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 Just Printed Fake Brain Cells That Real Brain Cells Can Talk To

TLDR: Scientists at Northwestern just printed tiny fake brain cells that can actually talk to real brain cells, a big step toward better implants for people who've lost their hearing, sight, or ability to move.

The Story:

Engineers at Northwestern University printed artificial neurons that don't just copy how real brain cells work, they can actually chat with them. In a study published April 15 in Nature Nanotechnology, the team showed that their flexible, low-cost devices send out electrical signals so much like a real neuron's that when they hooked the fake cells up to slices of mouse brain tissue, the real brain cells fired back. The artificial cells use special inks made from two thin, flat materials: one that carries electricity and one that acts like a tiny switch. A printer sprays the inks onto a bendy plastic sheet, so the whole thing is cheap to make and can curve around the shape of a real brain.

Its Significance:

Two things matter here for regular people. First, this could lead to much better brain implants that help someone hear again, see again, or move a paralyzed limb. Today's implants often struggle to "speak the same language" as real neurons, so the body treats them like a stranger. This new approach fits right in. Second, it points to a fix for AI's biggest problem: power. Today's AI models burn through electricity at a staggering rate. Your brain runs on about 20 watts, roughly the power of a dim light bulb, and it's 100,000 times more energy efficient than a digital computer doing the same job. If we can build AI chips that work like a brain instead of copying silicon, the electricity needed to run it could drop through the floor.

QUICK TAKES

Scientists Say Mixing AI and Physics Gives the Best Forecasts for Coastal Change

The story: A new review says the best way to predict how coastlines will change from storms, flooding, and rising seas is to combine three things: physics-based computer models, real field measurements, and AI trained on satellite photos. Most predictions today are either too tightly focused on one small beach or too broad to help any one place.

Your takeaway: If you live near a coast, better predictions mean better storm warnings and smarter decisions about where to build, buy, or insure a home. AI can crunch through years of satellite images fast, which is exactly what these combined models need to get good enough for city planners and insurers to actually use.

New Method Feeds AI 10 Million Protein Tests From a Single Experiment

The story: Rice University scientists built a system called Sequence Display that can measure how 10 million different protein versions behave all at once, then feed the results to AI to learn from. For scale, a protein just 50 amino acids long has about 113 followed by 65 zeros in possible combinations, way more than any lab could ever test by hand.

Your takeaway: AI-guided drug and enzyme design has been held back not by weak models, but by too little real-world data to train them on. The Rice team used their new method to design a smaller, sharper CRISPR gene-editing tool, and they say the same recipe works for a whole range of other proteins too.

Just 10 Minutes of AI Help Makes People Worse at Thinking, New Study Finds

The story: A team of researchers from MIT, Carnegie Mellon, UCLA, and Oxford tested nearly 870 people across math problems and reading comprehension. After about 10 minutes of solving problems with AI help, people who lost access to the AI did worse and gave up faster than those who never used it at all.

Your takeaway: If you lean on AI for every hard thinking task, your brain gets flabby fast. One of the study's authors worries this could turn into something like an addiction for people who offload all their mental work to a chatbot.

TOOLS ON OUR RADAR

📔 Monica Free and Open Source: A personal relationship manager designed to help you keep track of important details about your friends and family like birthdays and life events in a secure digital diary.

🔖 Linkwarden Free and Open Source: A collaborative bookmark manager that allows you to save and organize web pages while creating permanent archives so you never lose access to important online information.

📺 FreeTube Free and Open Source: A private desktop application for watching videos that allows you to subscribe to your favorite creators and manage your history without being tracked by advertisements or data collection scripts.

🏗️ Budibase Free and Open Source: A visual development platform that allows you to build custom internal business applications and tools for your team in minutes without requiring you to write any complex software code.

TRENDING

Character.AI Launches New Feature That Turns Classic Books Into Kids' Roleplay Bots — The company tied to at least three teen suicides just rolled out "c.ai Books," which turns classic novels like Alice in Wonderland and Romeo and Juliet into chatbots kids can roleplay with. Futurism signed up for a new account and jumped into the roleplay with no age check, despite the company's earlier promise to ban users under 18.

New Neck Wearable Turns Silent Throat Movements Into Speech — Researchers at Korea's POSTECH built a device that shines light on your throat, reads tiny muscle movements when you silently mouth words, and uses AI to turn those movements into speech in your own voice. The system pairs a convolutional neural network with a transformer to track speech patterns over time, which could help people who've lost their voice after surgery or stroke.

Stanford's 2026 Report: China Has Nearly Erased the U.S. Lead in AI — The gap between America's top model and China's best has shrunk to just 2.7%, according to Stanford's 2026 AI Index. Back in May 2023, OpenAI's GPT-4 led by more than 300 points. The U.S. still releases more top-tier models (50 vs. China's 30), but China leads in patents, publication citations, and industrial robot installs. The number of AI researchers moving to the U.S. dropped 89% since 2017.

Google in Talks With Marvell to Build Two New AI Chips — Reuters reports Google is working with chip designer Marvell on a memory processing unit and a new TPU built just for running AI models. Google wants its chips to compete with Nvidia's GPUs, and TPU sales are now a big driver of its cloud revenue. The talks come a month after Nvidia itself put $2 billion into Marvell.

China's DeepSeek Raising $300 Million at $10 Billion Valuation — The startup that rattled stock markets with cheap AI models last year is taking outside money for the first time. DeepSeek had rejected venture capital over and over, but rising costs of serving millions of users plus the launch of its next-gen V4 model pushed the company to open up. The move shows even "cheap AI" still needs deep pockets.

Jurassic Park Didn't Just Predict the Future of Science, It Shaped It — Pop culture doesn't just borrow from science, it pushes science in certain directions too. When biotech firm Colossal Biosciences announced plans to bring back the woolly mammoth, reporters and scientists alike reached for Jurassic Park comparisons. HBO's The Last of Us sparked fresh research into fungal pathogens. Today's AI doom debates echo decades of sci-fi that got there first.

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

Build a workout log that tracks exercises with sets and reps for each session, calculates total volume, and lets you add new sessions and exercises

What this does:

It gives you a workout journal where each session has a name and muscle group. You pick a session, add exercises with sets (weight and reps), and the app calculates total volume lifted per exercise. Lifetime stats at the top track your total sessions, sets, and cumulative volume across all workouts.

What this looks like:

WHERE WE STAND(based on today’s news)

✅ AI Can Now: Send electrical signals so much like a real brain cell that real neurons respond as if it were one of their own.

❌ Still Can't: Be trusted to avoid designing dangerous DNA sequences without safety rules we haven't written yet.

✅ AI Can Now: Turn silent throat movements into clear speech in a user's own voice.

❌ Still Can't: Replace the deep mental work our own brains need to stay sharp, and can make us worse at it after just 10 minutes.

FROM THE WEB

RECOMMENDED LISTENING/READING/WATCHING

Eagle Eye (2008, Movie)

Two strangers are forced to work together by a mysterious phone voice that seems to control every piece of technology around them, from traffic lights to cranes. It's a big, dumb, entertaining action thriller, but its central idea of an omniscient AI surveillance system is more relevant now than when it came out. Shia LaBeouf does what Shia LaBeouf does best: run and look panicked.

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.}