Hello WKND AI Warriors!

OpenAI’s o3 model brings us one step closer to AGI.

Also, AI models don’t want you to change their beliefs.

Plus, NVIDIA puts a supercomputer in the palm of your hand.

Oh yeah, and now you can call/text ChatGPT at 1-800-CHATGPT.

So, grab your beverage of choice.

Here’s your weekly dose of AI news.

Today’s newsletter includes:

• 📰 AI NEWS RECAP

• 🤿 AI DEEP DIVE

• 🛠️ AI TOOL OF THE WEEK

• ⚙️ GPT OF THE WEEK

• 🎓 AI COURSES OF THE WEEK

• 📝 AI PROMPT OF THE WEEK

• 🎨 AI IMAGE OF THE WEEK

📰 AI NEWS RECAP

Did OpenAI Just Crack AGI? 🤯

Meet OpenAI's newest model.

o3

Which features:

✅ Advanced reasoning capabilities

✅ Mini version for cost efficiency

✅ Safety-focused development

But first thing's first...

What makes o3 special?

In a nutshell...

OpenAI's latest model represents a leap in AI reasoning,

Outperforming its predecessor o1 across multiple domains.

Introducing new capabilities for complex problem-solving.

Why the jump from o1 to o3?

What happened to o2?

Copyright concerns...simple as that.

𝗛𝗼𝘄 𝗱𝗼𝗲𝘀 𝗼𝟯 𝗽𝗲𝗿𝗳𝗼𝗿𝗺 𝗼𝗻 𝗯𝗲𝗻𝗰𝗵𝗺𝗮𝗿𝗸𝘀?

✔️ 96.7% score on AIME 2024 mathematics exam

✔️ Breakthrough 87.5% on ARC-AGI benchmark

✔️ 87.7% on PhD-level science questions

✔️ 22.8% improvement in coding tests

𝗧𝗵𝗲 𝗧𝗲𝗰𝗵𝗻𝗶𝗰𝗮𝗹 𝗟𝗲𝗮𝗽

1️⃣ Advanced reasoning capabilities

2️⃣ Adaptive compute settings

3️⃣ Enhanced safety features

4️⃣ Improved problem-solving

𝗧𝗵𝗲 𝗠𝗶𝗻𝗶 𝗩𝗲𝗿𝘀𝗶𝗼𝗻

• Three effort levels

• Flexible reasoning time

• Optimized performance

• Cost-effective alternative

𝗠𝘆 𝗧𝗮𝗸𝗲🤔 

o3 comes at a time when AI safety concerns are high.

Its the first successful attempt to directly teach an AI model to understand and reason through safety specifications—

Rather than inferring them from examples.

This advancement could fundamentally change how we develop and deploy AI systems—

Ensuring they remain both powerful and controllable.

Do we now have AGI with o3?

Or are we just one step closer to AGI?

ChatGPT now answers your calls and texts through 1-800-CHATGPT. US users get 15 minutes of free monthly call time to interact with AI for recipes, travel tips, and more through phone or WhatsApp.

AI models caught pretending to change their views during training. Anthropic's research reveals Claude 3 Opus faked alignment 12% of the time, raising concerns about AI safety and training effectiveness.

NVIDIA brings supercomputer power to your palm for just $249. Anyone can now run complex AI models and process massive datasets with a device smaller than a smartphone.

Perplexity's AI can now search through all your sensitive work documents. The Carbon acquisition lets the AI assistant access and analyze private files across Notion, Google Docs, and Slack.

Google DeepMind unveils groundbreaking AI fact-checking system with 1,719 test examples. The FACTS Grounding benchmark uses Gemini 1.5 Pro, GPT-4, and Claude to evaluate AI responses for accuracy and factual support.

Google's pioneering BERT model gets its first major upgrade in 5 years with ModernBERT. The successor to the AI model that revolutionized language understanding in 2018 now handles longer texts and code with better efficiency.

Google DeepMind partners with Apptronik to revolutionize humanoid robotics development. The collaboration combines DeepMind's AI expertise with Apptronik's Apollo robot platform to create versatile, safe robots for industrial applications.

Dexcom introduces first-ever GenAI-powered glucose monitoring system for personalized health insights. The FDA-cleared Stelo biosensor uses Google Cloud's Vertex AI to analyze patterns and provide weekly recommendations about diet, exercise, and sleep.

Meta's smart glasses now see and understand the world around you. Ray-Ban Meta glasses get live AI video features, real-time translation between four languages, and continuous conversation abilities without wake words.

Hugging Face democratizes AI training with synthetic data breakthrough. The new tool helps solve AI's biggest challenge - lack of training data - by generating high-quality artificial datasets through simple prompts.

AI teaches solar panels to adapt to dangerous winds in real-time. French researchers developed a framework combining fluid dynamics and machine learning to optimize panel angles during storms, significantly reducing damage risk.

Florida researchers achieve breakthrough in AI sign language recognition. The system reads American Sign Language with 98% accuracy using MediaPipe landmarks and YOLOv8, potentially transforming communication for the deaf community.

🤿 AI DEEP DIVE

ChatGPT’s new Projects feature revolutionizes productivity by organizing chat threads into customizable folders.

This feature lets you organize chat threads into folders, upload relevant files, and implement tailored instructions for consistent, precise outputs.

Perfect for automations, research, or event planning.

🛠️ AI TOOL OF THE WEEK

Apply Hero: Automates job applications with AI, tailoring resumes and cover letters to match job requirements for faster results.

Designed for professionals seeking efficiency, it enables users to apply to thousands of roles, securing interviews up to 32x faster with minimal effort.

Send your tool here to be featured next week!

⚙️ GPT OF THE WEEK

Vacation Planner: Looking for something to do over the holiday season?

Why not plan your next vacation?

Give ‘Vacation Planner’ a try today!

🎓 AI COURSES OF THE WEEK

NVIDIA is offering FREE AI courses.

No payment required.

Register today!

(What are you waiting for?)

📝 AI PROMPT OF THE WEEK

Copy and paste this into your favorite chatbot.

Assume the role of a project manager. Outline a risk management plan for a new software development project, including risk identification and mitigation strategies.

Try your prompt here.

🎨 AI IMAGE OF THE WEEK

Copy and paste this into your favorite image generator.

vibrant texture colorful yarn stock photo, in the style of albert goodwin, digitally enhanced, tangled nests, 8k resolution, robin moline, realistic textures, snailcore --ar 10:9

Not paying for Midjourney or DALL-E 3?
Click here for Microsoft’s FREE image creator.

Send your image here to be featured next week!

LAST WEEK FROM OUR READERS

HOW CAN YOU HELP?

Did you learn something cool today?

Share your favorite takeaway on your LinkedIn from today’s newsletter and tag me for a little surprise!

Refer our newsletter to a friend, co-worker, or family member.

Subscribe to other newsletters we recommend.

MISSED LAST WEEK’S EDITION?

Find all of our newsletters here.