Engineering Pragmatism In Hardware, AI Nuances, and Material Failure Risks - Episode Hero Image

Engineering Pragmatism In Hardware, AI Nuances, and Material Failure Risks

Original Title:

TL;DR

  • Using PLA for 3D printed airplane parts is dangerous due to its low glass transition temperature (52-54°C), causing material failure under heat and air intake pressure, as demonstrated by a recent aircraft crash.
  • Integrating RTL-SDR dongles with Home Assistant via the RTL-HAOS plugin enables a unified home automation system to decode signals from a vast array of sub-gigahertz ISM band devices.
  • Typewriters with integrated pen plotters offered advanced graphical capabilities for their time, allowing for custom fonts, variable sizing, and chart generation without a computer, a forgotten technology.
  • Reverse sundials that track the sun using motors and an illuminated arrow, like the one built with a bent brass tube and spring steel wire, demonstrate clever mechanical design for timekeeping.
  • Trace Line Clocks utilize magnetic attraction for a frictionless, rotating hand mechanism, where a flexible linkage connects hour and minute indicators, showcasing novel applications of magnets.
  • Volumetric displays, like the one using a projector to illuminate bubbles within a glass cube, achieve 3D patterns by precisely mapping pixels to bubbles at varying depths.
  • Self-cleaning road cat's eyes, relying on rubber domes and water accumulation from rain, offer a passive, low-maintenance solution for road marking visibility, proving simple mechanical designs can be effective.
  • Large Language Models (LLMs) are sophisticated pattern-matching systems trained on vast datasets, but they lack true learning, reasoning, and creativity, distinguishing them from genuine artificial intelligence.

Deep Dive

The Hackaday Podcast episode "Ep 349: Clocks, AI, and a New 3D Printer Guy" highlights how practical engineering considerations, particularly material properties and project focus, are critical for successful hardware projects. It also delves into the nuances of artificial intelligence, contrasting current large language models with true learning and reasoning capabilities, and emphasizes that sophisticated mechanical design can often offer elegant solutions where complex electronics might be unnecessary.

The discussion on 3D printing material failures, specifically a PLA duct melting on an experimental airplane, underscores the critical importance of understanding material properties for safety-critical applications. The incident, where the part was purchased rather than printed by the user, reveals a systemic risk: misrepresentation or ignorance of material specifications can lead to catastrophic failures. This extends beyond aviation, as the hosts note the potential for 3D-printed components in cars and motorcycles to fail under heat. This suggests a need for greater accountability in the supply chain of 3D-printed parts and a more rigorous approach to material verification, especially when components are intended for high-stress environments.

Furthermore, the episode explores the concept of "productive hacker maturity," illustrated by Dan Malone's decision to purchase a reliable, turnkey 3D printer to complete a project rather than investing time in fixing or building a less reliable one. This contrasts with the "hacker not invented here" syndrome, where the process of building or fixing becomes the primary goal, potentially derailing the actual project objective. The implication is that experienced makers recognize when to leverage commercially available, functional tools to achieve a specific outcome, reserving complex DIY efforts for projects where the creation process itself is the objective. This pragmatic approach saves time and resources, especially when deadlines are involved, and prevents the common pitfall of a project becoming a secondary, more complex project in itself.

In the realm of artificial intelligence, the podcast critically examines the current hype around Large Language Models (LLMs). Maya Posch's argument, comparing LLMs to "worst summer interns ever," is expanded upon by distinguishing between knowledge distillation and true learning. While LLMs can access and process vast amounts of information to generate plausible answers, they do not learn from their environment or interactions in real-time, a characteristic considered fundamental to human and even animal intelligence. The hosts draw parallels with observed animal intelligence in crows and bears, suggesting that intelligence is multifaceted and not solely defined by data processing capability. This perspective challenges the notion that LLMs represent true artificial intelligence, framing them instead as sophisticated pattern-matching systems. The core implication is that current LLMs lack the dynamic learning, reasoning, and creative capabilities that define biological intelligence, and that future advancements will likely focus on systems that can learn and adapt over time, rather than fixed models trained on static datasets.

Action Items

  • Audit 3D printing material usage: Identify 5 critical components where PLA is used and evaluate for potential replacement with ABS or other higher-temperature resistant materials to prevent in-flight failures.
  • Create a material property testing checklist: For 10 common 3D printing filaments, document glass transition temperatures and maximum operating temperatures to ensure appropriate material selection for functional parts.
  • Implement a component sourcing verification process: For 5 purchased 3D printed parts, establish a protocol to confirm material composition against advertised specifications to prevent misrepresentation and component failure.
  • Design a runbook for 3D printer maintenance: Document 3 common failure modes (e.g., clogged nozzle, faulty sensor) and their resolution steps to enable efficient troubleshooting and reduce project delays.

Key Quotes

"basically it looks like the part should have been printed in abs was printed in pla and it melted pla is generally 55 to 65 i just looked it up so actually it's worse than pla according to test results here of the measured glass transition temperature for the sample was between 52 8 degrees c and 54 degrees c yeah and that's less than pla apparently so you know how it is pla gets hot it gets a little bit plastic right it gets a little bit flexible and the hotter it gets the more flexible it gets so i think there were a lot of comments on this you know what is the glass transition temperature you know when is soft enough soft enough and the answer for your airplane is when it can't hold up to the air intake pressure and collapses on itself but in other applications don't get your pla hot"

The authors, Elliot and Al Williams, highlight a critical failure in a 3D-printed airplane part. They explain that the part, intended to be made from ABS, was instead printed with PLA, a material with a lower glass transition temperature. This led to the part melting and collapsing under flight conditions, causing the engine to lose air intake and the subsequent crash. The Williamses emphasize the importance of understanding material properties for functional applications, warning against using PLA in high-temperature environments.


"if you are 3d printing things and you want it to be functional you need to know about the material properties of whatever it is you're 3d printing with yeah a lot of things in cars and motorcycles get hot too so i wonder how many people are you know driving around with something 3d printed that's just a matter of time"

Elliot and Al Williams underscore the necessity of understanding material science when 3D printing functional components. They extend this concern beyond aviation to automotive and motorcycle applications, suggesting that many people might be unknowingly using 3D-printed parts in heat-sensitive areas. The Williamses imply that this oversight could lead to future failures, as the material properties may not be suitable for the operating environment.


"rtl ha os that is home automation os sits in the middle of this rtl 433 library which does the sdr stuff and home assistant home automation library and makes them work together and so what this means is that if you've got a home automation system that is fairly smart but doesn't have enough sensors in it you throw this in and all of a sudden it can listen to anything radio and try to bring that into your system"

The authors describe the RTL HA OS project, a piece of software that bridges the RTL_433 library, which decodes signals from software-defined radios, with the Home Assistant home automation platform. Jaron McDaniel, the creator, developed this to allow home automation systems to integrate data from a wide range of wireless sensors that communicate on ISM bands. This project enables users to expand their smart home capabilities by listening to various radio frequencies and incorporating the data into their existing systems.


"the beauty about doing it with an sdr is that on one hand it's totally overkill but on the other hand this one device can listen to all of these different frequencies and no matter what kind of crazy device you've got on there using whatever kind of crazy encoding probably pretty easy to figure that out with an sdr"

Elliot and Al Williams discuss the versatility of using a Software Defined Radio (SDR) for decoding wireless signals. They acknowledge that it might seem like overkill, but an SDR's ability to tune into multiple frequencies and decode various encodings makes it a powerful tool. The Williamses suggest that an SDR can likely decipher signals from a wide array of devices, regardless of their specific communication protocols.


"the trace line clock you know hackaday we have a long history of showing people's clocks simply because well it's a pretty easy thing to build usually as a clock and then the real question is is points for style right everybody can show a clock on the screen but what are the other ways you can do it this one is just very cool what i like best about it is it uses the magic of magnets and i think we just don't do that enough right magnets are as far as i can tell magnets are pure alien magic and uh they this one actually rotates a hand using magnets and a little blue line just rotates around and there's a way to tell what time it is from looking at how the pieces line up against the circle"

The authors introduce the "Trace Line Clock," a hackaday featured clock that stands out due to its use of magnets for its display mechanism. They express a fondness for magnetic applications, describing magnets as "pure alien magic." The clock features a rotating hand, a blue line, that indicates the time by aligning with markings on a circular display, highlighting a stylish and unconventional approach to timekeeping.


"i think the key thing that none of this measures is creativity so you know i've been on that jag for a while i did a talk at a university i think about a year ago where i talked about there's generative ai which everybody likes to talk about mainly because most people are too lazy to create stuff but then there's assistive ai you don't hear much about that right so you think about it like a spell checker is assistive not that it's ai but it's assistive you don't go i'm going to write a novel so let me ask the spell checker how to start it out you just use it to check your spelling that's assisting you"

Maya Posch distinguishes between generative AI, which she notes is popular because people may be "too lazy to create stuff," and assistive AI. Posch uses a spell checker as an example of assistive AI, explaining that it aids in tasks like writing a novel by checking spelling rather than generating the content itself. She implies that generative AI, while capable of producing output, may lack the creative spark or genuine understanding that characterizes human intelligence.

Resources

External Resources

Books

  • "The British Air Accidents Investigation Branch Report" - Mentioned as a source for information on an airplane crash involving a 3D printed part.

Articles & Papers

  • "What is Chat GPT Doing and Why Does It Work" by Stephen Wolfram - Referenced for a detailed explanation of how ChatGPT functions.

People

  • Al Williams - Co-host of the Hackaday Podcast, discussing various hacks and tech topics.
  • Dan Maloney - Mentioned for his project involving a 3D printer and a generator, and for his article on focusing on projects.
  • Elliot Williams - Co-host of the Hackaday Podcast, discussing various hacks and tech topics.
  • Harry Waguna - Submitted a tip about a volumetric display hack.
  • Jaron McDaniel - Creator of the RTL 433 to Home Assistant bridge.
  • Maya Posch - Mentioned for her take on LLM intelligence.
  • Mike Ferney - Creator of a self-cleaning road cat's eye hack.
  • Tansurat Akinchi - Creator of an RP2350 module for Framework laptops.
  • Tyler - Mentioned as having written up the reverse sundial hack.
  • Vincent Wins - Creator of a production KiCad template.

Organizations & Institutions

  • Alps - Manufacturer of mechanisms used in pin plotter typewriters.
  • Hackaday - Platform for sharing hacks and tech projects, and the source of the podcast.
  • Hackaday.com - Website where listeners can find links and participate in discussions.
  • Home Assistant - Home automation software used in a hack.
  • New York Times - Source of daily computer games, including "Connections."
  • Wolfram - Company offering supercomputing services.

Tools & Software

  • KiCad - PCB design software.
  • Kibot - Program used for generating documentation for PCB projects.
  • MQTT - Protocol used for data transfer in home automation.
  • RTL SDR dongle - Software-defined radio used for decoding signals.
  • RTL 433 library - Software for decoding signals from various devices.
  • Wolfram Alpha - Service mentioned for running computations.

Websites & Online Resources

  • Hackaday.com/podcast - Webpage for the Hackaday Podcast, including forms for submitting guesses.

Other Resources

  • 3D printed duct - Part that failed in an airplane due to incorrect material choice (PLA instead of ABS).
  • 3D printing - Technology discussed in relation to various hacks and projects.
  • AI (Artificial Intelligence) - Concept discussed in relation to LLMs and animal cognition.
  • LLM (Large Language Model) - Type of AI discussed in terms of intelligence and capabilities.
  • Pla - Material used for 3D printing, which melted in an airplane duct.
  • Abs - Material that should have been used for the airplane duct.
  • Glass transition temperature - Property of plastics relevant to the airplane duct failure.
  • Rotary engine - Type of engine discussed in the "What's That Sound" segment.
  • ISM bands - Radio frequency bands used for sub-gigahertz communication.
  • Software Defined Radio (SDR) - Technology used for decoding radio signals.
  • Pin plotter typewriters - Devices from the 1980s that used pens to draw characters.
  • Alps mechanism - Component used in pin plotter typewriters.
  • Daisy wheel printers - Type of printer, mentioned in relation to Juki.
  • Juki 6100 - Model of a daisy wheel printer.
  • Diablo 630 - Printer model that Juki 6100 was a knockoff of.
  • Reverse sundial - Clock hack that uses an illuminated LED arrow pointing at the sun.
  • Equatorial mount - Telescope mount concept compared to the reverse sundial.
  • Sunstone - Legendary Viking navigation tool.
  • Trace line clock - Clock hack utilizing magnets and a rotating hand.
  • Volumetric display - 3D display technology using lasers and bubbles in glass.
  • Projection mapping - Technique compared to volumetric display.
  • Wiremap - An earlier volumetric projection concept from 2008.
  • Road cat's eye - Reflective road markers that are self-cleaning.
  • KiCad template - Production template for KiCad projects.
  • RP2350 - Microcontroller module for Framework laptops.
  • Framework laptop - Laptop with expansion ports for modules like the RP2350.
  • Voronoi style interactive LED wall panel art - Interactive art pieces with LED panels.
  • Bluetooth chip - Debuggable Bluetooth solution.
  • CH microprocessors - Microprocessors used in a debuggable Bluetooth chip.
  • LED hourglass - Hourglass using LEDs instead of sand.
  • Supercomputing - High-performance computing discussed in relation to Wolfram's services.
  • Generative AI - AI that creates new content.
  • Assistive AI - AI that helps users with tasks.
  • Spell checker - Example of assistive AI.
  • Crows - Animals mentioned for their intelligence and problem-solving abilities.
  • Primates - Animals discussed in the context of early language learning experiments.
  • Dogs - Animals discussed in relation to soundboard communication.
  • Reinforcement learning - AI paradigm focused on learning through experience.
  • Diffusion image generator models - AI models that create images from noise.
  • Markov chain - Statistical model for predicting sequences.

---
Handpicked links, AI-assisted summaries. Human judgment, machine efficiency.
This content is a personally curated review and synopsis derived from the original podcast episode.