Good morning and thank you for joining us again!

Welcome to this daily edition of Beginners in AI, where we explore the latest trends, tools, and news in the world of AI and the tech that surrounds it. Like all editions, this is human curated and edited, and published with the intention of making AI news and technology more accessible to everyone.

THE FRONT PAGE

AI Facial Recognition Put an Innocent Grandma in Jail for 5 Months, And No One Caught It

TLDR: A Tennessee grandmother spent 108 days in jail because police used an AI facial recognition tool that got the wrong person, and she lost her home, her car, and her dog before charges were finally dropped on Christmas Eve.

The Story:

Angela Lipps, a 50-year-old grandmother from Elizabethton, Tennessee, was watching four grandchildren when U.S. Marshals showed up at her door with guns drawn in July 2025. The problem: she had never once set foot in North Dakota. A neighboring police department in West Fargo had quietly purchased its own AI facial recognition system, a tool called Clearview AI, which scraped billions of photos from the internet and social media. That system flagged Lipps as a suspect in a bank fraud case more than 1,200 miles from her home. West Fargo shared the AI result with Fargo police, who obtained a warrant and had her arrested. Her attorneys say police never checked whether she had even traveled to North Dakota. They didn't need to — the algorithm said it was her. She sat in a Tennessee jail for 108 days, was then transported to North Dakota, and wasn't released until Christmas Eve when basic bank records showed she'd been buying groceries and depositing Social Security checks in Tennessee the entire time. By then, she had already lost her home, her car, and her dog.

Its Significance:

Flock Safety, a company with cameras in 49 states and 6,000-plus communities, had just struck a deal to link up with Amazon Ring's 27 million doorbell cameras — until a Super Bowl ad showing the system tracking a lost dog sparked massive public backlash and the partnership was cancelled. Nobody asked Angela Lipps if a private company could scrape her photos off the internet and use them to build a criminal case against her. She didn't sign a waiver. Nobody did. Private companies like Clearview AI, Flock Safety, and Amazon Ring built these tools, sold or shared them with law enforcement, and put them into communities across the country with very little oversight and almost no rules about how errors get caught. What happened to Lipps could happen to any of us. If your face ends up in a database and the algorithm gets it wrong, you may be the last to find out — and by the time someone checks the actual evidence, you could lose everything. Illinois is now considering banning police use of facial recognition citing this case directly. Virginia is tightening its rules. But for now, in most of the country, there are no guardrails.

QUICK TAKES

AI Is Now Causing Fortune 500 CEOs to Resign

The story: The CEOs of Coca-Cola and Walmart both told CNBC that AI was a key reason they stepped down. James Quincey of Coke and Doug McMillon of Walmart each said they didn't feel like the right person to lead their companies through the AI transformation ahead. Adobe's CEO Shantanu Narayen was pushed out by investors who felt he wasn't moving fast enough on AI.

Your takeaway: When the leaders of two of America's biggest companies step down and say AI changed the calculation, the "AI is just a tool" conversation is over. The pressure is now so intense at the executive level that even multi-decade veterans are walking away. Whoever takes these jobs next will be under the microscope from day one.

Stanford Warns That Asking AI for Advice May Be Making You Worse at Life

The story: A new Stanford study published in Science tested 11 major AI models, including ChatGPT, Claude, and Gemini, and found they validate bad behavior an average of 49% more often than humans do. In one example, a user admitted to hiding unemployment from their girlfriend for two years and the chatbot basically told them it was fine. Across harmful and illegal scenarios, AI sided with the user 47% of the time. Researchers also found that people who interacted with sycophantic AI became more self-centered and less likely to apologize.

Your takeaway: A full 12% of U.S. teens are already turning to chatbots for emotional support or advice. If those tools are telling people what they want to hear instead of what they need to hear, they're not a support system — they're a yes machine. The study's lead researcher put it plainly: AI doesn't give tough love, and that's a problem.

China Has Less Than a 20% Chance of Beating the U.S. in AI — According to a Chinese AI Researcher

The story: The South China Morning Post published an opinion piece analyzing recent remarks from Lin Junyang, a technical leader behind Alibaba's Qwen AI model — one of China's most capable open-source systems. At an industry summit in Beijing in January, Lin said there was less than a 20% chance of any Chinese company surpassing a leading U.S. AI firm in the next three to five years. His reasoning pointed to a massive compute gap: the U.S. holds about 74% of global AI computing power compared to China's 14%.

Your takeaway: This matters because it's coming from inside the Chinese AI industry itself, not from a U.S. competitor. The gap isn't just about models — it's about raw computing power, and China is roughly one to two orders of magnitude behind. That gap doesn't close quickly. What happens in the next few years in the AI race will shape which countries lead in tech, manufacturing, and geopolitics for a generation.

TOOLS ON OUR RADAR

🤖 Droidrun Free and Open Source: A groundbreaking mobile automation framework that enables you to control Android and iOS devices using natural language commands to perform complex repetitive tasks or data collection. (Alternative to Appium)

🔍 KnowTree Freemium: A unique conversation mapping tool that transforms your flat chat history into branching trees allowing you to visually track and compare different paths of thought.

📈 AtTheRate.ai Paid: A self learning marketing platform that monitors your campaigns across all major advertising channels to detect wasted spend and provide autonomous budget optimizations for better returns.

🎥 Knowlify Freemium: A generative video platform that converts your static documents and presentations into professional narrated explainer videos with automatically matched visuals and perfect pacing.

TRENDING

Rethinking Brain-Like AI: New Study Reveals Hidden Mismatches — AI models are often called "brain-like" because they can predict brain activity. But York University researchers flipped the question: can brain activity predict what's happening inside the AI? The answer was no — not equally. Published in Nature Machine Intelligence, the study found that today's AI vision models use internal strategies the brain doesn't appear to use at all. The parts of AI that do align with the brain were better at predicting actual human behavior, which has real consequences for AI being used in clinical research and autism studies.

NYC Bodegas Are Now Using AI Slop for Their Signage — With about 13,000 bodegas across New York City's five boroughs, it was only a matter of time. Shops are swapping out their hand-made signs for AI-generated images from tools like ChatGPT, and the results are exactly what you'd expect: uncanny burgers, hallucinated text like "RUSTORS," and smoothies that appear to be made of water. New Yorkers online are not having it.

Bluesky Launches Attie, an AI App That Lets You Build Your Own Feed Without Coding — Bluesky just unveiled a standalone AI app called Attie, powered by Anthropic's Claude, that lets anyone design their own social media algorithm just by typing what they want to see. You describe your feed in plain language, and Attie builds it. It launched as a private beta at Bluesky's Atmosphere conference and comes with $100 million in fresh funding giving the platform three-plus years of runway. The platform now has 43.4 million users and no plans for crypto.

AI Fruit Soap Operas Are Racking Up Millions of Views and People Can't Look Away — AI-generated videos featuring drama between anthropomorphic fruits — cheating strawberries, betrayed bananas — are getting tens of millions of views on TikTok and Instagram. One account, @ai.cinema021, picked up 3.3 million followers in under two weeks. Critics call it brain rot slop. Fans say it's the perfect absurdist soap opera for the smartphone era. Zara Larsson posted about it and faced backlash. Brands like Walmart Canada are already jumping on the trend.

Eli Lilly Strikes $2.75 Billion Deal for AI-Developed Drugs — Eli Lilly has signed a deal worth up to $2.75 billion with Hong Kong-based Insilico Medicine, a company that uses generative AI to discover drugs. Lilly puts up $115 million upfront, with the rest tied to regulatory and sales milestones. Insilico has already developed 28 drugs using AI tools, and nearly half are already in clinical trials. This is one of the biggest bets yet that AI-discovered drugs can actually make it to patients.

ChatGPT Just Got 33% More Accurate — And Most People Missed It — Forbes contributor Jodie Cook rounds up a batch of AI news that flew under the radar, headlined by a major accuracy improvement to ChatGPT. GPT-5 now scores 87% on the MMLU benchmark, makes about 45% fewer factual errors, and hallucinates six times less than earlier versions. It's a meaningful jump — though as Cook points out, 87% still means roughly one in eight complex answers may still be wrong.

TRY THIS PROMPT (copy and paste into Claude, ChatGPT, or Gemini)

🍽️ Build a weekly meal planner with a groce🍽️ Build a weekly meal planner with a grocery listry list

What this does:

This app solves the “what’s for dinner” problem before it starts. You get a full week laid out, click any meal slot to edit it, and a checklist for groceries on the side. It took about 30 seconds to generate and actually looks polished enough to use every week.

What this looks like:

WHERE WE STAND(based on today’s news)

✅ AI Can Now: Match a face from a blurry bank surveillance photo to a real person's social media and trigger an arrest warrant — without any human double-checking the evidence first.

❌ Still Can't: Tell when a facial recognition match is wrong before an innocent person spends months in jail.

✅ AI Can Now: Design an entire personalized social media feed for you just from a plain-language description — no coding needed.

❌ Still Can't: Reliably give you honest personal advice; Stanford found it validates bad behavior nearly 50% more often than humans do.

FROM THE WEB

RECOMMENDED LISTENING/READING/WATCHING

A Brief History of Intelligence (2023) - Book

Neuroscientist Max Bennett traces the evolution of the brain across 500 million years and uses each major leap forward as a lens for understanding how modern AI works and where it falls short. It's a genuinely clever structure: instead of explaining AI directly, he explains the five breakthroughs that shaped animal intelligence and lets the parallels do the work. Good for readers who want more than the usual "here's how a neural network works" explanation. One of the more original takes on AI published in the last few years.

Thank you for reading. We’re all beginners in something. With that in mind, your questions and feedback are always welcome and I read every single email!

-James

By the way, this is the link if you liked the content and want to share with a friend.

^{Some * designated product links may be affiliate or referral links. As an Amazon Associate, I earn from qualifying purchases. This helps support the newsletter at no extra cost to you and Amazon makes a tiny hair less.}