π€ Preparing young minds for an AI future
Dear curious mind,
While last week belonged to Anthropic with updates and launches of the Claude models, OpenAI has stepped back into the spotlight with several exciting new features. The AI landscape continues to evolve at a fast pace, and I'm here to help you navigate through these developments.
Before we dive in, I want to remind you about this Sunday's free NotebookLM webinar. I will show you how to leverage NotebookLM to transform your reading, thinking, and creative processes. You'll learn practical techniques to engage with texts more deeply and extract meaningful insights using this powerful AI tool. If you would like to join the webinar or get access to the recording, sign up in this Google Form.
In this issue:
π‘ Shared Insight
Teaching Our Children Responsible AI Use
π° AI Update
OpenAI Brings Advanced Voice Mode to Desktop Apps
OpenAI Makes Chat History Searchable
OpenAI Released Their New Web-Search
Anthropic Brings Voice Inputs to Mobile Claude Apps
Anthropic Released Desktop Apps
π Media Recommendation
Podcast: Aidful Book Summaries
π‘ Shared Insight
Teaching Our Children Responsible AI Use
The widespread adoption of AI tools like ChatGPT has created a significant challenge in education. Students increasingly use these powerful tools to complete their homework assignments, often without their teachers' knowledge or approval. Professor Ethan Mollick states in his article Post-apocalyptic education that we are now facing what he calls the "Homework Apocalypse", a reality where AI can complete most traditional homework assignments, rendering them ineffective as learning tools and assessment measures. A representative survey found that 82% of bachelor students and 72% of school students had used AI for school, with 56% using it for writing assignments and 45% for other types of schoolwork. Many students don't view getting AI help as cheating, though teachers often do.
Detecting AI-generated content has proven extremely difficult. Even linguistics journal editors and experienced teachers perform poorly at identifying AI writing, especially after multiple revisions. Well-prompted AI writing is often judged as more human than actual human writing, making traditional detection methods unreliable.
OpenAI recently made headlines for developing, but not releasing, a tool to detect ChatGPT-generated content. The idea is that by slightly changing the word probability in AI outputs, generated content can be identified. This watermarking of AI-generated content has emerged as a potential solution for the detection of AI-generated content. Recently, Google released their watermarking approach SynthID as open-source and, in theory, every model could use it.
In practice, watermarking alone can't solve this problem. Many AI models, including various open-source variants, don't implement watermarking. Students can easily switch to these alternatives or use these models to reformulate and with that destroy the watermarks in the original generated text.
Instead of focusing on detection, parents and teachers need to adapt their approach. The solution lies in teaching students to use AI responsibly as a learning tool rather than a shortcut. This includes:
Using AI as an assistant that enhances learning rather than replaces thinking
Demonstrating proper AI usage through collaborative classroom exercises
Designing assignments that emphasize understanding and process over final output
Using AI to help explain complex topics in ways that resonate with individual students
The future of education isn't in preventing AI use, but in teaching students to harness it responsibly. By embracing AI as a learning aid while ensuring students understand its limitations and proper role, we can prepare them for a world where AI assistance is increasingly common.
π° AI Update
OpenAI Brings Advanced Voice Mode to Desktop Apps (@OpenAI on π)
Bringing voice chat to desktop apps is a welcome addition, even though I am left behind as a Linux user. But I realized last week that advanced voice mode is now also accessible for European users through the iOS and Android apps.
OpenAI Makes Chat History Searchable (@OpenAI on π)
Finally! The feature is realized as a keyword search. Implementing semantic search would be far more powerful for finding past conversations. I would also love to see basic organizational features like folders and pinned chats to help manage the growing history of conversations.
OpenAI Released Their New Web-Search (OpenAI blog)
The answer to Google and Perplexity from OpenAI is live for Plus and Team subscribers of ChatGPT. You can trigger it with a click on the globe icon or by typing /search. It is even possible to set ChatGPT search as the default search tool in your Chrome browser. All of that and more is explained in the corresponding help page.
Anthropic Brings Voice Inputs to Their Mobile Apps (@AnthropicAI on π)
Voice inputs offer a natural way to interact with AI systems. However, the lack of voice output capabilities for Claude means that the app is so far not truly conversational. Consequently, it is not possible to have conversations where you are unable to read, e.g. while driving a car.
Anthropic Released Desktop Apps (@AnthropicAI on π)
The launch of Claude desktop applications marks an important step toward feature parity with OpenAI. Like for ChatGPT, there is no Linux app, but I think the user base is too small to make this a reality.
π Media Recommendation
Podcast: Aidful Book Summaries
You want to know if a book is worth your time? I have launched an AI-powered podcast that creates one engaging book summary per day.
I started with books I have previously read to ensure quality. Using NotebookLM, I carefully select the best AI-generated audio summaries from multiple outputs. Check out the first episodes and experience these surprisingly enjoyable AI-crafted book summaries for yourself:
Co-Intelligence: Living and Working with AI - Ethan Mollick
Building a Second Brain - Tiago Forte
Million Dollar Weekend - Noah Kagan
This Is Personal: The Art of Delivering the Right Email at the Right Time - Brennan Dunn
Four Thousand Weeks - Oliver Burkeman
I am quite impressed that the AI accurately captures key insights and reminds me of valuable passages I had forgotten over time. It is fascinating how NotebookLM transforms a complete book provided as input into an engaging audio summary that is actually fun to listen to. The information density exceeded my expectations. To learn more about the AI tool which does the work to create the podcast, join my free webinar about NotebookLM by signing-up on this Google Form.
Disclaimer: This newsletter is written with the aid of AI. I use AI as an assistant to generate and optimize the text. However, the amount of AI used varies depending on the topic and the content. I always curate and edit the text myself to ensure quality and accuracy. The opinions and views expressed in this newsletter are my own and do not necessarily reflect those of the sources or the AI models.