Quantum Technology Matters for Media Literacy Educators

Ekin Dogus Cubuk grew up in Western Turkey, writing basic physics simulations in QBasic, dreaming of becoming a theoretical physicist. Years later, he met Liam Fedus, a fellow physicist who would go on to help build ChatGPT and eventually run post-training at OpenAI. The two first bonded at the gym, flipping a tire. Their real obsession was physical science and a growing conviction that AI had hit a digital wall. An LLM can write flawless code, but it cannot make an atom move. Liam calls this the limit of "thinkism": an AI will never think for a trillion tokens and suddenly intuit a room-temperature superconductor.  So they founded Periodic to build what they call a "synthesis superintelligence," a closed-loop system of autonomous labs and AI that can intentionally design and synthesize physical materials. In this conversation, we get into what it actually takes to engineer matter, how the team cut a projected $30 million analysis cost down to sub-$1 million, and the dream experiments they would run with unlimited capital. Full episode linked in the comments.

The most powerful AI models are the ones you pay for. What happens when college students who are one emergency from dropping out can't afford access? In our second episode of Ignore All Previous Instructions, Stephanie O'Donoghue, Director of Advanced and Emerging Technology at the Technical College System of Georgia, shares how the college is putting these tools in every student's hands. Watch the full episode here: https://lnkd.in/gXJnBkNS #AIPolicy #AILiteracy #Georgia

Vitalik Buterin learns hard topics from AI by explaining the answer back to it in his own words. Vitalik shared how he learned obfuscation well enough to publish a blog series on it. He keeps asking questions, then restates everything in his own words and asks the AI whether he’s right. He repeats that until it stops listing improvements. His reasoning is that you learn better by writing than by reading, and it stops you from fooling yourself. I think that’s the practical lesson here. AI makes learning faster only if you test yourself against it, rather than just collecting answers. https://vist.ly/5pisr

Google opened SynthID Detector globally in English, letting people check images, video and audio for the company’s SynthID watermark. Google says the tool also checks supported content from partners including OpenAI, NVIDIA and Kakao, with Apple support planned. Important limit: this checks for SynthID signals, not every kind of AI-generated media. A missing match does not prove content was made by a person. #AI #Google #SynthID #MediaLiteracy #Technology

Andras Molnar of TUM Think Tank, Steve Wood of PrivacyX Consulting, Nick McQueen of Access Partnership, Urs Gasser of Technical University of Munich, and Sandra Cortesi of Technical University of Munich discuss how emerging policy responses address changing opportunities and risks regarding Generative AI and children International Report October 2026 #dataprotection #privacylaws https://lnkd.in/dCpPgmxD

With more people turning to LLMs like ChatGPT, Claude, and Gemini during personal challenges, concerns are growing over whether AI is truly equipped to offer emotional support. “A lot of people are using LLMs to help process the death of a loved one, or maybe their girlfriend broke up with them, or their boyfriend cheated on them,” said Glenn Matlin, lead researcher and Ph.D. student in the Georgia Tech School of Interactive Computing. To address this, Georgia Tech researchers have developed a new neural framework to evaluate how different training sources shape empathy and social reasoning in these models. Read more: https://lnkd.in/ewPNmPYX #GTComputing #TogetherWeCompute #ChangeTheGame BONUS 🐝- GT@CoLM 2026: https://lnkd.in/eAPU5Vd4

Four parties. Four conditions. One bridge. Education, government, AI, and the user/student increasingly overlap and influence the same person. For that relationship to work, four conditions have to be present: clarity, transparency, honesty, and trust. At the center is collaboration. Collaboration is the bridge that turns separate systems into a functioning community. Without it, the systems may be connected, but the relationship remains fractured. The person cannot simply be taught, evaluated, governed, monitored, or processed. They have to be part of the process. Technology can connect systems. Collaboration is what connects people. Figure created from the original conceptual framework by Amy Watts, with visual development assistance from ChatGPT by OpenAI, October 2026.

AI behaves more like a brain than a database – cognitive science’s role in its origin story helps explain why: Modern AI owes its existence to both cognitive and computer science. What you envision as its origin story can influence how you understand what large language models can do. https://lnkd.in/e8ZYp3vR #ArtificialIntelligence #CognitiveScience #MachineLearning #NeuralNetworks #LanguageModels

As more tasks are delegated to AI agents, human judgment becomes more important to orchestrate their operations. Another interesting article by Cassie Kozyrkov.

Terence Tao warns that the race to AGI is becoming toxic and scary, stating that the field is sacrificing both understanding and safety. He notes that while chatbot errors were once charming, AI systems are now being placed in charge of reward systems, raising serious concerns about oversight and control. The mathematician also highlights the impact on early-career researchers, noting that many junior research projects have been wiped out overnight and some papers have been retracted due to errors. Tao added that there is no clear picture of current developments and that no one from OpenAI is providing clarity to the research community. #ArtificialIntelligence #AIInnovation #EmergingTech #Innovation