Human-Written Or Machine-Generated: Finding Intelligence In Language Models

January 24, 2024 by 45 Comments

What is the essential element which separates a text written by a human being from a text which has been generated by an algorithm, when said algorithm uses a massive database of human-written texts as its input? This would seem to be the fundamental struggle which society currently deals with, as the prospect of a future looms in which students can have essays auto-generated from large language models (LLMs) and authors can churn out books by the dozen without doing more than asking said algorithm to write it for them, using nothing more than a query containing the desired contents as the human inputs.

Due to the immense amount of human-generated text in such an LLM, in its output there’s a definite overlap between machine-generated text and the average prose by a human author. Statistical methods of detecting the former are also increasingly hamstrung by the human developers and other human workers behind these text-generating algorithms, creating just enough human-like randomness in the algorithm’s predictive vocabulary to convince the casual reader that it was written by a fellow human.

Perhaps the best way to detect machine-generated text may just be found in that one quality that these algorithms are often advertised with, yet which they in reality are completely devoid of: intelligence.

Continue reading “Human-Written Or Machine-Generated: Finding Intelligence In Language Models” →

This Unique Flip-Flop Uses Chemistry And Lasers

January 24, 2024 by 8 Comments

One of the first logic circuits most of us learn about is the humble flip-flop. They’re easy enough to build with just a couple of NOR or NAND gates, and even building one up from discrete components isn’t too much of a chore. But building a flip-flop from chemicals and lasers is another thing entirely.

That’s the path [Markus Bindhammer] took for his photochromic molecular switch. We suspect this is less of an attempt at a practical optical logic component and more of a demonstration project, but either way, it’s pretty cool. Photochromism is the property by which molecules reversibly rearrange themselves and change color upon exposure to light, the most common example being glass that darkens automatically in the sun. This principle can be used to create an optical flip-flop, which [Markus] refers to as an “RS” type but we’re pretty sure he means “SR.”

The electronics for this are pretty simple, with two laser modules and their drivers, a power supply, and an Arduino to run everything. The optics are straightforward as well — a beam splitter that directs the beams from each laser onto the target, which is a glass cuvette filled with a clear epoxy resin mixed with a photochromic chemical. [Markus] chose spiropyran as the pigment, which when bathed in UV light undergoes an intramolecular carbon-oxygen bond breakage that turns it into the dark blue pigment merocyanine. Hitting the spot with a red laser or heating the cuvette causes the C-O bond to reform, fading the blue spot.

The video below shows the intensely blue dot spot developing under UV light and rapidly fading thanks to just the ambient temperature. To make the effect last longer, [Markus] cools the target with a spritz from a CO2 cartridge. We imagine other photochromic chemicals could also be employed here, as could some kind of photometric sensor to read the current state of the flip-flop. Even as it is, though, this is an interesting way to put chemistry and optics to work.

Continue reading “This Unique Flip-Flop Uses Chemistry And Lasers” →

Massive Water Rocket Is Impressive But Accessible

January 24, 2024 by 12 Comments

Water rockets are one of those projects that never get old, and bumping the size just adds to the challenge. In the video after the break, [ARRO Rockets] takes us through the launch of Gamma IV, his most ambitious water rocket project yet. Crafted with spliced soda bottles and standard household materials, this rocket is a testament to what one can achieve with simple components and a bit of ingenuity.

The rocket’s release mechanism demonstrates this — employing nothing more than a quick connect hose connection and a basic pulley system. The parachute recovery system is also a nice combo of modern electronics and simplicity. It uses a microcontroller with accelerometer to detect the apogee, and release the parachute to be ejected by another piece of soda bottle acting as a spring. It also records or the flight data on an SD card.

[ARRO Rockets] had some trouble with friction on the launch rail, which was partially solved with liberal application of silicone spray. The root cause might be the rail button flexing on launch, or just the change of the pressurized bottles.

We are especially impressed by how accessible this project is, a reminder that high-flying achievements don’t necessarily require deep pockets or hard-to-source parts. The entire setup is not only cost-effective but also opens up numerous possibilities for further experimentation and refinement, like adding a second stage or a precision release mechanism.

Continue reading “Massive Water Rocket Is Impressive But Accessible” →

Crusty: The Story Of The Mac SE That Could

January 23, 2024 by 25 Comments

Retrocomputing often involves careful restorations, rare components, and white gloves.  This story involves none of those. This is the story of two people who sought to answer one of the greatest questions in the universe: What does it take to kill a Mac SE?

Crusty’s mainboard as found

The star of the show here is Crusty, a Mac SE that was found on the loading dock of a scrap company. It sat out in the weather for at least 6 months, complete with the original leaking lithium battery.

Enter [RadRacer203], who is friends with the owner of this particular scrap company. [RadRacer203] and picked up Crusty, along with a few other classic Macs. He brought these machines to VCF East 2021, where our other hero comes in. [CJ] is something of a magician with CRTs and analog electronics. Trained under [Sark] himself, [CJ] has mastered the 5-finger exploding capacitor technique.

The battery had eaten through the mainboard and even into the chassis. But after a thorough cleaning, the damn thing booted up. Crusty was born.

This Mac was a survivor. Much like Top Gear and their plucky Toyota Hilux, [RadRacer203] and [CJ] devised a plan to put Crusty to the test.

Click through the break for more!

Continue reading “Crusty: The Story Of The Mac SE That Could” →

Cockpit of a Hawker Siddeley Trident with the moving map display

A Live Map Display In A 1960s Airliner

January 23, 2024 by 4 Comments

We tend take GPS navigation for granted these days, so it’s easy to forget that it became only available in the last few decades. Aviation navigation used to be significantly more challenging, so how was the Hawker Siddeley Trident, a 1960s airliner, fitted with a live updating map display? In a fascinating dive into aviation history the British Airliner Collection has spun up an insightful article on the magic behind these moving map displays.

Without access to satellite navigation or advanced electronics, engineers had to get creative. Enter the Trident’s moving map display, a marvel of ingenuity that predated the GPS systems. Using a combination of Doppler radar and some clever mechanics, pilots could accurately determine their position without relying on any external signals.

The system makes use of four Doppler radar beams, arranged in what was known as the Janus array. This configuration corrected for errors caused by changes in altitude or wind drift, ensuring accurate ground speed readings. The movable antennas mounted under the cabin floor could adjust its orientation to maintain alignment with the actual direction of travel, calculating drift angle precisely. Combined with compass information and flight time from a known start point to to indicate the current position with a pointer on a rolled paper map. The system was well ahead of it’s time, and significantly easier to use and more accurate than the Decca radio navigation system in use at the time.

It’s mind boggling to see the solutions engineers came up with without much of the digital technology we take for granted today. Gyroscopes for inertial navigation, the cavity magnetron for radar and radial engines were all building blocks for modern aviation.

Thanks for the tip [poiuyt]!

Reverse Engineering The Apple Touch Bar Screen

January 23, 2024 by 10 Comments

The Apple Touch Bar was an oddity on a fairly small number of Apple laptops which replaced the function key row with a touch display. Yet what is special about this display other than its odd form factor when you consider it as a generic touch display? As [Wenting Zhang] describes in a recent reverse-engineering video, this 2,170 x 60 pixel display is somewhat limited in that it doesn’t support the MIPI DSI video mode, only command mode, along with a special instruction (0x3C) for automatic address offsets. The results of this project can be found on the GitLab account.

In a way these limitations make sense when you consider Apple’s use case for these special MIPI-DSI displays. As a touch screen with dynamic controls being displayed on it, features such as video playback never were a goal, and thus Apple likely decided to save a few bucks, possibly also due to MIPI licensing costs. What this means is that if you had dreamed of snapping up an extremely long and narrow OLED display for a video project you’re in for somewhat of a bad time. Although animated content is possible – as [Wenting] demonstrates – this comes with all the limitations of command mode, meaning slower updates, higher power usage and a lot more overhead.

Continue reading “Reverse Engineering The Apple Touch Bar Screen” →

Autofeeding CNC Lathe Cranks Out Parts All By Itself

January 23, 2024 by 18 Comments

The trouble with building a business around selling low-margin widgets is that you have to find a way to make a lot of them to make it worth your while. And if the widget in question is labor-intensive to make, you’ve got to find ways to reduce your inputs. That sounds like a job for industrial automation, a solution that’s often out of reach of small shops, for all the obvious reasons. Not if you’re clever about things, though, as this fully automated CNC lathe work cell shows.

This build comes to us from the woodshop of [Maher Lagha], where he’s making wooden honey dippers. Wooden dowel blanks are dispensed from an infeed rack and chucked between centers on the headstock and pneumatic tailstock. A two-axis stage in front of the workpiece moves a tool against the spinning stock, carving out the honey dipper in just a few minutes. When the lathe work is done, the spindle stops, the tailstock pulls the honey dipper back off the headstock, and a pneumatic piston unceremoniously whacks the almost-finished part — it looks like it still needs a little manual post-processing — into a bin. Lather, rinse, repeat, profit.

[Maher] doesn’t provide many details, but just looking at the work cell shows a veritable feast of industrial automation equipment. The spindle and tailstock of the lathe sit on a bed made from a massive slab of aluminum extrusion, and the X- and Y-axes use linear rails and ballscrews. And mindful of the effects of wood chips on delicate mechanisms, [Maher] did a good job of containing the mess with a host of acrylic guards.

As we said when we saw [Maher]’s wooden coaster work cell a while back, the wood widget business must be pretty good to justify automation like this. What’s nice with both these rigs is that they look like they could be quickly reprogrammed and retooled to create other products. Pretty impressive.

Continue reading “Autofeeding CNC Lathe Cranks Out Parts All By Itself” →