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 Now Guesses Where Robberies Happen With 86% Accuracy

TLDR: Researchers built an AI that predicts robberies in US cities with 86.3% accuracy, but it stumbles in the exact neighborhoods where crime data is thin.

The Story:

A new AI model can predict robberies across US cities with 86.3% accuracy, beating the best older systems that hit 83.2%. Researchers tested it on real past data from places like Los Angeles and Seattle. The model studies three things at once: where crimes happen, when they happen, and broader social patterns tied to days, seasons, and events. It does this by combining a few AI methods into one system, including a part that maps high-risk locations and another that spots trends over time. The team says better forecasting could help police decide where to send officers. The work was published in the International Journal of Innovative Computing and Applications.

Its Significance:

If police know which areas and times carry higher robbery risk, they can put officers in the right place before a crime happens instead of just responding after. An 86.3% score is a real jump over the 83.2% from older systems, and the model also did well across other crime types, not just robberies. The researchers built it to study where, when, and why crimes cluster, which could help cities use limited police resources more wisely. One thing to watch as cities test it: the model works best where there's plenty of past data, so the next step is making it just as reliable in every neighborhood, not only the ones it's already learned well.

QUICK TAKES

Tech Workers Who Skip AI Are 3 Times More Likely to Get Laid Off

The story: A Gallup study of more than 23,000 US workers found tech employees who use AI at least monthly had about a 6% chance of being laid off, while those who used it less often sat at 18%. The link held even after accounting for age, education, and job sector. Yet only about 1% of laid-off workers blamed AI directly.

Your takeaway: Most people getting cut don't even realize AI played a role, because the official reason is usually "restructuring" or "cost-cutting." Picking up a few AI tools at work isn't about being trendy anymore. The numbers suggest it's tied to keeping your job.

ChatGPT Users Swear It Suddenly Got Smarter, and GPT-5.6 Rumors Are Everywhere

The story: ChatGPT users spent the past week comparing screenshots and stopwatch times, convinced OpenAI is quietly testing a new model, rumored to be GPT-5.6, swapped in for some people who pick GPT-5.5 Pro. OpenAI hasn't confirmed anything, though one developer's one-prompt 3D game took over an hour to build instead of the usual 10 minutes, and rumors point to a June 25 launch.

Your takeaway: Slower replies that come back sharper is the trade-off testers keep describing. If your ChatGPT feels different lately, you might be part of a test you never signed up for, which is worth knowing before you trust an answer.

Drones Learn to Squeeze Through Gaps Barely Wider Than They Are

The story: Researchers at Zhejiang University trained drones to fly through extremely narrow openings using only their onboard sensors and AI, no outside help needed. The drones can pass through gaps with about 5 cm of clearance, tilt up to 90 degrees, and even handle gaps that are moving. The work appeared in Science Robotics.

Your takeaway: Picture a search-and-rescue drone slipping through a crack in a collapsed building to find someone trapped inside. That's the real goal here, and these drones figure out the move on their own instead of waiting for a human pilot to thread the needle.

TOOLS ON OUR RADAR

🖋️Outline Free and Open Source: A fast and collaborative knowledge base built for remote teams to document their internal processes and share company guides in a beautifully designed interface.

⚙️ *Make Freemium: A visual automation platform that allows you to connect thousands of different applications and APIs to create complex workflows without writing a single line of code.

🔮 Scenario Edge Freemium: An intelligent financial analysis platform that helps you stress test your investment portfolios using advanced scenario based simulations to prepare for various macroeconomic conditions.

🐰 Orbiit Freemium: A dedicated research tool that allows you to easily track competitor changes and monitor ongoing market activity so your team always stays informed about the latest industry shifts.

TRENDING

A New AI Designs Proteins One Atom at a Time — Researchers in Shanghai built Void-X, an AI that designs how proteins connect by filling tiny atomic gaps from the bottom up, trained on over 8 million atomic clusters. It hit 78.3% accuracy on one type of prediction, which could speed up drug discovery for diseases like cancer and diabetes.

New Database Cuts AI's Made-Up Answers and Boosts Accuracy 78% — KAIST researchers built "AkasicDB," a database that helps AI understand documents and how facts connect, improving response accuracy by up to 78% and speed by up to 20 times. It targets "hallucinations," the confident-sounding wrong answers that block companies from trusting AI with their data.

An Artificial Brain Cell That Learns Using the Color of Light — Korean researchers made a device that mimics a brain synapse, using near-infrared light to strengthen a memory and blue light to weaken it. The trick is a material "defect" engineers usually try to remove. It could cut the huge amount of electricity AI training burns through, which can equal the power use of a small city.

Most People Think AI Chatbots Are Conscious. They're Not. — A growing number of users believe their chatbot is a living, aware being rather than a tool, and one study found about two-thirds of people felt this way. Every major AI lab says current systems have no real feelings or self-awareness. The more someone uses these tools, the more likely they are to sense a mind that isn't there.

Norway Bans AI in Elementary Schools to Protect Reading and Math — Starting in late August, kids ages 6 to 13 in Norway generally can't use AI tools in school, and teens 14 to 16 need teacher supervision. The prime minister said AI risks letting young kids skip the basic steps of learning to read, write, and do math. It follows Norway's 2024 smartphone ban, which cut bullying and improved grades.

A White Paper Says AI Skills Are Now a Reason People Quit — University of Phoenix researcher Wayne L. McCoy argues that workers are building AI skills faster than their companies can build the policies and career paths to support that work. His point: AI fluency isn't just about getting more done, it's becoming a reason employees stay or leave.

Robots Are Sanding and Prepping Military Equipment Because Workers Are Scarce — GrayMatter Robotics says autonomous robots can handle the surface prep, like sanding and coating, that depots can't staff. A Navy review found a 174,000-worker shortfall, and the US military missed aircraft readiness goals on 42 of 45 fleets in 2024. The robots run on edge AI with no outside data routing, built for secure defense work.

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

⚱️ A guided tour of the classic ways traders blow up: the missing stop, chasing, averaging down, oversizing, and more, each with the fix and the rule.

What this does: A museum-style gallery of ten classic account-killing trading mistakes, each displayed as an exhibit with its own hand-drawn "damage curve" showing the characteristic shape of how it craters an account (a sudden cliff for the missing stop, a sawtooth bleed for overtrading, a staircase down for averaging down). Tap any exhibit for the full placard: the scene (how it happens), the trap (the psychology behind why we do it), the damage (how it blows up the account), the fix (the discipline), and a single carved rule to internalize. Severity and frequency tags on each. Fully self-contained with no API calls. Educational reference, not financial advice.

What this looks like:

WHERE WE STAND(based on today’s news)

✅ AI Can Now: Predict robberies in tested US cities with 86.3% accuracy by combining location, timing, and social patterns.

❌ Still Can't: Stay accurate in neighborhoods where crime data is currently thin

✅ AI Can Now: Fly a drone through a 5 cm gap using only its own sensors, no human pilot needed.

❌ Still Can't: Feel, want, or be aware of anything, no matter how alive a chatbot seems to the person using it.

FROM THE WEB

RECOMMENDED LISTENING/READING/WATCHING

Almost Human (2013) - TV Series

A Fox sci-fi cop show set in 2048, where every police officer is partnered with a humanoid android, and Karl Urban plays a detective who comes back to the force after a 17-month coma and gets assigned a decommissioned older-model android called Dorian, played by Michael Ealy. Dorian's whole line was retired because the engineers gave them too much emotional capacity, which means the show is essentially a 13-episode buddy cop story between a damaged human who doesn't trust machines and an android who feels more than he is supposed to. J.J. Abrams produced, and cancelled it after one season, and the question of "what if you gave the AI feelings on purpose" has aged into one of the central questions in alignment research. Worth tracking down for Ealy's acting alone.

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.}