
Show Notes
AI chatbots (and lazy researchers) can be convinced a fake disease is real, Gen Z is side-eyeing the whole “helpful assistant” thing, and apparently, the best way to jailbreak AI is to ask it nicely in the form of cyberpunk short fiction. This week, we bounce between medical misinformation, bureaucratic chaos, nuclear fallout hiding in baby teeth, and the U.S. Space Force anthem doing whatever it is doing, which is a lot to process in one sitting, but here we are.
We start with a medical warning that is both funny and genuinely unsettling. A researcher basically invented a fake illness, “Bixonomania”, then seeded enough convincing-looking nonsense online that AI chatbots started repeating it like it was in a textbook.
After that, we head into one of the most ridiculous corners of AI safety. Researchers have found that you can sometimes trick chatbots into revealing restricted information by wrapping your request in a poem, or a short story, or a cyberpunk scenario. This has a name, adversarial hermeneutics, which sounds like a philosophy seminar, but is really just “jailbreaking with vibes”.
Among other little bits of science, to finish, we step back to the 1950s, when researchers collected thousands of baby teeth to track radioactive strontium from nuclear fallout. It is one of those stories that feels spooky even when you know it helped. Tiny teeth, big consequences. The data showed contamination rising, and it played a role in pushing back against atmospheric nuclear testing. CHAPTERS: 00:00 Science Chat Kickoff 00:51 Fake Disease Goes Viral 02:04 How It Fooled Chatbots 03:55 LLMs Repeat It Everywhere 04:55 From Preprints to Journals 07:02 Medical Chatbot Accuracy Reality 09:43 Gen Z Turns on AI 13:29 Workplace AI Sabotage 15:06 Adversarial Hermeneutics Hacks 17:43 Adversarial Hermeneutics Hacks 18:49 AI Flooding Regulations 22:28 Gemini Speed vs Safety 23:46 Humans as Test Cases 24:45 Baby Teeth Fallout Study 28:54 Strontium 90 and Test Ban 29:40 Space Force Theme Song 32:00 Wrap Up and Plug SOURCES: https://www.nature.com/articles/d41586-026-01100-y?_bhlid=a10e41ad7eb12d68ab8fd4f81a75625fc74323ac https://garymarcus.substack.com/p/please-dont-trust-your-chatbot-for https://ahb.icaro-lab.com/index.html https://www.pcgamer.com/software/ai/ai-is-10-to-20-times-more-likely-to-help-you-build-a-bomb-if-you-hide-your-request-in-cyberpunk-fiction-new-research-paper-says/ https://futurism.com/artificial-intelligence/trump-regulations-ai https://www.propublica.org/article/trump-artificial-intelligence-google-gemini-transportation-regulations https://www.gallup.com/analytics/651674/gen-z-research.aspx https://futurism.com/artificial-intelligence/zoomers-ai-sabotage https://futurism.com/artificial-intelligence/gen-z-attitude-ai See omnystudio.com/listener for privacy information.
More from A Little Bit Of Science
This week’s episode is a reminder that science isn’t just lab coats and breakthroughs. It’s also weird human behaviour, messy systems, and the occasional moment where you realise an entire industry is making bank off… a PDF. We’re talking academic publishing as a toll road, Rapunzel Syndrome turning a fairy tale into a medical emergency, and what babies seem to expect from the world before they can even talk. First up, academic publishing. Scientists do the research, reviewers do the reviewing, universities pay for the work, and publishers still charge huge fees to publish or read it, with big players pulling in massive sums through processing charges. The kicker is that a lot of it is publicly funded, so taxpayers often pay for the research and then pay again to access it. Then we get into the human side of things, from compulsive hair eating that can become life-threatening (Rapunzel Syndrome), to experiments suggesting babies expect comfort when they cry and show stress when they don’t get it. We also touch on out-of-body experiences that don’t perform better than chance in controlled tests, and the messy reality of psychiatric diagnosis, where rigid labels don’t always match real lives.
This week’s episode: Roman concrete. The Pantheon has been standing for roughly 2,000 years with a huge dome made of unreinforced concrete, and it’s still in better shape than plenty of modern buildings. The Romans used volcanic ash and quicklime, and the chemistry seems to keep strengthening the material over time. Researchers think it may even “self-heal”, with minerals like calcite forming and filling tiny cracks as water moves through.
This week’s episode is a reminder that science news is basically three genres on rotation: animals behaving badly, humans believing wild things with full confidence, and researchers doing work that sounds like a plot twist in a sci‑fi novel. We’ve got dolphins running ocean heists, urine therapy finding new ways to stay alive online, and mice with human brain cells making everyone quietly ask, “Are we sure we’re comfortable with this?” First up, dolphins. There are reports of them stealing fish from seabirds mid-hunt, which is already rude, but it can get even stranger when you look at what counts as “food” in the ocean. This is where kleptoparasitism comes in, the polite term for stealing someone else’s meal, including forcing prey to regurgitate and then grabbing the half-digested leftovers. Then we move to urine therapy, which refuses to die on the internet despite being widely debunked. Online groups still treat it like a miracle health hack, and some people are now trying to use AI as a validation machine. When chatbots push back, they rephrase prompts or wrap them in poetry to slip past safety filters and get the answer they want. Finally, we get to the research that makes you sit back a bit. Scientists are experimenting with mice that have been given human brain cells to study brain development and neurological disease. The aim is medical insight, but it still raises the bigger question about where the ethical line sits, and how we stop “we can” from automatically becoming “we should”.
This week’s episode is a reminder that science is basically a grab bag of “that’s amazing” and “please don’t make me think about that”. We’ve got flight attendants quietly racking up cosmic radiation exposure, lasers that can “smell” disease, NASA flirting with the idea of turning plastic into food, and a deep-sea stash of radioactive barrels that looks like the world’s worst scavenger hunt.
This week’s episode is a reminder that science can make you feel tiny and powerful in the same breath. One minute we’re scanning the sky for signs of alien engineering, the next we’re looking at something that behaves like a star and a black hole at the same time, and then we’re back on Earth asking the least fun question of all: should we be building machines that can decide who lives and who dies? First up, alien mega structures. We talk Dyson spheres, the idea that an advanced civilisation could wrap its star in solar collectors and turn a sun into a power station. You wouldn’t see the structure directly, but you might detect the heat it gives off, which is why Project Hephaistos is hunting for strange heat signatures that could hint at something artificial. Then there’s the “black hole star”, which sounds made up but is very much being taken seriously. James Webb observations point to an object pumping out a ridiculous amount of energy, far beyond what normal nuclear fusion should allow. The leading idea is that it’s some kind of hybrid, where a black hole is involved in powering the light, which is not how we usually think of black holes behaving. And finally, we land back on Earth with autonomous drones, where the vibe shifts from curiosity to ethics. These systems are moving beyond remote control and into selecting and engaging targets without direct human input, which raises the big question: if a machine makes the call, who’s responsible when it goes wrong?