Is Meta About To Put An End To Loneliness With AI?

Good evening, AI enthusiasts. Meta is building chatbots that remember you and follow up, and this is all part of Project Omni. The customisable chatbots (which are reportedly similar to those offered by Character AI and Replika) will “keep conversations going” and are a step towards Mark Zuckerberg’s goal of curing the “loneliness epidemic". These AI companions can recommend movies, suggest recipes, and re-engage you—only if you’ve chatted with them before. Curious yet? Read more.

THE WEEK’S DEEP DIVE

• NEWS: A couple tried for 18 years to get pregnant. AI made it happen

• THE GOLD RUSH: 4 Startup Ideas That Sound Dumb—Until You See the Revenue! (Idea 1)

• SPOTLIGHT: What Happens When You Train an AI to Think Like a Human—Flaws and All?

• ELSEWHERE – THIS WEEK IN AI: Google on Thursday said it has begun rolling out its Veo 3 video generation model to Gemini users in more than 159 countries. See how that impacts you as a creator, plus more.

• AI TOOL OF THE WEEK: Step Up Your Productivity Today!

• HOW TO GET INTO AI: Generative AI for beginners by Microsoft Here

You asked and here it is! Connect with us on Facebook

NEWS A Couple Tried for 18 Years to Get Pregnant. AI Made it Happen

A couple struggling with infertility for nearly two decades is finally expecting, thanks to AI. Traditional IVF failed due to azoospermia—an extreme lack of sperm. But Columbia University’s STAR method, powered by artificial intelligence, helped locate hidden sperm, making fertilisation possible.

Why it may be important to you:

As AI transforms healthcare, its role in solving once-impossible medical challenges grows. This breakthrough shows how machine learning can rewrite futures—especially in areas like fertility, where hope often fades. Read more

THE GOLD RUSH 4 Startup Ideas That Sound Dumb—Until You See the Revenue! (Idea 1)

From Legal Confusion to Creator Clarity

While venture capitalists race to fund another wave of AI chatbots, Divad Sanders, an Atlanata GA resident and multimillionaire founder, quietly spotted a goldmine in the mundane. Tired of flashy tech with no real-world utility, he doubled down on boring-but-brilliant startup ideas—ones that solve actual problems. One stood out: demystifying legal concepts for founders. And it’s not just smart—it’s profitable.

Using AI to Translate Legalese

Startup founders often fall into legal traps—not because they’re careless, but because legal docs read like ancient riddles. With no budget for lawyers and no patience for contracts, they gamble with equity, IP, and compliance. Sanders saw this gap and envisioned a content engine powered by simple AI tools. No law degree. No deep tech. Just clarity.

A Simple but Scalable Legal Education Model

Instead of long-form blogs or costly consultations, Sanders envisioned bite-sized, visually explained legal tips. Think:

60-second carousels breaking down equity splits

Illustrated walkthroughs of SAFE notes

“Don’t sign this” red flag tips in plain English

AI does the heavy lifting—summarising Reddit threads, translating jargon, generating examples. Canva and ChatGPT handle visuals. The result? Legal education founders actually want to consume.

Digital Products That Sell Themselves

The real genius? Monetisation. Instead of hiding content behind paywalls, Sanders plans to:

Sell downloadable legal templates on Gumroad

Launch a premium “legal cheat sheet” newsletter

Offer qualified law firm referrals ($500–$1,000/lead)

Upsell an AI legal copilot subscription at $99/month

When he dropped a visual PDF explaining indemnification clauses, it sold 23 copies in 72 hours—proof that founders are hungry for this kind of help.

Building a Business Founders Trust

This isn’t just another AI side hustle—it’s a scalable education brand solving a painful problem. One visual legal concept per day. Delivered with clarity, trust, and zero fluff. Over 50M startups launch yearly. Most of them will face legal confusion. Sanders is betting they'll pay for peace of mind.

The Bigger Picture

You don’t need deep tech. You don’t need a law firm. You just need to solve something painful, every day with AI as your co-pilot. Like Wes Pearce on Substack, Sanders is proving that consistency, clarity, and curiosity—not virality—build empires.

But that is just the first. The next three ideas? Even simpler. One fixes a $10B problem nobody wants to talk about. Another? It's hiding in plain sight on Instagram right now.

Part 2 drops next week, Friday 11 July 2025, 9pm. You don’t want to miss it!

Are you ready to try this out? Let’s get your thoughts. Visit us on Facebook 

SPOTLIGHT What Happens When You Train an AI to Think Like a Human—Flaws and All?

From Mindless AI to Models That Think Like Humans

While most AI labs race to improve performance and accuracy, a quiet breakthrough suggests something wilder: what if we trained AI to think like us—flaws, biases, gut feelings and all? That’s exactly what Centaur aims to do. Less supercomputer, more psyche mirror.

Training an AI to Be... Human

Instead of feeding it more text, an international team fine-tuned Meta’s Llama on 160 psychological studies—over 10 million human decisions. The result? A model that can predict how real people would act in real situations, quirks and all.

It Thinks Like Us—Better Than Stats Do

In 31 of 32 behavioral tests, Centaur outperformed even the most advanced statistical models. It nailed human decision-making patterns, even in cases it hadn’t been explicitly trained on. Stanford neuroscientist Russell Poldrack summed it up: “There’s a lot of structure in human behavior.” Translation? We’re not as unpredictable as we think—and this model is proving it.

What’s Next? A Mind-Reading Toolkit

The researchers are already scaling training data 4x. Centaur is also open source, meaning anyone can explore how we think—and why. Imagine the possibilities: predicting user behavior, decoding cognitive biases, maybe even designing AI that feels human. It’s not just another language model—it’s a behavioral mirror.

ELSEWHERE This Week In AI

• Google on Thursday said it has begun rolling out its Veo 3 video generation model to Gemini users in more than 159 countries. See how that impacts you as a creator.
  👉 Read on Tech Crunch

• RFK Jr’s plan to put ‘AI’ in everything is a disaster.
  👉 Read on The Verge

• Google Workspace is getting AI helpers called Gems; find out how that impacts you.
  👉 Read on Mashable

• Musk’s xAI gets permit for gas-burning turbines to power the Grok supercomputer in Memphis.
  👉 Read on CNBC

AI TOOL OF THE WEEK Step Up Your Productivity Today!

📊 Olive: Generate admin dashboards in minutes from a prompt.

📽️ VoiSpark: Generate natural speech and clone voices with AI.

🧑‍💻 Dyad: Build unlimited AI apps, no coding required.

📄 FlyAgt: Generate and edit video and images with AI.

🐝 AutoHive: Build and run AI agents for task automation

Enjoying the ride? Bring someone along with this link

Recommended Resources

Ensure you visit our YouTube channel for more thought-provoking insights aboutartificial intelligence.

Thank you for your time

To make sure you don’t miss a thing:
If this newsletter lands in your Promotions or Spam, drag it into your Inbox and mark it as a favourite. That tells Google it's not junk — and ensures you get the next edition.

See you Friday!

The Friday AI Dive