🧠 Brain Scan Study Says Heavy ChatGPT Use May Sap Mental Muscle

MIT researchers wired up volunteers with fMRI scanners while they wrote essays under three conditions: using ChatGPT, using a standard web search, or relying on no digital tools. The study—summarized in a recent report—tracked neural connectivity and short-term recall immediately afterward. Participants using the LLM produced smooth, coherent text, often praised by human graders for relevance and flow. But the scanners told a different story.

📉 What the Scans Revealed

Brain images showed that people who leaned on ChatGPT displayed weaker links between regions involved in critical thinking and memory. Although their essays scored well, researchers called the writing “formulaic” and noted that those same participants struggled to quote or summarize their own work in a follow-up session. By contrast, the “search engine” group showed middling neural activity, while the pen-and-paper cohort retained the strongest connectivity and recall.

🪤 The Risk of Cognitive Off-Loading

The findings suggest that outsourcing research and wording to a language model can dull the mental circuits you’d normally exercise while brainstorming, fact-checking, and drafting. One subject even admitted they “felt like a beginner again” when asked to write without AI support. Researchers warn that routine dependence on generative tools could encourage a vicious cycle: the more you rely on the model, the less practice your brain gets at creating and remembering ideas—so you rely on the model even more.

🔄 Hope for a Rebound

The study isn’t all doom. In a second phase, groups swapped methods. Participants who ditched ChatGPT and wrote unaided showed renewed neural connectivity, suggesting that cognitive function can rebound once users re-engage their own problem-solving skills. That plasticity gives educators and employers leverage: mix AI with offline tasks to keep minds limber.

⚙️ What Happens Next?

Researchers propose simple safeguards—rotating AI and non-AI activities, inserting reflection breaks, and teaching users to interrogate model outputs rather than copy-paste. They also call on platform builders to add optional timers and “mental-effort meters” that flag extended passive use. As workplaces embed LLMs into everyday workflows, the study is a timely reminder: AI can speed the road, but you still need to keep steering—or watch your cognitive engine idle.

🎥 MidJourney Enters the AI-Video Arena

MidJourney’s new “v1” video model lets users generate 10-second clips in the same Discord interface that made its image tool popular. Early demos show a neon cityscape morphing into a rainforest and a hyper-real coffee swirl that zooms into microscopic beans—all rendered in about two minutes.

🆚 Facing Veo and Sora

Unlike Google’s Veo or OpenAI’s Sora, MidJourney sticks to a community-driven workflow: type /imagine video plus a prompt, then upscale or remix frames. Resolution tops out at 1080p for now—lower than Veo’s 4K, but MidJourney’s strength is artistry and rapid iteration. Creators praise its cinematic color grading and painterly textures, noting it “feels like an animated concept painting come to life.”

💼 Monetization and Moderation

The company will charge one “GPU minute” per render on paid plans; free-tier users can watch previews but must upgrade to export MP4s. Moderation rules mirror image generation: no explicit violence, political persuasion, or likeness of minors. A “video remix” toggle lets users evolve clips while preserving the original seed, fostering collaborative storytelling threads.

🌟 Why It Matters

With MidJourney, Runway, Pika, Veo, Sora, and emerging Chinese models all vying for attention, AI video is moving from tech demo to creator staple. Analysts predict short-form ads, music videos, and game trailers will flood social feeds by year-end, forcing agencies to master prompt-craft instead of pricey shoots.

This Week’s Scoop 🍦

⚖️ AI-powered law firm launches with Sequoia backing

🚕 Baidu plots Robotaxi rollout to Singapore and Malaysia

💰 Meta offers $100 million payday to AI engineers

🏢 Microsoft axes sales staff, pivots to AI-first strategy

👔 UK’s AI “tsar” resigns after six months in role

✝️ Pope Leo makes AI’s existential risk a signature issue

🎥 Weekly Challenge: Launch Your First MidJourney Video

Ready to unlock AI art and video? This week’s challenge helps you sign up and kick off your first creation in under 10 minutes.

Here’s how:

🔑 Easy sign-up: Head to midjourney.com and click Join the Beta. You can register with Discord or simply choose “Continue with Google” if you have an account.

📧 Verify in seconds: Whether you use email or Google, check your inbox (or approve the Google prompt) to confirm your account and access the MidJourney server.

🚪 Enter the server: Once verified, Discord opens to the MidJourney community. Find any #newbies channel on the left sidebar—this is your launchpad.

💡 First image tip: In #newbies, type /imagine followed by a simple scene (e.g., “golden hour forest path”). Keep prompts short—two to three key phrases—to see clear, striking results.

🎞️ Quick video test: Try /imagine video plus a brief action (“butterfly flapping wings over a lake”). A 6-second clip renders in about two minutes—perfect for experimenting without burning through credits.

📂 Find your work: Everything you create appears in midjourney.com/app under Gallery—no more hunting through chat logs.

Share your first masterpiece on Discord or social media, and showcase your debut creation next week!

Want to sponsor Jumble?

Click below ⬇️

That’s it for this week! ChatGPT use causing cognitive decline, video dreams, and Robotaxi schemes—AI’s momentum isn’t slowing. What story surprised you most? Hit reply and let us know. Until next!

Stay informed, stay curious, and stay ahead with Jumble!

Zoe from Jumble