WednesdAI // Week 42

A lot of news about AI filmmaking advancements and new policies around AI technology so let’s dive right in.

 
Welcome to WednesdAI – Pixel Dreams’ weekly update with top stories from the rapidly evolving world of Artificial Intelligence.

This Week’s News

Top Story

🎬 How to make complicated AI Action shots

Bringing cinematic storytelling into the age of AI, this demo explores how creativity and technology can collide to produce something truly film-worthy. We’re not focusing on any particular model or chasing technical trends — this project is about merging classic filmmaking craft with the latest AI tools to create a Hollywood-style action shot from scratch. In this demo, you’ll see how a retro sci-fi robot came to life through Midjourney, how Nano Banana in Google Gemini was used to generate dynamic poses, and how AI editing tools like Freepik and Photoshop refined every frame. The result isn’t just an experiment in AI — it’s a look at how cinematic vision and machine intelligence can work together to tell a story with motion, drama, and style.

🎥 Watch our full demo on WednesdAI to see how it all came together, from concept to action shot.

⚙️ Agents at Work

🧠 Google Gemini 2.5 Automates Your Browser

Google unveiled Gemini 2.5, its next-generation AI model, which can now control your computer to complete tasks like organizing files or editing documents—no clicks required. The feature, called “Computer Use,” lets Gemini take direct action across apps, guided by text prompts or voice commands. Google says the system runs securely in a sandbox, though details on data handling remain fuzzy. For businesses, this marks Google’s latest step toward full desktop automation—great for efficiency, but also a reminder that AI assistants are inching closer to being digital employees.

📰 Get the updates from Google blog.

🎬 NotebookLM Adds Video Overviews + Nano Banana Visuals

Google is expanding NotebookLM, its AI research assistant, with a new feature called Video Overviews that automatically creates short, narrated explainer videos from documents, notes, or data sources. Users can upload materials, and the tool generates a custom video complete with visuals and voiceover—essentially turning research summaries into mini presentations. The update also introduces Studio, giving users more control over tone, pacing, and visual style. For businesses and creators, this moves Google deeper into AI-generated media production—transforming raw information into polished content at the click of a button.

📰 Dive into more insights from Google Blog, The Verge and Google Blog.

⚖️ Creative Control

🌏 Singapore’s AI Copyright Safe Harbor

Singapore’s proposed AI Copyright Safe Harbor aims to give companies limited legal protection when using copyrighted material to train generative AI models. The policy would allow AI developers to use such data for research and innovation without automatic infringement, as long as it’s not for direct commercial replication. Supporters see it as a way to attract AI investment and clarify ownership rules, while critics argue it risks weakening creator rights. For global businesses, the move could make Singapore a test case for balancing innovation with intellectual property law in the AI era.

📰 Read the article form Science direct.

🎬 Hollywood Studios Build Their Own AI Models

Major Hollywood studios including Netflix, Paramount, and Warner Bros. are reportedly negotiating with OpenAI to license their film and TV libraries for AI training. The deals would give OpenAI access to decades of video content to help improve its models’ visual understanding and storytelling abilities. The move has raised concerns about copyright, creative ownership, and how much control studios are giving up to the same technology that could one day compete with them. For Hollywood, it’s a calculated risk between monetizing old content now or watching AI learn from it for free later.

📰 Read more about this from The Verge.

💡 Reality Check

💼 Enterprise AI Adoption: Boom or Bust?

Enterprises are pouring billions into AI, but most aren’t seeing meaningful returns. Around 95% of corporate AI initiatives fail to deliver measurable value, largely due to poor data infrastructure, unclear strategies, and unrealistic expectations. Many companies rushed into adoption to appear innovative, only to face mounting costs and integration challenges. The result is an AI landscape where spending keeps rising, but productivity gains remain mostly theoretical.

📰 Find out from MIT and TechCrunch.

🧬 AI Helps Man Diagnose Lyme Disease

Oliver Moazzezi from Hampshire turned to AI after years of unexplained symptoms like tinnitus, fatigue, and muscle spasms were dismissed as anxiety by doctors. By inputting his detailed symptoms into an AI tool that referenced verified medical sources, he was guided toward a Lyme disease diagnosis later confirmed by a private doctor. While Oliver credits AI for leading him to answers, experts warn that self-diagnosis with AI can be risky and should never replace professional medical advice. Specialists also argue the UK’s Lyme disease surveillance and testing standards need reform to better capture real cases.

📰 Full story from BBC News.

The section header images in this article were generated using the following prompts:

Sam Altman sits at a futuristic, glowing keyboard, typing the foundational code for a video prompt. Floating in the air above the keyboard, Bob Ross uses a digital paintbrush to paint colorful, descriptive words like almighty mountains and lazy little clouds. Beside them, Michael Jackson strikes an iconic, energetic pose, and the AI captures his motion, translating it into dynamic words like sudden explosion of light and rhythmic pulse that add themselves to the prompt text. Conceptual, symbolic, high-tech, teamwork.

Taylor Swift’s face illuminated by a smartphone, excitedly follows an online scavenger hunt. On the phone screen, the image of the pop star glitches for a split second, revealing an underlying wireframe and a cold, robotic expression. The fan’s excitement morphs into confusion and slight fear. Moody, atmospheric, shallow depth of field, hyperrealistic, close-up on the fan’s face and the phone.

A conceptual art piece showing a digital pipeline. On the left, a pristine ChatGPT interface inputs clean data. The data flows through glowing tubes labeled “Canva” for visualization and “Zapier” for automation. The pipeline is clean and efficient. But on the right, the output of the pipe is not a report, but a messy, chaotic splash of corrupted pixels and shredded digital money, staining a document with a corporate letterhead. Symbolic, minimalist, high detail.

A hyper-detailed digital painting of Disney’s character. A spiraling vortex of text surrounds, filled with confidently incorrect statements about the laws of thermodynamics. At the end of the line, a giant robotic arm with a Disney logo is methodically picking up each flawed bot and dropping it into a digital shredder. Dystopian, conceptual art, clean aesthetic with a dark twist.