Welcome to The Daily AI Show Newsletter, your deeper dive into AI that goes beyond the latest news and our live weekday show. In this issue:

• What's on Your AI Shipmas List? Dreaming Big for AI in 2025

• Smarter AI, Less Data: The Revolution of Model-Based Transfer Learning

• Canvas Meets Custom GPTs: Your Workflow, Reimagined

Plus, we discuss this week’s 12 Days of OpenAI announcements, Google’s claim on the multiverse, what happens when effort is no longer a signal of value, and all the news stories that caught our eyes and ears this past week.

It’s Sunday morning

Time to check in with AI and see how close we are to it running our lives.

(spoiler: pretty close)

The DAS Crew

What's on Your AI Shipmas List? Dreaming Big for AI in 2025

As the AI world closes out 2024, the wish lists for future developments are brimming with creativity, ambition, and practicality. From smarter multimodal applications to co-writing tools and better integrations, the "Shipmas List" is a collective reflection of what users hope AI will deliver next.

A recurring theme is accessibility and efficiency—tools that simplify everyday interactions with AI and remove the technical barriers to productivity. Whether it’s a true co-writer that can edit documents contextually, multimodal AI that can seamlessly combine video, audio, and text, or new ways to interact with AI through more intuitive apps, the Shipmas wish list reveals a shared vision for AI that feels more personal, useful, and embedded into daily life.

WHY IT MATTERS

Smarter Multimodal Interactions: Tools that integrate video, audio, and text capabilities in real-time could revolutionize how we work, learn, and create content.

Enhanced Productivity Tools: From co-editing documents to enabling seamless app connections, these innovations would drastically reduce friction in workflows.

Personalized AI Support: A wish for better AI personalization, like custom tools that adapt dynamically to user preferences, shows how deeply users want AI to integrate into their unique needs.

Expanding Accessibility: From hands-free interactions on mobile devices to co-writers that intuitively understand context, the focus is on making AI easier to use for everyone, regardless of technical expertise.

Setting the Stage for 2025: These requests reflect where AI could head next—more agent-like behaviors, intuitive interfaces, and tools that empower creativity and innovation across industries.

Smarter AI, Less Data: The Revolution of Model-Based Transfer Learning

Model-Based Transfer Learning (MBTL) is reshaping how we think about AI training, introducing a more efficient way to teach models by focusing on only the most impactful data. Rather than overwhelming models with vast amounts of mixed-quality information—often redundant or confusing—MBTL first identifies the training data that provide the highest value improvements in performance. By selecting specific higher value data and minimizing unnecessary training, MBTL not only reduces costs but also accelerates the enhancement of effective AI systems.

While MBTL's initial focus is on complex problems like optimizing traffic flow, its potential applications extend far beyond. The ability to pinpoint impactful data could be a game-changer in logistics, agriculture, and even sports analytics, making AI systems more adaptable and scalable.

This method also hints at a future where smaller businesses can afford sophisticated AI, relying on leaner, smarter models that don’t require enormous datasets to perform effectively.

WHY IT MATTERS

More Efficient AI Training: MBTL drastically reduces the time and computational power needed to train AI, making advanced-capacity models more accessible to smaller organizations and projects.

Scalable Solutions for Complex Problems: From traffic flow optimization to logistics and agriculture, MBTL allows AI to tackle multi-variable systems in real-world settings with greater ease.

Cost Reduction for Businesses: By training on only the most relevant data, MBTL cuts costs associated with traditional training methods, broadening access to powerful AI tools.

Applications in Everyday Life: Beyond industries, MBTL could improve mundane tasks like daily commutes, reducing traffic jams and delays with smarter, real-time AI interventions.

Foundations for Specialized AI: MBTL’s targeted approach could pave the way for domain-specific AI systems that achieve “specialized AGI” in areas like supply chain management or urban planning.

Canvas Meets Custom GPTs: Your Workflow, Reimagined

The integration of OpenAI’s Canvas with Custom GPTs is a game-changer, offering users a way to bring AI-driven workflows into a more interactive, document-based space. While it may seem like mearly an upgraded chat thread at first glance, the real power of Canvas lies in how it enhances specific, repeatable tasks. By combining the adaptability of Custom GPTs with Canvas’s collaborative, real-time editing interface, users can unlock productivity like never before.

Below are three practical use cases that Brian used to highlight the potential of this new integration.

You can try out and use these Custom GPTs demos at the links below, and view the full custom GPT instructions for all 3 GPTs here.

Sales Email Drafting: Need to write personalized emails to prospects? A Custom GPT designed to draft sales emails uses Canvas to pull prospect data and offer tailored drafts. Canvas makes it easy to refine subject lines, adjust messaging, and create A/B test variations—all in a collaborative editing environment that streamlines the process.

Smart To-Do Lists: Forget static to-do lists. With Canvas, a Custom GPT can create a dynamic, gamified to-do list that helps you prioritize tasks, track progress, and even re-prioritize when urgent changes arise. Canvas adds visual clarity, enabling you to see updates and edit directly, turning your AI into a productivity partner.

Python Programming Made Simple: For those learning or experimenting with Python, Canvas provides an intuitive space to create, edit, and run scripts directly. A Custom GPT acts as your coding assistant, guiding you through writing and debugging code while Canvas handles execution and live adjustments. This makes even complex tasks feel manageable.

Just Jokes

12 Days of OpenAI

This week we gave our live reactions to days 3-7. We will continue to do live reactions for the rest of the 12 days. Days 8-12 will start tomorrow and go to Friday.

Sora: Text-to-Video Generation

OpenAI introduced Sora, a text-to-video model that transforms textual prompts into dynamic video content. Building upon the success of DALL-E, Sora enables users to generate short video clips by simply describing the desired scene. This tool aims to democratize video creation, offering storytellers and content creators a novel medium for expression. However, early reviews indicate that while Sora shows promise, it occasionally struggles with complex actions and detailed scenes, suggesting it's still in developmental stages.

Canvas: Collaborative AI Interaction

The Canvas feature enhances custom GPT's functionality by providing a collaborative workspace where users can co-write text and code alongside the AI. This interface allows for real-time editing and interaction, streamlining the creative process and making AI assistance more intuitive and accessible. By facilitating a more hands-on approach, Canvas aims to boost productivity and creativity in various projects

Apple Intelligence Integration

In collaboration with Apple, OpenAI has integrated ChatGPT into Apple's operating systems, allowing users to access AI capabilities directly through Siri and other native applications. This partnership enhances the functionality of Apple devices, enabling more complex queries and tasks to be handled seamlessly. The integration emphasizes user privacy, with data processed securely to protect personal information.

Advanced Voice Mode with Vision

OpenAI expanded ChatGPT's capabilities by introducing an advanced voice mode that incorporates visual elements. This feature allows users to engage in voice-based interactions complemented by visual aids, enriching the conversational experience and making information more accessible and engaging. The addition of visual context aims to enhance understanding and retention of information during AI interactions.

Project Folders: Organized AI Engagement

To improve user experience, OpenAI launched "Projects," a feature that enables users to organize their AI interactions into distinct folders. This system allows for better management of various tasks and activities, with options to edit project titles, set icon colors, and add relevant files and instructions. By consolidating related chats and materials, Projects facilitate a more structured and efficient workflow within ChatGPT.

Did you know?

AI is helping physicists explore multiverse theories by analyzing vast datasets from particle collisions. At CERN's Large Hadron Collider, AI algorithms sift through massive amounts of data to detect anomalies that could indicate new particles or phenomena, potentially offering insights into the multiverse concept.

Additionally, AI techniques like Generative Adversarial Networks (GANs) simulate various multiverse scenarios, allowing researchers to model and study different possible universes.

These AI applications enable scientists to explore complex theories about the multiverse more efficiently, providing new avenues for understanding the fundamental nature of reality.

This Week’s Conundrum
A difficult problem or question that doesn't have a clear or easy solution.

The Effortless Signal Paradox:

For centuries, effort has been a proxy for value. A well-written novel, a detailed business plan, or even a well-studied exam result signals expertise, diligence, and mastery. AI disrupts this framework by making high-quality outputs cheap and ubiquitous. But if everyone can produce polished essays, detailed reports, or stunning art at a fraction of the time, the ability to produce these outputs is no longer scarce, and effort is no longer a meaningful signal.

The conundrum: When effort becomes decoupled from value, what happens to systems—like education, hiring, or awards—that depend on effort as a measure of worth? Do we embrace AI-driven abundance, accepting a new system where the focus shifts to creativity, purpose, or originality, even if it upends centuries of tradition? Or do we hold onto existing systems and attempt to evolve new markers of human effort, at the risk of excluding or devaluing AI-generated contributions?

Canadian Government Invests $240 Million in AI Data Center
The Canadian government announced a $240 million investment in Cohere’s multi-billion-dollar AI data center. This funding aims to boost Canada's competitiveness in AI infrastructure and innovation.

DeepHealth AI Improves Breast Cancer Detection
DeepHealth, owned by RadNet, revealed a study where their AI technology improved breast cancer detection rates by 21%. The study involved nearly 750,000 mammograms and showcased the potential of AI in enhancing diagnostic accuracy.

YouTube Rolls Out Auto-Dubbing for Knowledge-Based Content
YouTube’s new auto-dubbing feature converts creator audio into multiple languages, targeting knowledge-focused content like tutorials. This feature aims to expand creators’ global reach by breaking language barriers.

Neural Attention Memory (NAM) Enhances Transformer Efficiency
A new study introduced Neural Attention Memory (NAM), an optimized system for transformers to store and retrieve information more efficiently. This technology improves AI model performance across domains, including vision and reinforcement learning.

Chiba University Explores Robots for Elderly Care
Researchers at Chiba University are developing home care robots to support the growing elderly population. With one in six people projected to be over 60 by 2030, these robots aim to fill gaps in caregiving services.

Google Quantum Chip “Willow” Sets New Speed Record
Google unveiled its latest quantum chip, Willow, which performs calculations in under five minutes that would take supercomputers septillions of years. The chip also addresses error correction challenges, marking a significant leap in quantum computing.

WordPress Owner Acquires AI Tool Builder WP AI
Automatic, WordPress’s parent company, acquired WP AI, a developer of AI tools for WordPress sites. This acquisition promises seamless integration of AI-powered features into the WordPress ecosystem.

Humane Launches Cosmos OS for Unified AI Management
Humane introduced Cosmos OS, an operating system designed to integrate and manage all user-facing AI systems across devices. This platform acts as a central interface, orchestrating various AI assistants.

Did You Miss A Show Last Week?

Enjoy the replays here on YouTube or take us with you in podcast form on Apple Podcasts or Spotify.