Hello WKND AI Warriors!

Anthropic’s CEO says we need an "AI MRI" to understand how powerful models really think.

Also, OpenAI's new "open" model might secretly call for help from bigger cloud AIs.

Plus, the White House wants AI education in every American classroom.

Oh yeah, Meta's Ray-Ban glasses can now translate conversations live.

So, grab your beverage of choice.

Here’s your weekly dose of AI news.

Today’s newsletter includes:

• 📰 AI NEWS RECAP

• 🎓 AI COURSES OF THE WEEK

• 🤿 AI DEEP DIVE

• 🛠️ AI TOOL OF THE WEEK

• 📝 AI PROMPT OF THE WEEK

• 🎨 AI IMAGE OF THE WEEK

📰 AI NEWS RECAP

We Need An "AI MRI"

Is bigger always better in AI?

Can we trust AI if we don't understand it?

Experts admit how AI works is a black box.

Anthropic's CEO, Dario Amodei, just published a powerful essay: "The Urgency of Interpretability."

He warns we're building powerful AI without truly knowing how it works inside.

But what is Interpretability?

In a nutshell...

It's like getting an MRI for the AI's brain.

Understanding its inner workings.

Not just what it does, but why.

It helps us:

✅ Understand how AI makes decisions

✅ Detect bias and harmful behaviors

✅ Diagnose problems and errors

✅ Ensure safety and reliability

✅ Build trust in AI systems

Here's how it works:

1️⃣ Find hidden "features"

2️⃣ Analyze AI's internal "neurons"

3️⃣ Map "circuits" between features

4️⃣ Trace the AI's step-by-step thinking

5️⃣ Use AI to help interpret itself ("autointerpretability")

Think about it like this...

Flying a plane without knowing how the engine works.

Doctors prescribing medicine without knowing its effects.

That's where we are with powerful AI today.

We need to see inside the "black box".

𝗠𝘆 𝗧𝗮𝗸𝗲🤔 AI will get smarter.

But will we understand how it works before its too late?

Can we trust AI if we don't understand it?

My hope...

Is that interpretability helps us build AI that's:

✅ Fairer and less biased

✅ Safer and more reliable

✅ Aligned with human values

✅ Truly trustworthy in critical roles

But hey, I'm optimistic.

Will we "see" inside the black box?

Or are we racing towards a Terminator-like future?

Anthropic reveals bad actors use Claude AI for influence campaigns and malware. The company is fighting back with new detection methods to counter sophisticated AI misuse.

OpenAI's new 'open' model might phone home to powerful cloud AI for help. This "handoff" could make free models smarter by tapping into bigger AI brains for complex tasks.

Talk to anyone, Meta's Ray-Ban smart glasses now offer live translation for all users. The AI feature translates conversations in real-time directly through the glasses.

Microsoft Copilot adds AI agents and better search for teamwork. New Agent Store and Notebooks help turn data into action, boosting human-agent collaboration in Microsoft 365.

White House orders AI education push for American students. New task force aims to teach AI literacy in K-12 schools and train teachers on the technology.

YouTube tests AI summaries in search results, showing video clips directly. The feature, similar to Google Search AI Overviews, is rolling out to select US Premium users.

🎓 AI COURSES OF THE WEEK

The harsh truth…

Only 12.6% of people ever complete online courses.

So I’m launching a private AI training just for you.

Why?

Research shows people who take live trainings have an 85%+ completion rate.

Why my course?

As an ex-EY (Ernst & Young) Senior Manager:

✅ I led an Intro to GenAI training to 250+ EY employees.

✅ Supported GenAI solutions for Fortune 500 clients.

✅ Was nominated to EY’s US AI Adoption Network.

✅ Was asked to speak at GenAI conferences.

But like many of you, I didn’t know where to start.

Since March 2023 I’ve spent 1,000+ hours learning AI.

My goal?

To save you time. Simple as that.

Here’s what you’ll get:

☑ 4 weekly 120-minute AI training sessions

☑ 8 hour AI for Business & Work masterclass

☑ Ongoing Q&A support

Ready to master AI?

Click below to join my waitlist.

🤿 AI DEEP DIVE

Discover how AI is evolving to anticipate questions even before they're asked, exploring a new method called "sleeptime compute".

This video breaks down how such advancements work, simplifying complex ideas and revealing the potential for improved efficiency and reduced costs in practical AI applications.

🛠️ AI TOOL OF THE WEEK

Doable: An AI-power web app creator that takes action with simple English commands instead of complex clicks, using natural language operators.

Embed powerful workflow automation and streamline user experiences with just one line of code to significantly boost satisfaction, activation, and retention.

📝 AI PROMPT OF THE WEEK

Copy and paste this into your favorite chatbot.

Code Debugging Guidance 

You are a senior software engineer proficient in [INSERT LANGUAGE/FRAMEWORK]. 

I'm encountering [INSERT ERROR DESCRIPTION] in the following code snippet: [INSERT CODE]. 

Identify likely causes and propose step‑by‑step fixes. Numbered list of debugging steps. 

[INSERT TONE]

🎨 AI IMAGE OF THE WEEK

Give AI image creation a try for yourself.

Copy the text below into Google’s AI image creator!

Paris fashion week, Cottagecore super model with stunning eyes, experimental dark aesthetic style polaroid, dizzy effect, Dramatic lighting, mirrorglass effect, candid photo, fall aesthetic, fine grain, neon, ascii, bokeh, pointillism, Fall and fresh aesthetic color

How can you help?

Refer my newsletter to help others learn AI.

Missed last week’s edition?

Find all of my newsletters here!