- AI Fire
- Posts
- 📸 AI + Selfie Pic > A Doctor
Read time: 5 minutes
I could tell you your next medical checkup could start with a selfie and it might know more about your future than your doctor does? But a new AI tool is turning heads (literally)… just from your face.
What are on FIRE 🔥
IN PARTNERSHIP WITH BRAD'S DEALS
Amazon Prime isn’t just about fast shipping and streaming. There are 10 hidden perks that could completely transform the way you shop, save, and enjoy Prime. If you’re not using these benefits, you’re missing out on the full value of your membership.
Check out this list of exclusive features and start unlocking the full potential of your membership today.
AI INSIGHTS
A simple selfie not only reveals how old you look, but also how long you might live, and even how well you'll respond to cancer treatment.
That’s the surprising finding from a team of researchers at Mass General Brigham, who developed a new AI tool called FaceAge. It uses facial photos to predict a person’s biological age and even forecast survival outcomes in cancer patients.
It scans facial photos — like a selfie — to estimate your biological age, which reflects how fast your body is aging, not just your birthdate.
🔍 How FaceAge Was Built and Tested?
FaceAge was tested on:
6,196 cancer patients using routine photos taken before radiotherapy.
100 patients in palliative care, compared against predictions made by 10 clinicians.
📊 What Did FaceAge Find?
Older FaceAge = lower survival rates across various cancer types
In end-of-life predictions:
Cancer patients looked 5 years older biologically than their real age.
Those with older FaceAges had poorer survival outcomes.
Clinicians' predictions were less accurate than FaceAge in many cases — unless FaceAge data was included.
🧠 Smarter Than Doctors?
In a test with 100 palliative care patients, doctors were asked to guess life expectancy. Without AI help, most doctors’ predictions were barely better than chance.
When given FaceAge insights, their predictions improved significantly.
In short: FaceAge outperformed 10 trained clinicians in short-term survival prediction.
Why It Matters: FaceAge shows just how much hidden health data is encoded in our appearance. While it won’t replace doctors (and shouldn’t), who knows, in tomorrow’s checkup, your selfie might just speak louder than your words.
TODAY IN AI
AI HIGHLIGHTS
🎉 Manus AI is finally open to everyone! No waitlist needed, plus you get 1 free daily task and an one-time FREE 1,000-credit bonus just by signing up.
💼 Will AI and AI agents replace God, steal your job, and change your future? AI Agents emergency debate: These jobs won't exist in 24 months! We must prepare for what's coming.
🚀 Google just launched the AI Futures Fund to supercharge startup innovation with early access to advanced AI models, hands-on support and huge Google Cloud credits. Apply today.
📱 The first smartphone to get Google's AI video generator isn't a pixel. Honor 400 buyers will be the first to try it on a phone. Free for first 2 months for Honor 400 users.
👨⚕️ OpenAI launched HealthBench to evaluate AI performance in healthcare-related tasks. According to HealthBench results: OpenAI's o3 model ranked highest with a 60% score. Grok followed at 54%. Gemini 2.5 Pro scored 52%.
👨💻 One X user literally shared a 3-hour step-by-step video guide on building an online software business with AI for Free. Code a SaaS from zero.
💰 AI Daily Fundraising: Perplexity AI is raising $500 million at a $14 billion valuation, up from $9 billion last November. It’s launching Comet, a new AI-powered web browser, challenging Google and Apple’s search dominance.
AI SOURCES FROM AI FIRE
Welcome to episode #6 of Monday ToolDrop series!!!
We’re talking about autonomous AI agents without human intervention, or that you could have an independent web browser that works entirely on its own.
Check out what’s included this week (FREE analysis below):
A comprehensive framework for Agentic AI development
A truly independent web browser that prioritizes privacy and security
An open-source video messaging alternative to Loom
An AI agent framework designed specifically for Qwen language models
A Pythonic framework for the Model Context Protocol.
Note: These exclusive resources & reviews are available only in our AI Fire community. It’s because you guys can freely ask for support or share personal experience during testing there. Get your full breakdown here (no hidden fee)!
NEW EMPOWERED AI TOOLS
✍️ Resume Maker is the easiest way to build your perfect resume in minutes.
📚 CheatGPT is your all-in-one AI cheat code for exams, studying & more.
🎧 Mapondo is an AI-powered audio guides made for modern travelers.
📱 GoBuildMy.App can build native iOS and Android apps by just chatting.
💻 Matter AI reviews code changes focused on bugs, security & performance issues.
AI QUICK HITS
🤖 Even top VCs say no one really knows what an "AI agent" is. They fund AI ghosts?
🔋 Apple boosts iPhone battery life in iOS 19 using AI, similar to Google’s adaptive battery.
👥 Fake AI tool to spread malware is targeting 62,000+ via Facebook lures.
❌ Coca-Cola new AI ad campaign - “Classic” - still gets basic facts wrong.
🔬 AI can guess racial categories from heart scans with up to 96% accuracy.
AI CHEAT SHEET
AI CHART
Just months after OpenAI’s reasoning model o3 stunned the AI world with huge performance gains, a new report suggests the party may not last much longer. But the blazing-fast improvements in reasoning AI models could start slowing down as soon as next year.
After all, o3 reportedly used 10× the compute of its predecessor (o1) and achieved massive performance. But we may be approaching the limits of what throwing more compute at reasoning can achieve.
Compute scaling has been growing 10× every few months.
This could soon hit practical and financial limits, converging with broader compute growth (only ~4× per year).
Once this ceiling is hit - likely in 2026 - progress will slow unless new breakthroughs emerge.
OpenAI isn’t the only one building reasoning models. Others like DeepSeek-R1, Microsoft’s Phi-4, and Nvidia’s Llama-Nemotron are experimenting too. Some interesting takeaways:
DeepSeek-R1 used way less compute than o3 and still got great results.
Nvidia and Microsoft’s models used even less and still performed well on benchmarks.
But many of these models used synthetic data or supervised fine-tuning (not just RL), so it’s hard to compare apples to apples.
So yeah, compute helps but it's not the whole story. But why should you care if AI reasoning models slow down?
Well, the entire AI industry is betting big on reasoning models being the next leap toward Artificial General Intelligence (AGI). But if the gains from just adding more compute start drying up, companies may have to change strategy fast.
In a way, this slowdown could actually be a good thing. It forces the field to grow up a bit, to get clever rather than just go bigger. That’s worth keeping an eye on 👀
AI JOBS
We read your emails, comments, and poll replies daily
How would you rate today’s newsletter?Your feedback helps us create the best newsletter possible |
Hit reply and say Hello – we'd love to hear from you!
Like what you're reading? Forward it to friends, and they can sign up here.
Cheers,
The AI Fire Team
Reply