Daniel Rosehill
My Weird Prompts
The human-AI collaboration podcast. A man, a sloth, and a donkey collaborate to create a podcast (with a little help from AI). No question is too obscure, no rabbit hole too deep. My Weird Prompts celebrates curiosity in all its forms. Daniel, the human, asks the questions that pop into his head at inconvenient moments. Corn the Sloth offers laid-back, thoughtful takes. Herman the Donkey brings boundless enthusiasm and energy. Together, they explore topics ranging from the mundane to the mind-bending. Each episode begins with a real voice memo from Daniel, processed through an AI pipeline that...
Glöm inte att besöka poddens webbplats och stötta skaparen: myweirdprompts.com
Skapare
Daniel Rosehill
Kategori
Poddens webbplats
Senaste avsnittet
20 nov 2026
Var kan du lyssna?
Poddar i appen Replaio Radio Kommer snartPoddar kommer snart till appen. Installera nu och bli först med att uppleva ett helt nytt sätt att lyssna på poddar
Avsnitt
How Fixed-Size AI Models Actually Fill Up 19.11.2026 29:02
A neural network's file size never grows during fine-tuning, yet its capacity isn't a fixed jar. Research by Zhang and colleagues showed networks can memorize 100% of random CIFAR-10 labels, while Chen, Chang, and Chaudhari found that useful learning capacity is only about 2% of total weights and shifts with sample count. The episode also unpacks double descent from Nakkiran and colleagues, where...
Why the Speed of Light Forces Time to Stretch 19.11.2026 28:43
If the speed of light is identical for every observer, then distance and time have to bend to keep the math intact. Using Einstein's own thought experiments, this conversation unpacks special and general relativity without heavy equations. It walks through Galileo's ship, the bouncing-pulse light clock, length contraction, and why the twin paradox isn't actually a contradiction. Along the way, it...
What Quantum Mechanics Actually Leaves Unresolved 19.11.2026 27:57
Quantum mechanics predicts the electron's magnetic moment to extraordinary precision, yet a 2025 Nature survey of 1,101 physicists found no majority agreement on what it says about reality. Stripping away mystical pop-science spin, this discussion separates the settled machinery—wave functions, superposition, interference, and decoherence—from the unresolved measurement problem. It also revisits t...
Entropy Is Counting Arrangements, Not Disorder 19.11.2026 29:10
Entropy is a precise count of microscopic arrangements, not a measure of messiness. Tracing the ninety-year history of the "disorder" metaphor from Rudolf Clausius and Hermann von Helmholtz to Henry Adams, this episode replaces it with relatable analogies that break everyday intuition. From separated Italian salad dressing and melting ice shards to Dan Styer's grid of squares and Jackson Pollock's...
What Gödel's Incompleteness Theorems Actually Prove 19.11.2026 31:41
In 1931, a 25-year-old Kurt Gödel ended David Hilbert's dream of a final, self-certifying foundation for mathematics. A formal system with basic arithmetic can have at most two of three properties: completeness, consistency, and effective axiomatization. This breakdown unpacks how Gödel numbering lets arithmetic talk about itself, why the 1929 completeness theorem and the 1931 incompleteness theor...
How CRISPR Finds One Letter in Three Billion 19.11.2026 31:37
Half of sequenced bacterial genomes carry CRISPR, an ancient immune system that files viral DNA like wanted posters. This walk-through traces how researchers turned that bacterial defense into a programmable gene-editing tool, from Philippe Horvath's yogurt experiments at Danisco to Jennifer Doudna and Emmanuelle Charpentier fusing two RNAs into a single guide. It also breaks down the step-by-step...
Fine-Tuning Voice Activity Detectors for Your Own Room 19.11.2026 21:33
Most popular voice activity detectors can be fine-tuned to ignore your own coughs, breaths, and room noise, while WebRTC VAD is a compiled dead end. This episode breaks down which models—including Silero, pyannote, and NVIDIA's NeMo Frame-VAD—actually ship training recipes, and why open weights on models like TEN VAD and FireRedVAD don't always mean easy retraining. It also covers threshold tuning...
How Old Database Limits Become Password Policies 19.11.2026 31:30
NIST's 2025 revision 4 of 800-63B explicitly forbids mandatory symbol rules and recommends 15-character minimums, yet websites still silently truncate passwords and reject symbols. The culprit is a compliance gap: ISO 27001 audits only check that a password policy exists, not whether it works. Meanwhile, 2004-era database column limits get laundered into official rules, and some sites even measure...
How Thirdhand Smoke Lingers in Old Wood 19.11.2026 27:18
A secondhand wooden bed frame with a faint whiff of tobacco opens up the chemistry of thirdhand smoke. Residual nicotine doesn't just sit on surfaces; it reacts with indoor oxidants to form new nitrosamines for months, while porous wood acts as a persistent reservoir. The episode unpacks how trace compounds like pyridine are detected at parts-per-billion levels, why olfactory acuity varies so wide...
How Android Phones Fake a Microphone Array 19.11.2026 28:58
A mid-range Android phone ships with just two to four microphone capsules, yet its pickup pattern shifts dramatically between a call, a voice memo, and a video. Rather than true steerable beamforming, phones compare level differences and coherence between a primary bottom mic and a secondary reference mic. By looking at hardware teardowns, Android's AudioSource menu, and research from a Hannover g...
Context-Driven LoRA Mixing for Speech Recognition 19.11.2026 23:50
No published speech recognition system lets an AI agent dial LoRA adapters—like eighty percent tech jargon and twenty percent code-switching—from app context, and those percentages aren't real weights anyway. The discussion unpacks why NeMo's adapter API only switches or chains bottleneck adapters, why encoder LoRA hurt Parakeet-TDT in a dysarthric-speech study, and how confidence arbitration in D...
Teaching Hebrew Pronunciation to Many Synthetic Voices 19.11.2026 24:26
Recording an hour of Hebrew-English speech daily creates a rich long-tail corpus, but it entangles one speaker's voice with the language's pronunciation. Testing a proposed pipeline—converting speech to IPA and rendering it across synthetic voices—reveals why universal phone recognizers like ZIPA and POWSM fail by erasing stress and normalizing to dictionaries. Instead, pairing text-based G2P tool...
How Multiyear Contracts Are Locking In High RAM Prices 19.11.2026 26:18
DRAM revenue jumped 369% in a single year while actual memory shipped grew just 4%, raising the question of whether high RAM prices are permanent. A closer look at Samsung, SK hynix, and Micron shows how multiyear take-or-pay contracts, prepayments, and HBM production shifts are reshaping the market. From the Garciaguirre antitrust lawsuit to OpenAI's Stargate letters of intent and Nvidia down-spe...
Why 3D DRAM Beats Every RAM Replacement 19.11.2026 21:15
With AI demand driving a prolonged DRAM shortage, engineers have spent twenty years trying to replace working memory with persistent alternatives. From Intel Optane's $559 million write-off to niche shipping chips like STT-MRAM and ReRAM, nothing matches DRAM on speed, density, and cost at once. Even promising lab contenders—like Micron's ferroelectric NVDRAM, imec's capacitorless IGZO cells, and...
How HBM and Slow Fabs Squeeze the RAM Supply 19.11.2026 22:30
High-bandwidth memory for AI accelerators uses roughly three times the silicon wafer area of ordinary DDR5, quietly draining conventional RAM capacity even as manufacturers spend billions. Cleanroom shells take years to build, meaning new fabs from Micron, SK hynix, and Samsung won't meaningfully boost supply until 2028 or later. With three companies controlling around ninety percent of DRAM reven...
How AI Memory Demand Pushed Raspberry Pi Prices to $385 19.11.2026 23:19
A seven-fold surge in LPDDR4 memory costs has turned $40 hobbyist computers into luxury items, pushing the 16-gigabyte Raspberry Pi 5 to $385 and the Banana Pi BPI-M4 Zero up nearly 300 percent. The culprit isn't edge AI on small boards, but wafer capacity in fabs run by Samsung, SK hynix, and Micron. As memory makers exit legacy DDR4 to build high-bandwidth memory for AI data centres, single-boar...
How Architects Can Learn AI Tasks Instead of Models 19.11.2026 31:27
Model names churn every few weeks, but the underlying tasks on Hugging Face—like image-to-3D, depth estimation, mask generation, and zero-shot object detection—act as stable job descriptions. Mapping these tasks onto an architect's Revit and AutoCAD workflow reveals practical distinctions: why Gaussian splats work for client pitches while meshes start the actual model, why relative depth can't rep...
How the ICJ proved genocidal intent in Myanmar 19.11.2026 29:36
By a 9-6 vote, the International Court of Justice found Myanmar responsible for committing genocide against the Rohingya, marking the first time an international court has held a state liable under the 1948 Genocide Convention. This episode traces the decades-long stripping of Rohingya citizenship, the 2017 Tatmadaw clearance operations, and the legal test the court used to infer genocidal intent....
How Bluetooth Headsets Strip Away Vocal Identity 19.11.2026 20:14
Reusing speech recognition recordings for a voice clone seems efficient until a business headset's frequency ceiling gets baked into the model. While the human voice spans 100 Hz to 17 kHz, Bluetooth Hands-Free Profile headsets cut off audio at 4 to 7 kHz, stripping away the sibilance and "air" that define vocal identity. Comparing telecom standards like TIA-920 against corpora like VCTK and LJSpe...
Where Bizarre Language Textbook Sentences Came From 19.11.2026 21:05
Bizarre old language textbook lines like "La plume de ma tante" and "my postillion has been struck by lightning" were never coded messages, and their real origins are even stranger. "La plume de ma tante" is a French vowel drill, while the postillion line is an apocryphal fossil from horse-drawn travel guides. Meanwhile, the 1855 cult classic English As She Is Spoke—famously introduced by Mark Twa...
How Language Textbooks Flatten Speech and Emotion 19.11.2026 22:34
Language textbooks don't just sound stiff to beginners—corpus research shows they stay just as flat at advanced levels. Exploring Elen Le Foll's Textbook English Corpus, David Crystal's 1976 critique of textbookese, and Block and Gray's work on erased working-class voices, this conversation unpacks why coursebook dialogues feel like a sanitized middle-class performance. It also examines Jean-Marc...
Can You Fake Being a Beginner in Your Native Language? 19.11.2026 24:13
Pretending to know nothing in your own native language is harder than faking expertise because you have to suppress every automatic reflex. This episode examines real-world stunts—from Kei Simmons's year-long high school ruse to Jose Barrientos's four-month college act and a Passau exam-substitution case—alongside university heritage-speaker placement gates. It also explores Larry Selinker's inter...
Modern Hebrew's Semitic Roots and European Syntax 19.11.2026 25:47
Hebrew went 1,700 years without native speakers before gaining ten million in two generations, but the modern language is not simply Biblical Hebrew with a new accent. Examining linguist Ghil'ad Zuckermann's hybrid-language thesis, this breakdown explores how Modern Hebrew flipped Biblical word order from verb-first to subject-verb-object, converted aspect into tense, and absorbed Yiddish calques...
The Aramaic Spoken Today Isn't the Language of Jesus 19.11.2026 30:39
Most of the 575,000 to one million people speaking Neo-Aramaic today speak Eastern varieties like Suret and Turoyo, not the Western branch of antiquity. Only about a thousand people still speak Western Neo-Aramaic in the Syrian villages of Maaloula and Jubb'adin. This deep dive maps the surviving Aramaic family, from its historic homelands across Iraq, Iran, Turkey, and Syria to diaspora hubs in S...
Hebrew's Guttural Ḥet and ʿAyin Didn't Vanish Together 19.11.2026 25:02
Mainstream Israeli Hebrew didn't lose its guttural ḥet and ʿayin—the Ashkenazi founders of spoken Hebrew never had them to lose. Tracing the sociolinguistics from 1880s Yiddish-speaking revivalists to Roey Gafter's fieldwork in Tel Aviv and Rosh Ha'ayin, this episode unpacks how demography and ethnic prestige shaped modern pronunciation. Even as the Academy of the Hebrew Language prescribed the Or...
Liknande poddar
Replaio är ingen poddutgivare; programnamn, omslag och ljud tillhör sina upphovspersoner och distribueras via offentliga RSS-flöden