Capturing Knowledge: From Cave Art to Artificial Intelligence Models
Retaining knowledge, particularly complex or vast amounts, is a hurdle humanity has been trying to overcome for centuries. As remarkable as it is, the human brain has limitations when it comes to remembering and retrieving information. As a result, we’ve always been searching for methods and tools to aid us in capturing, recalling, and making sense of knowledge.
Capturing Knowledge
The saga of our struggle and innovation in capturing knowledge traces back to the earliest days of human civilization. From the primal etchings on cave walls where our ancestors chronicled their lives and the world around them to the scribbled marginalia by monks and scholars in the pages of ancient texts, we’ve always sought not just to consume but also record and remember.
The advent of the Renaissance witnessed scholars turning their books into interactive knowledge repositories, complete with their personal insights, thoughts, and observations. As the Enlightenment period rolled around, scholars adopted the practice of diligently copying informative passages into notebooks for future reference, arguably one of the earliest instances of formal note-taking.
Contemporary Note-Taking Systems
Fast forward to modern times, and our methods have evolved and become more systematic. We’ve seen the birth of systems like the Cornell Method, which encourages active engagement with learning material through distinct sections for cues, notes, and summaries. The visually appealing mind maps, which graphically represent interrelated concepts to aid comprehension and recall, have also become a mainstay.
Yet, as effective as these methods may be, they all share one common drawback — they are incredibly time-consuming. While these detailed processes can enhance learning and recall, they demand a significant investment of time and patience, something not everyone can afford. And even for those who can make the trade-off, they often find themselves torn between spending more time diving deeper into new information or revisiting and better organizing their existing notes.
Technological Solutions
With the advent of the digital age, digital note-taking tools have become increasingly popular, aiming to harness technology to make the process more efficient and organized. From basic tools like Microsoft OneNote or Google Keep to more specialized apps like Evernote or Notion, digital note-taking methods offer exciting advantages, including easier editing, faster searching, and storing and organizing vast amounts of information.
However, while these digital solutions have made note-taking more convenient in many respects, they still pose challenges. Most notably, these digital solutions can still be quite time-consuming. To make the most out of these systems, users often need to invest considerable time learning how to use them effectively and organizing their notes in a way that best suits their needs.
Moreover, a significant issue arises when reading physical books. Using these note-taking systems can be cumbersome if you're not reading digitally. Readers often resort to scribbling notes in the book margins or on separate pieces of paper, which they later have to transcribe into their digital note-taking system manually. This process is tedious, disrupts the flow of reading, and can significantly extend the time it takes to finish a book.
Digital Search
Furthermore, even once all the information is digitized, searching remains a formidable task unless a complex tagging system is in place. Typical digital search functionalities require you to pinpoint the exact term used in the text (lexical or keyword search). So, if an alternate word was used, your search may miss the intended note entirely.
Adding insult to injury, navigating through the multitude of notes can be overwhelming even when you manage to search with the right term. When dealing with a large volume of notes, each search can return an extensive list of results. The only viable option is to painstakingly read through each note that came up in the search to find the specific information you’re seeking. This once again leads to a significant drain on your time.
The Ideal System
While digital note-taking tools have undoubtedly improved the organization and searchability of our notes, they’re not without their limitations, and they haven’t fully addressed the challenges of time and effort commitment, especially when dealing with physical books.
There remains a significant need for a solution that can bridge the gap between the physical and digital worlds, making note-taking efficient, effective, and seamlessly integrated into our reading process, regardless of the format of our reading material. The perfect solution would address all these challenges.
This system would cater to digital readers and those who prefer physical books, eliminating the taxing task of digitization. While impressive, current technologies like Optical Character Recognition (OCR) systems are not without flaws. They help digitize the text, but this process can be time-consuming. OCRs are prone to errors and often output poorly formatted text, causing yet another layer of work for the user — proofreading and reformatting.
The dream of such an intuitive, seamless, and accurate system has seemed like a fantasy. As a devoted reader, I have grappled with these limitations and frustrations for years. I have tried every method and system available, from traditional note-taking to the latest digital tools. Each time, however, I’ve run into the same stumbling block: the time and effort required to maintain these systems is substantial, detracting from learning and absorbing information from the books I love to read.
But if I didn’t make those systems, then it was almost impossible to retain the knowledge from the books I was reading effectively. Even if I highlighted sections or made notes, the lack of categorization made searching them incredibly tedious. It's a very frustrating dilemma.
Machine Learning
Only in the last couple of years has technology advanced enough to bring us closer to this elusive ideal. The dawn of machine learning has opened up new possibilities, presenting us with unthinkable solutions just a few years ago. But even so, creating a system that harmoniously combines all these elements remains daunting.
Even though I’m now immersed in AI, I must admit that I’ve always approached the field with a degree of skepticism. With a background in psychology, cognitive science, and philosophy, I’m all too aware of the profound complexity of the human mind and the vast gulf that separates even the most advanced AI from human intelligence.
The history of AI is littered with overblown predictions and missed deadlines. A famous example comes from Marvin Minsky, a pioneer of AI research, who in 1966 assigned an undergraduate student the “summer project” of solving computer vision. Minsky assumed the problem was simple enough to be solved within a few months. Over half a century later, despite major advancements, computer vision still isn’t fully solved. Although in the last decade, we have made significant progress.
No one encapsulated this skepticism of AI more than philosopher and cognitive scientist Hubert L. Dreyfus. In his book “What Computers Can’t Do: A Critique of Artificial Reason” he argued that AI’s reliance on formal symbolic manipulation, removed from the full context of the real world, was a dead end. He suggested that human intelligence couldn’t be reduced to formal rules and procedures.
Dreyfus’s critique was heavily influenced by the philosophical tradition of phenomenology, which emphasizes the importance of embodied experience and context in understanding human cognition. In a nutshell, Dreyfus argued that computers lack the necessary ‘worldliness’ and embodied existence to understand or replicate human thought truly.
Recently, Erik J. Larson echoed and expanded upon Dreyfus’s arguments in “The Myth of Artificial Intelligence: Why Computers Can’t Think the Way We Do”. Larson argues that the current approach to AI is fundamentally flawed because it underestimates the complexity of human cognition and overstates the capabilities of machines. He contends that computers, as they currently exist, cannot genuinely understand, make judgments, or engage in creative thought.
I share these concerns, which are very valid and have few good counterarguments. However, my skepticism was soon to be changed. Not necessarily by new philosophical foundations but by reframing the topic and focusing on the pragmatic utility of machine learning.
Large-Language-Models
Despite my skepticism towards AI, the advent of Large Language Models (LLMs) from Open AI has drastically altered my perspective. To be clear, I still hold many of the critiques discussed earlier. LLMs don’t exhibit ‘intelligence’ in the same sense humans do. They don’t have anything resembling consciousness and don’t genuinely ‘understand’ text, neither the inputs nor what they spit out.
However, they are excellent at one thing: processing and generating human-like text based on vast amounts of data. And they do it so convincingly that, for many practical purposes, it almost stops mattering whether it’s ‘real’ intelligence.
Each new version of GPT represented a quantum leap in the model’s capacity to generate human-like text. These models were trained on vast amounts of text data, learning to predict what word will likely come next given a particular context. As a result, they became remarkably good at mimicking human writing styles, understanding context, and providing coherent and relevant responses to a wide array of prompts.
The power of these models is nothing short of astonishing. We’re witnessing an exponential growth in AI’s capabilities. I genuinely feel like we’re seeing decades of progress compressed into a few months. The potential of LLMs is vast, and I believe we’re just scratching the surface of what they can achieve.
Ever since I discovered Chat-GPT, I haven’t stopped using it. It was just too good. Not perfect, and sometimes horrible. But most of the time, it was an incredible tool. It is used more and more, culminating in everyday use. Whenever GPT was down, I felt as if the electricity went down. I either couldn’t do what I needed to do, or now I had to resort to what it felt like an “stone-age” method that took me five times as long.
Over time, I learned what it was good for and wasn’t, how to optimize my prompts, and what to expect with a given content. I was completely hooked, and I currently have hundreds of hours of using GPT, along with other LLMs I found along the way, each with its strengths and limitations.
The Magic Solution Was Found
With this newfound potential, I realized that LLMs could solve my long-standing frustration with traditional note-taking methods. By harnessing the power of AI, we could create a system that simplifies note-taking, making it easy to capture, categorize, and index notes in a digital environment. This was the birth of Raven Notes.
I recognized that LLMs held the key to resolving the age-old note-taking problem. By integrating this technology into a user-friendly platform, we could automate the most time-consuming aspects of note-taking, making the process more efficient and less labor-intensive.
A Revolution in Note-Taking
This made me create Raven. An AI-powered note-taking system that revolutionizes how we interact with knowledge from books. It leverages the power of LLMs to provide an intuitive, efficient, and, most importantly, effortless note-taking experience, bridging the gap between physical and digital worlds.
Working with Raven is simple. Whether reading a physical or a digital book, you just need to highlight the passages you’re interested in and send them to Raven. If you’re dealing with a physical book, you can snap a photo of the highlighted section and send it to the platform. The built-in OCR system in Raven is highly effective in converting these images into digital text and uses AI to clean up any mistakes.
This is where the magic begins. Raven, powered by LLM, processes the highlighted sections, extracts the key points, and structures the notes coherently and concisely. It doesn’t just transcribe your highlights; it intelligently digests the information and summarizes it in a way that is easy to understand and remember.
Raven goes beyond mere summarization. It uses its advanced AI to categorize the notes, making them easy to search and cross-reference later. By identifying the core concepts and ideas in your notes, it organizes your notes to find exactly what you’re looking for with just a few clicks. No other existing note-taking system can do this, and it drastically cuts down the time you spend organizing and searching through your notes.
Raven is also capable of cross-referencing your notes across multiple books. This is especially useful when studying a particular topic from multiple sources. Raven’s AI understands the relationships between ideas and concepts and can connect the dots between different notes, giving you a more holistic understanding of the subject matter.
The Holy Grail
In essence, Raven is an innovative tool that revolutionizes the note-taking process. It solves the problem I just discussed by transforming an otherwise labor-intensive task into an effortless, seamless process. By bridging the divide between the physical and digital worlds, Raven allows for efficient note-taking, regardless of whether your source material is a physical or digital book.
This remarkable advancement enables us to concentrate on what truly matters: delving into and learning from the rich knowledge presented in our chosen books.
An Evolutionary Leap in Learning
With Raven, we usher in a new era in note-taking, finally offering a tangible solution to the longstanding challenge of wanting to make comprehensive notes without the associated time drain. Raven is far more than just a mere upgrade to current methods. It symbolizes an entirely new approach to capturing and interacting with our notes.
This shift has been made possible due to our tremendous strides in AI and LLMs. Although the quest for the perfect note-taking system continues, I firmly believe that with Raven, we’ve taken a significant step toward that goal.
Getting Started
Raven is in its development phase and unavailable for public use. But we will soon start onboarding the first batch of beta testers.
By participating in our early access program, you will be at the forefront of a community dedicated to advancing how we capture, process, and utilize knowledge. Your insights will help us tailor Raven better to serve all readers, students, and academics worldwide.
Embark on this journey with us at Raven. Visit www.ravenotes.com to join our early access list and be part of crafting the future of note-taking.
