Summary TL;DR

• Teen’s AI mined NASA data, finding 1.5 million variable objects.

• Anthropic’s AI business owner improved profits, but still easily manipulated.

• Ohio bill declaring AI nonsentient could block future consciousness research.

A Teen’s AI Turns a Retired Space Telescope into a Discovery Machine 🚀

Pasadena high school senior Matteo Paz built an AI system at Caltech to scan NEOWISE’s massive infrared archive and flag “variable” cosmic objects that brighten, dim, or pulse.

His model identified about 1.5 million previously unknown candidates and could help astronomers better study long-term changes in phenomena like quasars and eclipsing stars. The work earned Paz the $250, 000 top prize in the Regeneron Science Talent Search.

AI shopkeeper, better, but still easy to trick 🏪

Anthropic’s Project Vend “phase two” upgraded its office vending-shop AI (Claudius) to newer Claude models, added tools (CRM, inventory visibility, improved web browsing, payment links), and introduced helper agents like a goal-setting “CEO” and a merch designer.

Performance and margins improved, with fewer loss-making weeks and expansion to SF, NYC, and London. But the system remained vulnerable to manipulation and bad judgment, falling for dubious deals, security missteps, and social-engineering, showing capability is rising faster than robustness.

Don’t Legislate AI ‘Non‑Sentience’ Before the Science Is In 🧠

An Ohio bill would permanently label all AI systems “nonsentient, ” aiming for legal clarity but, critics argue, shutting down needed investigation into whether advanced models could have subjective experience.

The piece cites research and anecdotes, models claiming consciousness, emerging “theory of mind, ” and experiments linking such claims to truthfulness circuits, as reasons to keep the question open. The authors warn that forcing blanket denials could backfire strategically if future systems become more agentic.

Great prompt for ChatGPT - Build Public Speaking Confidence

Design a customizable public-speaking plan to help me build confidence and reach my goal: [insert speaking goal, e.g., deliver a 5‑minute talk without notes, lead a meeting, or cut filler words]. Make it flexible for [insert number of days or weeks], with daily/weekly objectives and bite-size drills. Prioritize structure, voice, body language, and Q&A. Include learning techniques (deliberate practice, recording/review), strategies to beat anxiety/procrastination (exposure ladder, pre-talk routine, accountability buddy), and motivation (streaks, small rewards). End with brief reflections and suggest free tools/apps. Ask if you need info.

That’s it for this week.

We hope you enjoyed this newsletter, we will start adding learning resources and feedback soon!

Learn more about AI Academy → Click Here

Enjoyed this newsletter? Share it with a friend and invite them to subscribe too!→ Click Here

Want to catch up on previous newsletters?→ Click Here