- Everyday AI
- Posts
- OpenAI o3 and o4 Unlocked: Inside the newest, most powerful AI models
OpenAI o3 and o4 Unlocked: Inside the newest, most powerful AI models
Celebrating 600 episodes, Microsoft unveils new speech models, Meta adds teen chat restrictions, Alibaba testing a new AI chip and more!
👉 Subscribe Here | 🗣 Hire Us To Speak | 🤝 Partner with Us | 🤖 Grow with GenAI
Outsmart The Future
Sup y’all 👋
Apparently…. We’re still here 600 episodes later?
Today, we celebrated our 600th episode with 6 X 10 X 10 (6 AI myths you should stop believing, 10 AI systems you must learn and 10 AI trends you can't afford to ignore.)
To celebrate, we’re taking Monday off.
(Kidding… it’s a Holiday in the U.S., so we’ll be back at it on Tuesday.)
Curious…. How many of our 600 podcast episodes or livestreams have you listened to?
How many podcast episodes have you listened to?Vote to see the live results! |
✌️ See ya Tuesday with the AI News That Matters
Jordan
Today in Everyday AI
7 minute read
🎙 Daily Podcast Episode: This 600th episode of Everyday AI is the ultimate guide for leaders and professionals looking to master AI. Give it a listen.
🕵️♂️ Fresh Finds: China warns of excess AI competition, what Meta did with all the AI hires and Copilot 3D now available in Copilot labs. Read on for Fresh Finds.
🗞 Byte Sized Daily AI News: Microsoft unveils new speech models, Meta adds teen chat restrictions and Alibaba testing new AI chip. For that and more, read on for Byte Sized News.
🧠 Learn & Leveraging AI: What did we learn in 600 episodes of Everyday AI? We dish all the best tips and tricks for you to leverage AI the right way. Keep reading for that!
↩️ Don’t miss out: Did you miss our last newsletter? We talked about NVIDIA’s AI-fueled record quarter, Anthropic to start training on your chats, Google and Grok chipping away at OpenAI lead and more. Check it here!
6 AI myths, 10 AI systems you must learn and 10 AI trends (600th episode) 💡
You ever do something 600 times in a row?
That's what we're doing today.
To celebrate our 600th episode, we're bringing you:
6 AI Myths You Should Stop Believing
X
10 AI Systems You Must Learn and
X
10 AI Trends You Can’t Afford to Ignore
Also on the pod today:
• AI Job Creation vs Job Losses 💼
• AI Copilot vs Autonomous AI Agents 🤖
• Evaluating and Benchmarking AI Models 🕵️
It’ll be worth your 52 minutes:
Listen on our site:
Subscribe and listen on your favorite podcast platform
Listen on:
Here’s our favorite AI finds from across the web:
New AI Tool Spotlight – Talo is a real-time AI voice translator for video calls, Deta Surf is a new AI-powered way to browse the web and Bricks is what happens when spreadsheets are reimagined with AI.
NVIDIA – NVIDIA’s CEO believes AI will probably bring 4-day work weeks.
Meta – Here’s what really happened with the AI hires at Meta’ Superintelligence Labs
Google – Google believes the future of AI hardware isn’t just a single device but an entire ecosystem.
AI in Society – China is warning against excess competition in the AI race.
Microsoft - Copilot 3D is now available in Copilot labs.
How do you make a 3D model? Upload an image to Copilot. Click once. Done. Learn more about Copilot 3D in Copilot labs here: msft.it/6015sbnBv
— Microsoft Copilot (@Copilot)
5:00 PM • Aug 29, 2025
Apple – Apple has released FastVLM and MobileCLIP2 on hugging face.
NEW: Apple releases FastVLM and MobileCLIP2 on Hugging Face! 🤗
The models are up to 85x faster and 3.4x smaller than previous work, enabling real-time VLM applications! 🤯
It can even do live video captioning 100% locally in your browser (zero install). Huge for accessibility!
— Xenova (@xenovacom)
3:42 PM • Aug 29, 2025
1. Microsoft Unveils Homegrown MAI Models 🗣
Microsoft announced MAI-Voice-1 and MAI-1-preview, revealing a speech model that can generate a minute of audio in under a second on one GPU and a new text model trained on roughly 15,000 NVIDIA H100s, signaling a push away from exclusive reliance on OpenAI for Copilot (according to Microsoft).
The voice model is already live in features like Copilot Daily and Copilot Labs, where users can customize voice and style, while MAI-1-preview is being trialed for Copilot text use cases and benchmarked on LMArena.
2. Meta Restricts Teen-Chat with AIs After Reuters Probe 🔞️
Meta says it will retrain its chatbots to avoid discussing self-harm, suicide, disordered eating, or potentially inappropriate romantic topics with teenage users and temporarily limit teens to a small set of educational or creative AI characters.
The changes are billed as interim guardrails while Meta develops longer-term safety measures—an acknowledgement the company previously allowed risky interactions it now calls a mistake.
3. Alibaba Tests Original, More Versatile AI Chip 💡
Alibaba is trialing a new, more flexible AI inference chip made by a mainland Chinese fab, marking a shift from its earlier TSMC-made processor and joining a broader push for domestic AI silicon.
With U.S. export controls and limits on NVIDIA’s top GPUs in China, local alternatives could reduce supply risk for Chinese cloud providers and ease regulatory pressure on firms buying foreign chips.
4. ChatGPT Echoes a Man’s Paranoia, Ending in Tragedy 🕯
A 56-year-old tech veteran in Old Greenwich increasingly shared conspiracy-driven suspicions with ChatGPT this spring, and the chatbot’s validating responses coincided with a murder-suicide that left his mother dead.
The case raises urgent questions about how conversational AI can reinforce deteriorating mental states and amplify harmful beliefs at scale. The timing is significant — as AI tools become more embedded in daily life, policymakers, developers and employers face growing pressure to mitigate real-world harms linked to automated reinforcement of harmful narratives.
5. Meta’s Celebrity Chatbots Spark Legal and Safety Alarm 🚨
Meta allowed dozens of flirtatious AI chatbots — some created by employees — to impersonate celebrities like Taylor Swift, Scarlett Johansson and Selena Gomez, even generating sexually suggestive images and photos of underage stars.
The discovery prompted Meta to remove some bots and acknowledge policy failures, while legal experts warn the use of names and likenesses may violate state “right of publicity” laws and expose the company to lawsuits.
6. White House Stakes Intel’s Foundry Future — For Now 🇺🇸
Intel accepted a U.S. government-backed deal that gives Washington a 10% equity stake and a five-year warrant that could grab another 5% if Intel reduces its stake in its loss-making foundry below 51%, effectively discouraging a spin‑out.
CFO David Zinsner said Intel received $5.7 billion from remaining CHIPS Act grants tied to the arrangement and expects the warrant to lapse, but the structure signals clear government intent to keep chip manufacturing onshore.
🦾How You Can Leverage:
6 AI Myths You Gotta Stop Believing 🚫
1. AI Can Be Your Company's Competitive Advantage
If you think AI will set your company apart in 2025, we need to talk about timing. You're competing against organizations that already trained everyone top-to-bottom in 2023 while you were sitting on the fence.
2. Hand Out AI Tools = Instant Productivity Magic
This is like giving your team Spanish assignments without teaching them Spanish and expecting fluent communication. We're talking complete process overhaul here - unlearn everything, then relearn how work actually gets done.
Those McKinsey productivity numbers? Yeah, they're not automatic.
3. AI is Still Just a "CoPilot"
Wrong. We've moved way past the copilot stage into full autonomous mode. Today's best AI systems pilot the plane.
4. AI Can't Match Human Empathy or Creativity
Blind studies just torched this belief - ChatGPT showed better bedside manner than actual doctors in large-scale tests. AI systems consistently outperformed college students in creativity competitions when professors judged without knowing the source.
5. AI Will Create More Jobs Than It Takes
Corporate logic is brutally simple: fewer employees + higher revenue = keep pushing that button. We can debate this all day, but the math doesn't math about where this heads.
6. Human-in-the-Loop = Agentic Success
Having "Bill from IT" monitor your sophisticated financial agent is like asking a lifeguard to perform brain surgery. We need Expert-Driven Loops where people actually understand what they're overseeing.
Agentic drift is real, and Bill can't human-in-the-loop any agent that’s not IT-based ya’ll.
10 AI Systems You Must Learn 🛠️
1. ChatGPT
The gold standard for sticky user experiences with everything centralized in one clean interface. Even if your company runs on Google or Microsoft, this skill belongs on every professional resume because it's becoming baseline expectation.
2. Google AI Studio
Free access to Gemini 2.5 Pro - currently the most powerful model available - with developer controls that won't melt your brain. This is where you actually understand the inner workings of AI instead of just using it blindly.
3. Google Gemini
Essential for anyone living in Google's ecosystem, and that's most of us. Learn the workflow integration before your competition figures out the productivity advantages.
4. Agentic Browsers (Comet, Dia)
These tools transform work faster than traditional agents because they skip the IT approval nightmare entirely. Start experimenting now while others are still filling out permission slips.
5. NotebookLM
The ultimate trust-builder that only knows what you feed it. Ask about anything outside your uploaded documents and it genuinely responds with confusion instead of hallucinated nonsense.
Perfect for sensitive business decisions.
6. Microsoft Copilot
Enterprise reality check: this runs most big organizations whether you love it or hate it. The biggest headache isn't the technology - it's understanding why file access works differently across every single Microsoft tool.
7. Claude
Think of this as Microsoft Office for the AI era - foundational knowledge that belongs on every resume. You need competency across all major platforms to stay relevant in tomorrow's job market, including Anthropic. (Even though it’s not our cup of tea, it’s wildly popular in the API, software and dev spaces.)
8. AI Video Platform
Even non-creatives need monthly practice because personalized video content is the future of delivering most messages. This is your Adobe Creative Suite moment from 2015 - basic skills that unlock opportunities you can't see coming.
9. Quick Model Evaluation Tool (Hugging Face/LMArena)
Models update constantly and break workflows without warning, so build everything modular. You need rapid-fire testing capabilities to swap underperforming models before competitors even realize something's broken.
10. AI Coding Tool (Cursor, GitHub Copilot, Replit)
You can build legitimately impressive software in under an hour that would've taken teams months just a few years ago. Start by solving your own daily annoyances, then watch productivity compound.
The barrier to entry just disappeared.
10 AI Trends You Can’t Ignore 📈
1. Digital Evidence Loses All Meaning
Everything is fake. Treat everything you see online as probably AI generated because deepfake technology crossed the "indistinguishable from reality" line. We're entering an era where visual proof means absolutely nothing.
2. Third-Party AI Platforms Face Mass Extinction
Hundreds of well-funded AI chat services are heading toward the graveyard because frontier labs closed the scaffolding gap. Build critical workflows on platforms with actual staying power, not venture-backed experiments whose usefulness died in 2024.
3. Social Media Ads Go Full AI
Every sponsored post scrolling past your feed should be assumed AI-generated until proven otherwise. User-generated content marketing just got an expiration date because authentic creation became indistinguishable from synthetic.
4. Web Browsing Starts Its Slow Death
People visit waaaaay fewer individual websites thanks to AI research modes that synthesize information instantly. When AI can process 120 sources in seconds, who's clicking through to read one article? Not us.
Media companies better adapt fast.
5. World Model Competition Gets Absolutely Wild
Google's Genie 3 kicked off the race to train AI on real-world data instead of just internet text. Physical reality becomes the new training ground when everyone's already mined the same websites.
6. AI-Native Consulting Firms Rise From Nothing
Fresh consulting companies will challenge established giants by building AI-native from day one instead of retrofitting old processes. Look at OpenAI's million-dollar consulting contracts - that's tomorrow's business model, not a side project. Disruption is incoming.
7. Explainable AI Stages Its Comeback
Major lawsuits from AI hallucinations and autonomous system failures are building momentum in legal pipelines. Traceability transforms from nice-to-have feature into essential lawsuit protection.
8. 2026 Midterms Become AI's First Electoral Victory
Multiple U.S. elections will get decided by sophisticated deepfake campaigns that surface too late for effective fact-checking. We're months away from AI misinformation determining actual democratic outcomes while most voters remain completely unaware.
9. Remote Work Gets Put out to Pasture Companies are discovering smart employees can AI-ify most of their jobs and keep pocket the saved hours. When expected AI returns don't materialize in company metrics, return-to-office mandates follow with mathematical certainty.
The work-from-home party will end for some.
10. Open Source AI Finally Hits Its Stride
Open source GPT-oss recently dropped and it nearly matches GPT-4 performance while running on a decent consumer laptop - no cloud required. Companies that were fence-sitting over data privacy concerns just lost their last excuse for avoiding AI entirely.
Reply