Rapidly Digesting Documents Using AI with Humata’s Cyrus Khajvandi and Dan Rasmuson

In this episode of the FYI podcast, Cyrus Khajvandi and Dan Rasmuson, co-founders of Humata, discuss the innovative AI tool that revolutionizes document interaction. With a focus on accelerating scientific discovery and knowledge transfer, Humata allows users to chat with their documents, ask questions, and receive fact-checked answers. The tool has found applications in various sectors, from academia to legal and R&D, enhancing decision-making by ensuring information accuracy. The co-founders share their backgrounds, the inception of Humata, and its potential to transform how professionals across industries work with documents. They also touch on the challenges of AI's truthfulness and the importance of human oversight in managing the increasing volume of encoded organizational information.

Takeaways

• 🚀 **Innovative AI Tool**: Humata is a groundbreaking AI tool designed to digest documents rapidly, providing summaries and answers to natural language queries directly from the text.
• 🤖 **AI's Hallucination Problem**: The creators acknowledge AI's tendency to generate false information (hallucinate), and Humata addresses this by allowing users to fact-check responses with highlighted references and citations.
• 🔍 **Contextual Understanding**: Humata can understand and respond to questions across multiple documents, expanding the context window significantly beyond what traditional AI models can handle.
• 📚 **Academic and Corporate Applications**: Initially aimed at accelerating scientific discovery for researchers, Humata has found applications in various industries, including legal, R&D, and customer support.
• 📈 **User Engagement Metrics**: The success of Humata is measured by user engagement, including the number of questions asked, frequency of use, and the completeness of answers provided.
• 🛡️ **Data Privacy and Security**: Humata prioritizes data privacy and security, offering end-to-end encryption, control over data, and role-based access permissions within organizations.
• 📈 **Knowledge Transfer**: The tool aids in the transfer of specialized domain knowledge within organizations, which is particularly valuable during periods of rapid workforce expansion or retirement.
• 🔗 **Integration and APIs**: There is potential for Humata to be integrated into other services and platforms via APIs, allowing for broader accessibility and functionality.
• ⚙️ **Product Development Focus**: The Humata team focuses on building solutions based on customer feedback, ensuring that the product remains relevant and useful for a wide range of problems.
• 🌟 **Ease of Use**: One of the key selling points of Humata is its ease of use and quick setup, allowing users to start benefiting from the tool without extensive onboarding or technical expertise.
• ✨ **Transformative Impact**: Users have reported that Humata has transformed their productivity and learning by making it easier to comprehend, learn from, and work with documents.

Q & A

What is the main focus of the Humata tool discussed in the podcast?

Humata is a tool designed to accelerate scientific discovery and knowledge transfer by allowing users to chat with their documents, ask questions using natural language prompts, and get faster summaries. It is particularly useful for academic researchers, legal teams, and various industries that require quick access to accurate information across numerous documents.

How does Humata address the truthfulness problem associated with AI?

Humata tackles the truthfulness problem by providing a side-by-side comparison of the user's document or documents along with highlighted references and citations. This allows users to fact-check information on the spot, ensuring the correctness of the answers they receive.

What is the background of Cyrus Khadjvandi, the CEO and co-founder of Humata?

Cyrus Khadjvandi has founded several companies before Humata. His previous ventures include a company in the crypto space and a biotech startup called Denovo, which was based on cellular reprogramming research conducted at Stanford and aimed at finding a potential cure for hair loss.

What was the initial problem that Dan Rasmuson identified, leading to the creation of Humata?

Dan Rasmuson identified the difficulty of staying on top of advanced scientific journals and publications as a researcher at Stanford. He noticed that this was a common challenge across different levels of academia and research. This led to the idea of combining AI's proficiency in reading and writing with the need for efficient information processing.

How does Humata's approach to document interaction differ from other AI solutions?

Humata differs by enabling users to ask questions across many different documents simultaneously, expanding the context window. It also allows users to trace answers back to the source document, reducing the risk of misinformation or 'hallucination' that can occur with AI.

What are some of the use cases for Humata outside of academic research?

Beyond academic research, Humata is used in various industries such as legal for creating medical chronologies or handling bankruptcy cases, by oil and gas companies for training and decision-making in the field, and by sales teams for complex products to provide customers with accurate information and documentation.

How does Humata solve the context window problem in AI models?

Humata addresses the context window problem by creating various artifacts from the documents and using an agent to decide what content to include in the context based on the user's question. This process requires intelligence and is designed to work effectively for different types of queries.

What is the significance of the 'Ask Every Page' feature in the legal space?

The 'Ask Every Page' feature is significant for legal professionals as it allows them to create thorough timelines from documents, such as in personal injury cases, without losing important contextual information that could be crucial for building a case strategy.

null

null

How does Humata facilitate knowledge transfer within organizations?

Humata facilitates knowledge transfer by enabling users to quickly access and verify information from a vast array of documents. This is particularly useful when experienced workers retire, as their expertise can be retained and made accessible to newer recruits.

What are some of the technical challenges Humata faces in providing accurate and relevant information?

Technical challenges include effectively managing the context window to ensure AI models can handle increasingly larger documents, retrieving the correct information from long inputs, and maintaining performance as more documents are added to the system.

How does Humata ensure data privacy and security for its users?

Humata ensures data privacy and security by end-to-end encrypting all data, allowing users to own and control their data, providing the ability to delete documents from the system, and isolating data per organization to prevent data leakage.

What is the future roadmap for Humata in terms of product development and market expansion?

Humata's roadmap includes developing tools for organizations to analyze their usage and knowledge surfaced, expanding integrations into more services where data is stored, and focusing on quality enhancements. They also aim to announce new products that build upon their existing platform to support the delivery of work products.

Outlines

00:00

🎙️ Introduction to FYI Podcast and Humata

The podcast FYI, focused on technological disruption, welcomes its audience and emphasizes the importance of understanding innovation for investment. It features an interview with Dan and Cyrus, co-founders of Humata, a tool designed to accelerate scientific discovery by summarizing and answering questions about documents using AI. The co-founders discuss their backgrounds and the genesis of Humata, highlighting its ability to combat AI's 'hallucination' problem by providing factual, citable information.

05:01

🔍 Use Cases and Differentiators of Humata

The discussion explores various use cases of Humata, including legal and medical chronologies, and the importance of fact-checking in making informed decisions. The co-founders explain Humata's unique ability to manage context across numerous documents and its application in training and decision-making processes. The conversation also touches on the technical aspects of expanding the context window for AI models and the challenges of maintaining performance with large document inputs.

10:02

🚀 Navigating Rapid AI Advancements at Humata

The co-founders discuss the fast-paced evolution of AI and how Humata is guiding its resources and development. They emphasize building for customer needs, gaining insights into core problems, and focusing on the transition towards AGI. The conversation also addresses the challenges of upskilling novice salespersons and the potential of AI to streamline knowledge transfer within organizations.

15:03

📚 Knowledge Transfer and Economic Implications

The discussion highlights the use of Humata for knowledge transfer, particularly in specialized domains, and the economic benefits it presents for organizations dealing with retiring experts. The co-founders share an example of a major oil and gas company using Humata to manage the loss of expertise due to workforce retirement. They also touch on the broader economic implications of AI in reducing the reliance on high-earning employees.

20:03

🌐 Impact of Foundation Models on Humata's Development

The co-founders reflect on the rapidly changing landscape of foundation models and their impact on Humata's roadmap. They discuss the advancements made by OpenAI and how Humata differentiates itself by focusing on delivering assistance and retrieval capabilities. The conversation also explores the potential commoditization of certain AI functionalities and the importance of specialization for startups in the AI space.

25:05

🛡️ Addressing Data Privacy and Security Concerns

The discussion addresses concerns related to data privacy and security, with the co-founders outlining Humata's measures to protect user data. They mention end-to-end encryption, data ownership by users, and the platform's compliance with SOC 2 standards. The co-founders also discuss the isolation of data per organization and the introduction of team management features for enhanced security and accessibility.

30:06

🔗 Future Integrations and the Evolution of Work with Humata

The co-founders share their vision for future integrations of Humata, including potential APIs and the expansion into more services. They discuss the possibility of enabling not just reading but also writing within the platform, and the importance of providing immediate value to users. The conversation also highlights the potential for Humata to transform how people comprehend, learn, and work on a daily basis.

35:07

📈 Tracking Success and Encouraging Feedback

The final paragraph focuses on how Humata tracks customer success, emphasizing engagement metrics such as the number of questions asked and the quality of answers received. The co-founders express their interest in understanding which departments and individuals within an organization find the most value in using Humata. They also invite user feedback to continuously improve the product and shape its future development.

Keywords

AI

AI, or Artificial Intelligence, refers to the simulation of human intelligence in machines that are programmed to think like humans and mimic their actions. In the context of the video, AI is central to Humata's technology, which uses AI to read, write, and summarize documents, thereby aiding in scientific discovery and decision-making across various industries.

Humata

Humata is a company that has developed a disruptive technology enabling users to interact with their documents through natural language processing, powered by AI. It is positioned as a tool to accelerate scientific discovery and is also being used in various other sectors like legal and customer support, as mentioned in the script.

Natural Language Processing (NLP)

NLP is a field of AI that enables machines to understand, interpret, and generate human language. In the video, NLP is a core technology behind Humata's platform, allowing users to ask questions and receive summaries from documents using conversational language.

Disruptive Innovation

Disruptive innovation refers to a new technology or idea that disrupts an existing market or creates a completely new market, eventually displacing established methods or products. The video discusses how Humata's AI technology is a form of disruptive innovation in the way documents are digested and understood within various professional fields.

Chat with Your Document

This phrase describes the innovative feature of Humata's platform that allows users to have a conversation with their documents as if they were human. It encapsulates the user's ability to ask questions and get answers directly from the text of documents, which is a central theme in the video.

Fact-Checking

Fact-checking is the act of verifying the accuracy of information. In the context of the video, Humata emphasizes the importance of fact-checking with its side-by-side comparison feature, allowing users to compare AI-generated summaries with the original document to ensure the information's accuracy.

Knowledge Transfer

Knowledge transfer is the process of sharing information, expertise, or intellectual capital. The video discusses how Humata can facilitate knowledge transfer, especially in organizations facing challenges with an aging workforce and the impending retirement of skilled employees.

Context Window

The context window refers to the amount of information a system can take into account when generating a response. In the video, it's mentioned that AI models have limitations in handling large context windows, and Humata addresses this by intelligently selecting content for inclusion in the context to provide accurate responses.

Enterprise Solutions

Enterprise solutions are services or products designed to address the needs of large organizations. The video highlights how Humata's technology is being used by enterprises for various applications, including legal, R&D, and customer support, showcasing its adaptability as an enterprise solution.

Data Privacy and Security

Data privacy and security involve protecting information from unauthorized access, use, or disclosure. Humata emphasizes end-to-end encryption and data ownership by the user, addressing concerns about privacy and security, which is a critical aspect when handling sensitive corporate documents.

API

An API, or Application Programming Interface, is a set of protocols and tools that allows different software applications to communicate with each other. The video briefly touches on the idea of Humata potentially offering APIs to integrate its document-interaction capabilities into other services and platforms.

Highlights

Humata is a tool that uses AI to digest documents, providing faster summaries and enabling natural language queries.

Cyrus Khajvandi, CEO of Humata, has a background in cellular reprogramming research and has founded several companies.

Dan Rasmuson has a decade of experience in software, including working with AI and founding the machine learning labeling tool, Labelbox.

Humata was initially created to solve the problem of staying updated with scientific journals, a challenge faced by researchers and academics.

The platform provides a side-by-side comparison of documents with highlighted references for immediate fact-checking.

Humata's users include academic researchers, attorneys, R&D groups, and customer call support centers due to its versatility.

The ability to trace answers back to the original document helps prevent AI-generated hallucinations or confabulations.

Humata can expand the context window, allowing users to ask questions across many different documents simultaneously.

The platform is particularly useful for legal teams creating medical chronologies or handling bankruptcy cases.

Humata is also valuable for onboarding new employees in challenging processes, such as in oil and gas companies.

The tool helps sales teams of complex products provide accurate information to customers and reference product documentation efficiently.

Humata's approach to solving the context window problem involves intelligent decision-making on what content to include in the context.

The platform is designed to handle long input and retrieve the correct information, a challenge for current language models.

Humata is focused on building for customer needs, gaining insight into core problems, and iterating based on feedback.

The platform is designed to be user-friendly, allowing even non-technical users to leverage its capabilities without extensive training.

Humata ensures data privacy and security by offering end-to-end encryption and giving users control over their data.

The platform is continuously improving, with upcoming features aimed at enhancing quality and user experience.

Humata differentiates itself by offering a ready-to-use solution that doesn't require companies to build and maintain their own AI models.

The tool provides immediate value and can be integrated into various services where organizational data is already stored.

Humata is transforming how people comprehend, learn, and work by making document interaction more intuitive and efficient.