Reliant's paper-scouring AI takes on science's data drudgery | TechCrunch

🌈 Abstract

The article discusses the potential of AI models to automate menial tasks in research and academia, particularly in the area of literature review and data extraction. It highlights the work of a startup called Reliant, which has developed a specialized AI-powered tool called Tabular to streamline these time-consuming processes.

🙋 Q&A

[01] Reliant's Approach to Automating Research Tasks

1. What are some of the menial tasks in research and academia that Reliant aims to automate?

• Literature review and data extraction from large volumes of scientific publications
• Systematic reviews that require analyzing data from thousands of studies

2. How does Reliant's Tabular product differ from the performance of language models like ChatGPT?

• ChatGPT was able to extract data with an 11% error rate, which is not accurate enough for the needs of researchers
• Reliant's Tabular product, which is based on the LLama 3.1 model but augmented with proprietary techniques, was able to perform the same task with zero errors

3. What are the key features of Reliant's Tabular product?

• It can process a large number of documents and extract the desired data, whether it is well-labeled and structured or not
• It provides a user interface that allows researchers to work with all the extracted data at once, edit the data, and link back to the original literature

4. Why does Reliant focus on building its own hardware rather than renting from cloud providers?

• Owning the hardware allows Reliant to better predict and handle the compute-intensive nature of its AI-powered tasks
• It gives the company more control and the ability to optimize the problem-solving process, which is important for providing accurate and timely responses to researchers

[02] Reliant's Vision and Approach

1. What is Reliant's overall vision for applying AI to research and academia?

• The goal is to "improve the human experience" by reducing menial labor and allowing researchers to focus on more important work
• Reliant aims to become a specialized provider of AI-powered tools that can accelerate science across technical domains

2. How does Reliant approach the challenge of ambiguity and domain-specific knowledge in its AI models?

• The company invests in understanding specific scientific domains and building proprietary techniques to resolve ambiguities and assumptions
• This allows the models to provide more consistent and reliable outputs, which is critical for the needs of researchers

3. What is Reliant's strategy for commercializing its technology?

• The company is initially focused on establishing that its tech can pay for itself by serving for-profit companies, as they provide the necessary funding
• Reliant is not interested in selling its products at a loss, but rather in building a sustainable business model

4. How does Reliant view the competition from larger AI companies like OpenAI and Anthropic?

• Reliant sees the improvements in the broader AI ecosystem as beneficial to its own progress, as it can build on top of the advancements in language models and other technologies
• The company believes its domain-specific expertise and proprietary techniques give it a competitive advantage in the research and academic market