritz observed, “Authors sifted through 3,500 scientific papers, many of which proved irrelevant. The enormous time investment for a sliver of valuable information signifies a prime process for AI to streamline.”
Recent tests with language models like ChatGPT, which processed data with an 11% error rate, showcase the potential, despite not meeting actual necessity standards.
“But that’s far from sufficient,” remarked Moritz. “Accuracy is crucial for these tedious academic tasks, where errors are inadmissible.”
Tabular, Reliant’s main offering, leverages a version of LLM (LLaMa 3.1) enhanced with proprietary methods, proving vastly more accurate. In the aforementioned document analysis task, their technology achieved a zero-error rate.
Essentially, you input a batch of documents, define the data to extract, and Reliant meticulously identifies and analyzes it, regardless of documentation structure, displaying results in a user-friendly interface for detailed exploration.
Moritz underlines the objective: enabling users to interact with and understand vast data sets comprehensively, facilitating a direct transition from aggregated data to specific literature.
This nuanced and practical AI utilization could notably expedite advancements across diverse scientific fields. The startup has attracted a $11.3 million seed investment led by Tola Capital and Inovia Capital, with contributions from angel investor Mike Volpi.
Reliant’s approach necessitates substantial computing power, prompting the decision to purchase rather than rent computing resources. This move, while financially ambitious, grants them the flexibility to deeply explore specialized problem areas.
Moritz elucidates the challenge of delivering precise answers in constrained timeframes, emphasizing pre-emptive solution formulations for potential user queries, significantly enhancing efficiency and accuracy.
Proactively handling data extraction and analysis not only streamlines the process but also allows for adjustment and improvement before user interaction, a smart allocation of computing resources.
Pre-analyzing data aids in clarifying ambiguities inherent in the vast scope of scientific research, a critical step in delivering reliable, unambiguous results, as Moritz highlights.
Reliant prioritizes demonstrating the economic viability of their technology, mindful of balancing ambitious goals with tangible, marketable products, as outlined by Moritz.
Despite potential competition from giants like OpenAI and Anthropic, Bellemare sees their proprietary technology and unique data sources as a substantial advantage in the evolving AI landscape of biotech and research.
Reliant has strategically positioned itself in the burgeoning AI-driven transformation of research and biotech, setting a solid foundation for growth and innovation.
Moritz concludes with a pragmatic outlook, focusing on precision and accuracy for clients where they matter most. This meticulous attention to detail separates them from the competition, emphasizing quality over broad applicability.
Compiled by Techarena.au.
Fanpage: TechArena.au
Watch more about AI – Artificial Intelligence
