Last week might go down as one of the busiest and most dramatic weeks in the AI world this year. Everyone’s still buzzing about GPT-5 and Google’s new Genie 3, but those headlines overshadowed dozens of other fascinating developments. From new creative tools you can try today, to corporate drama, to fresh AI research—there’s a lot worth unpacking.
So, let’s take a calm walk through all the AI stories that flew under the radar, but absolutely deserve your attention.
🖼️ Microsoft Copilot 3D: Image to 3D in Seconds
Let’s start with something you can actually play with today. Microsoft launched Copilot 3D, a feature inside Copilot Labs that converts any 2D image into a 3D model.
Unlike text-to-3D research models, this tool doesn’t require prompts. Instead, you upload a photo or graphic and Copilot 3D spits out a 3D object you can rotate, inspect, and export.
What’s it useful for?
- Game developers → Quickly prototype assets.
- Designers → Add 3D visuals to websites or VR scenes.
- Makers → Prep objects for 3D printing.
- Educators → Turn diagrams into interactive learning objects.
I tested it with two very different cases:
- A silly AI-generated taco roller skate.
- A real-world photo of a guitar.
Both rendered within ~20 seconds—staggeringly fast compared to older image-to-3D pipelines that used to take hours on cloud servers. The results weren’t flawless (the guitar headstock was a bit warped), but for a free, instant tool, it’s impressive.
💡 Pro tip: If you know Blender or Unreal Engine, you can clean up Copilot’s 3D models for production use.
🎬 Perplexity Expands Into Video
We usually think of Perplexity as a research search engine, but now it’s adding video generation.
Subscribers get a limited number of short video generations per month (5 for Pro, 15 for Max). These are ~8-second clips, complete with sound.
Here’s what makes it stand out:
- Works from text prompts (“a cowboy riding a kangaroo on an erupting volcano”).
- Can also do image-to-video, animating a still photo into motion.
- Uses prompt optimization behind the scenes to boost results.
The system isn’t confirmed, but the length and style suggest it’s powered by Google’s Video V3 model.
Is it revolutionary? Not exactly. But if you’re already using Perplexity for research, this is a sweet bonus feature for quick creative experiments.
🎭 OpenArt’s One-Click Story Generator
If you’ve seen “AI slop” or “brain rot” meme-style videos on TikTok, you’ll know what this is about. OpenArt, a startup founded by ex-Googlers, now offers a One-Click Story generator.
Here’s how it works:
- Type a sentence or script (e.g., “Nina goes skydiving”).
- Choose a character or upload your own.
- Let OpenArt generate a short narrative video with characters, music, and janky charm.
The output is intentionally rough, but that’s part of the appeal. The “low-fi” look is exactly what makes these clips viral on platforms like TikTok and Instagram.
👉 While I personally worry about the flood of low-quality AI content, this tool is worth noting as it democratizes animation and storytelling—even if it’s not Hollywood-grade.
🧠 Recall: Your Personal AI Memory
If you consume podcasts, YouTube videos, or articles daily, you probably forget 90% of them by the next week. That’s where Recall comes in.
What Recall does:
- Saves and summarizes web content in one click.
- Auto-tags and organizes everything.
- Lets you chat across your knowledge base.
Example: Instead of digging through old notes, you can ask Recall, “What image-generation tools have I tested already?” and it finds the answer.
It recently added the ability to import past notes, bookmarks, and even TikToks, making your entire knowledge history searchable.
So far, this feels like a game-changer for anyone drowning in information.
📚 Notebook LM Gets Video Overviews
Google’s Notebook LM continues to evolve. Initially known for turning collections of documents into podcasts or mind maps, it now supports video overviews.
Upload your sources (articles, transcripts, etc.), and Notebook LM creates a slideshow-style video with narration.
Imagine:
- Teachers auto-generating study explainers.
- YouTubers creating content pipelines with minimal effort.
- Businesses turning reports into shareable presentations.
The style is closer to Google Vids (slideshow generator) than full-blown generative video, but it’s still powerful—and very accessible.
💸 Google’s New AI Finance & Travel Tools
Google announced two AI-powered services:
- Google Finance AI → Lets you ask natural language finance queries with real-time data.
- Google Flight Deals → AI-assisted affordable flight search (beta in US, Canada, India).
Both are rolling out gradually. While not live for everyone yet, they hint at a future where we query markets and travel plans conversationally.
🔑 AI Memory Wars: Gemini & Claude
Chatbots are getting personal. Memory is the new frontier:
- Gemini → Adds “temporary chats” (incognito) and long-term memory. Rolling out slowly.
- Claude → Now remembers past conversations, but only when you explicitly allow it.
ChatGPT pioneered this, but now competitors are catching up. Memory means AI can give more personalized advice, though it raises privacy concerns.
💥 Silicon Valley Drama
What’s an AI news roundup without a bit of chaos?
- Elon Musk vs. Sam Altman: Elon accused Apple of antitrust favoritism toward OpenAI. Sam clapped back, accusing Elon of manipulating X. Elon challenged Sam to sign an affidavit. Even Grok, Elon’s own AI, hilariously sided with Sam.
- xAI co-founder exits: Igor Babushkin left Elon’s AI company to start his own VC firm. Elon publicly thanked him, but the timing—right after GPT-5 drama—raised eyebrows.
- Microsoft poaches Meta talent: Offering multimillion-dollar packages, despite being close collaborators with Meta.
Sometimes it feels like AI news is just reality TV for nerds.
🤖 New Language Models
Several smaller but important model releases dropped:
- Gemma 3 (270M) → A lightweight Google model designed for mobile devices.
- Mistral Medium 3.1 → Incremental improvements to their mid-range model.
- Qwen updates → Alibaba extended context windows to 1M tokens.
- Claude Sonnet 4 → API now supports 1M tokens.
- OpenAI → After backlash, ChatGPT-4o was restored as an option.
Trend: Huge context windows everywhere, making it easier to dump entire books into a single chat.
🎮 Matrix Game 2.0: Open-Source Genie
Following Google’s Genie, Skywork AI launched Matrix Game 2.0, which creates explorable 3D worlds from images.
Caveat: Running it locally requires a GPU with 24GB+ VRAM, so most users will wait for cloud versions. Still, it’s a peek into the future of interactive AI worlds.
🎨 Creative AI Updates
- MidJourney → HD video generation now available for standard users.
- Pika Labs → Teased lightning-fast lip-sync model (6s for HD video).
- Higsfield AI → Added “draw-to-video,” letting you sketch edits directly on images.
Together, these make creative AI more accessible for hobbyists and professionals alike.
♟️ AI Chess Tournament
Google’s Kaggle hosted an AI vs. AI chess tournament. Surprisingly, GPT-3.5 (ChatGPT’s o3 model) beat out Grok 4 in the finals.
This shows smaller models can still excel in structured domains like games, where efficiency matters more than raw size.
⚠️ Geoffrey Hinton’s Warnings
AI pioneer Geoffrey Hinton issued chilling warnings:
- 10–20% chance AI could wipe out humanity.
- We must train AI with maternal instincts to care for humans.
This ties into the alignment problem—making sure AI goals don’t diverge dangerously from human values.
Sobering stuff from someone often called the “Godfather of AI.”
🧺 Robots Folding Laundry
Let’s end on a lighter note. Figure Robotics unveiled a humanoid robot that can move laundry and fold it.
Last week we saw robots loading washers. This week, they’re folding shirts. It’s still early, but it’s one of the most practical robotics demos in years.
❓ FAQ
Q: Is Copilot 3D free?
Yes, all you need is a Microsoft account.
Q: Does Perplexity really use Google’s Video V3?
Not confirmed, but video length strongly suggests so.
Q: Which chatbot has the best memory right now?
Claude’s implementation feels the most polished so far.
Q: Why does context window size matter?
It determines how much text/models can “remember” in a single conversation. With 1M tokens, entire books or datasets fit at once.
Q: Should we worry about AI wiping us out?
Experts are divided, but Hinton believes it’s a real possibility.
⚠️ Disclaimer
This article is for informational purposes only. Many of the tools mentioned are experimental or rolling out gradually. Features, pricing, and availability may change. Always check official websites for the latest details.
Tags: Copilot 3D, Perplexity AI, OpenArt AI, Recall app, Notebook LM, Google AI, Claude memory, Gemini memory, AI drama, Geoffrey Hinton, Figure Robotics, Matrix Game 2.0, MidJourney, Pika Labs, Higsfield AI, Kaggle chess
Hashtags: #AINews #Copilot3D #PerplexityAI #OpenArt #RecallAI #NotebookLM #ClaudeAI #GeminiAI #MatrixGame #MidJourney #Hinton
✨ And that wraps up our August AI Roundup. From wild drama to practical tools, this month shows how fast—and unpredictable—the AI space really is.
