The Next Chapter of Readwise: Our Own Reading App - Inside a radical new project to democratize AI

The Next Chapter of Readwise: Our Own Reading App - Inside a radical new project to democratize AI

In the age of technology, where information is readily available at our fingertips, reading remains a fundamental practice for personal and professional growth. However, despite spending hours immersed in books, we often struggle to retain the knowledge we acquire. This problem begs the question: why doesn't software help us internalize the valuable insights we come across?

At Readwise, a company dedicated to improving the practice of reading, we have recognized the need for a solution. Our mission is to enhance the reading experience through software by an order of magnitude. To achieve this, we have developed our own reading app, designed to address the challenges of retention and content overload.

The idea of building our own reading app was born out of the demands of our users. We observed that while technology has disrupted traditional publishing and distribution models, the experience of reading itself has remained largely unchanged. Physical books continue to outsell e-books, indicating a persistent preference for the tangible reading experience ("Physical Books Still Outsell E-Books — And Here's Why," Lucy Handley). The advent of e-readers and reading apps in the late 2000s seemed promising, but in retrospect, it appears that these innovations primarily focused on content distribution rather than advancing the practice of reading.

To tackle the issue of retention, we prioritize what happens after reading. We believe that to extract the maximum value from reading, we must first address the problem of forgetting. Our reading app is built on a local-first foundation, enabling lightning-fast, full-text search capabilities across your entire library of books, articles, annotations, and highlights. With this feature, you can effortlessly find specific information, even if you only remember a single word. Moreover, our app operates offline, ensuring uninterrupted access to your reading materials.

Another challenge we aim to overcome is content overload. In today's digital age, we often save more reading material than we can realistically consume. Consequently, our digital gardens become digital graveyards, with an overflowing inbox of articles and books waiting to be read. To combat this, our app integrates highlighting as a first-class feature. Unlike other reading apps, which treat highlighting as a secondary function, we recognize its importance in extracting meaningful insights. By encouraging users to write between the lines, we enhance the reading experience and facilitate knowledge retention.

While many platforms strive to make reading a social experience, we have consciously chosen a different path. Rather than attempting to create a vast internet-scale annotation layer, we focus on improving the individual reading experience. As toolmakers, we see immense potential in enhancing single-player mode before exploring multiplayer functionality.

Now, shifting our focus to the field of artificial intelligence, we encounter another groundbreaking project that aims to democratize the development of AI models. The project, called BigScience, brought together over 1,000 volunteer researchers to create BLOOM (BigScience Large Open-science Open-access Multilingual Language Model). Unlike other famous large language models, such as OpenAI's GPT-3 and Google's LaMDA, BLOOM prioritizes transparency. The researchers openly share details about the data it was trained on, the challenges faced during development, and the evaluation of its performance.

BLOOM's accessibility is a significant advantage. It is available for anyone to download and experiment with for free on Hugging Face's website. Developers can leverage this model as a foundation to build their own AI applications. With 176 billion parameters, BLOOM surpasses OpenAI's GPT-3 in size and offers similar levels of accuracy and toxicity control.

What sets BigScience and Hugging Face apart is their commitment to openness and responsible AI. While other tech giants restrict access to their models, BigScience and Hugging Face have taken steps to provide transparency and accountability. Meta, for example, released its large language model (OPT-175B) only upon request and limited its use to research purposes. In contrast, Hugging Face goes a step further by introducing the Responsible AI License. This license acts as a deterrent against using BLOOM in high-risk sectors or for harmful purposes, ensuring ethical usage.

BigScience's ethical guidelines played a significant role in shaping BLOOM's development. These guidelines served as guiding principles for the project, ensuring that the model aligns with ethical considerations and avoids perpetuating biases present in other models. By rallying volunteers from around the world, BigScience was able to build data sets in various languages, even those not well represented online.

While some companies argue that the sexist and racist language embedded in their models makes them too dangerous for public use, the creators of BLOOM believe that having an open language model for research purposes has a long-term positive impact.

In conclusion, both Readwise and the BigScience project represent exciting advancements in their respective fields. Readwise aims to revolutionize the reading experience by addressing retention and content overload through their own reading app. Meanwhile, BigScience seeks to democratize AI development by creating a transparent and accessible language model. By incorporating these actionable insights into our own lives, we can enhance our learning and stay at the forefront of knowledge in an ever-evolving world.

Actionable advice:

1. Prioritize retention: Use tools like Readwise to improve your knowledge retention by organizing and easily accessing your library of books, articles, annotations, and highlights.
2. Embrace highlighting: Make highlighting a core part of your reading process. By actively engaging with the material and writing between the lines, you enhance comprehension and facilitate knowledge internalization.
3. Explore responsible AI: Stay informed about the latest developments in AI and explore projects like BigScience that prioritize transparency and ethical considerations. Understand the potential of AI models and leverage them responsibly for research and innovation.