algorithm

36 Articles

Can Digital Poison Corrupt The Algorithm?

June 27, 2025 by 24 Comments

These days, so much of what we see online is delivered by social media algorithms. The operations of these algorithms are opaque to us; commentators forever speculate as to whether they just show us what they think we want to see, or whether they try to guide our thinking and habits in a given direction. The Digital Poison device  from [Lucretia], [Auxence] and [Ramon] aims to twist and bend the algorithm to other ends.

The concept is simple enough. The device consists of a Raspberry Pi 5 operating on a Wi-Fi network. The Pi is set up with scripts to endlessly play one or more select YouTube videos on a loop. The videos aren’t to be watched by anyone; the device merely streams them to rack up play counts and send data to YouTube’s recommendation algorithm. The idea is that as the device plays certain videos, it will skew what YouTube recommends to users sharing the same WiFi network based on perceived viewer behavior.

To achieve subtle influence, the device is built inside an unobtrusive container. The idea being that it could be quietly connected to a given WiFi network to stream endlessly, in turn subtly influencing the view habits of other users on the same network.

It’s difficult to say how well this concept would work in practice. In many cases, sites like YouTube have robust user tracking that feeds into recommendation algorithms. Activity from a random user signed into the same network might not have much of an influence. However, conceptually, it’s quite interesting, and the developers have investigated ways to log the devices operation and compare it to recommendations fed to users on the network. Privacy provisions make this difficult, but it may be possible to pursue further research in this area. Files are on Github for the curious.

Ultimately, algorithms will always be a controversial thing as long as the public can’t see how they work or what they do. If you’re working on any projects of your own in this space, don’t hesitate to let us know!

[Thanks to Asher for the tip!]

Fitting A Spell Checker Into 64 KB

March 26, 2025 by 16 Comments

By some estimates, the English language contains over a million unique words. This is perhaps overly generous, but even conservative estimates generally put the number at over a hundred thousand. Regardless of where the exact number falls between those two extremes, it’s certainly many more words than could fit in the 64 kB of memory allocated to the spell checking program on some of the first Unix machines. This article by [Abhinav Upadhyay] takes a deep dive on how the early Unix engineers accomplished the feat despite the extreme limitations of the computers they were working with.

Perhaps the most obvious way to build a spell checker is by simply looking up each word in a dictionary. With modern hardware this wouldn’t be too hard, but disks in the ’70s were extremely slow and expensive. To move the dictionary into memory it was first whittled down to around 25,000 words by various methods, including using an algorithm to remove all affixes, and then using a Bloom filter to perform the lookups. The team found that this wasn’t a big enough dictionary size, and had to change strategies to expand the number of words the spell checker could check. Hash compression was used at first, followed by hash differences and then a special compression method which achieved an almost theoretically perfect compression.

Although most computers that run spell checkers today have much more memory as well as disks which are orders of magnitude larger and faster, a lot of the innovation made by this early Unix team is still relevant for showing how various compression algorithms can be used on data in general. Large language models, for one example, are proving to be the new frontier for text-based data compression.

Math On A Checkerboard

January 3, 2025 by 6 Comments

The word “algorithm” can sometimes seem like a word designed to scare people away from math classes, much like the words “calculus”, “Fourier transform”, or “engineering exam”. But in reality it’s just a method for solving a specific problem, and we use them all the time whether or not we realize it. Taking a deep dive into some of the ways we solve problems, especially math problems, often leads to some surprising consequences as well like this set of algorithms for performing various calculations using nothing but a checkerboard.

This is actually a demonstration of a method called location arithmetic first described by [John Napier] in 1617. It breaks numbers into their binary equivalent and then uses those representations to perform multiplication, division, or to take the square root. Each operation is performed by sliding markers around the board to form certain shapes as required by the algorithms; with the shapes created the result can be viewed directly. This method solves a number of problems with other methods of performing math by hand, eliminating other methods like trial-and-error. The video’s creator [Wrath of Math] demonstrates all of these capabilities and the proper method of performing the algorithms in the video linked below as well.

While not a “hack” in the traditional sense, it’s important to be aware of algorithms like this as they can inform a lot of the way the world works on a fundamental level. Taking that knowledge into another arena like computer programming can often yield some interesting results. One famous example is the magic number found in the code for the video game Quake, but we’ve also seen algorithms like this used to create art as well.

Continue reading “Math On A Checkerboard” →

Custom Drone Software Searches, Rescues

October 24, 2024 by 17 Comments

When a new technology first arrives in people’s hands, it often takes a bit of time before the full capabilities of that technology are realized. In much the same way that many early Internet users simply used it to replace snail mail, or early smartphones were used as more convenient methods for messaging and calling than their flip-phone cousins, autonomous drones also took a little bit of time before their capabilities became fully realized. While some initially used them as a drop-in replacement for things like aerial photography, a group of mountain rescue volunteers in the United Kingdom realized that they could be put to work in more efficient ways suited to their unique abilities and have been behind a bit of a revolution in the search-and-rescue community.

The first search-and-rescue groups using drones to help in their efforts generally used them to search in the same way a helicopter would have been used in the past, only with less expense. But the effort involved is still the same; a human still needed to do the searching themselves. The group in the UK devised an improved system to take the human effort out of the equation by sending a drone to fly autonomously over piece of mountainous terrain and take images of the ground in such a way that any one thing would be present in many individual images. From there, the drone would fly back to its base station where an operator could download the images and run them through a computer program which would analyse the images and look for outliers in the colors of the individual pixels. Generally, humans tend to stand out against their backgrounds in ways that computers are good at spotting while humans themselves might not notice at all, and in the group’s first efforts to locate a missing person they were able to locate them almost immediately using this technology.

Although the system is built on a mapping system somewhat unique to the UK, the group has not attempted to commercialize the system. MR Maps, the software underpinning this new feature, has been free to use for anyone who wants to use it. And for those just starting out in this field, it’s also worth pointing out that location services offered by modern technologies in rugged terrain like this can often be misleading, and won’t be as straightforward of a solution to the problem as one might think.

Supercon 2023: Teaching Robots How To Learn

September 3, 2024 by 2 Comments

Once upon a time, machine learning was an arcane field, the preserve of a precious few researchers holed up in grand academic institutions. Progress was slow, and hard won. Today, however, just about anyone with a computer can dive into these topics and develop their own machine learning systems.

Shawn Hymel has been doing just that, in his work in developer relations and as a broader electronics educator. His current interest is reinforcement learning on a tiny scale. He came down to the 2023 Hackaday Supercon to tell us all about his work.

Continue reading “Supercon 2023: Teaching Robots How To Learn” →

Robots Collaborate To Localize Themselves Precisely

August 26, 2024 by 3 Comments

Here’s the thing about robots. It’s hard for them to figure out where to go or what they should be doing if they don’t know where they are. Giving them some method of localization is key to their usefulness in almost any task you can imagine. To that end, [Guy Elmakis], [Matan Coronel] and [David Zarrouk] have been working on methods for pairs of robots to help each other in this regard.

As per the research paper, the idea is to perform real-time 3D localization between two robots in a given location. The basic idea is that the robots take turns moving. While one robot moves, the other effectively acts as a landmark. The robots are equipped with inertial measurement units and cameras in a turret, which they use to track each other and their own movements. Each robot is equipped with a Raspberry Pi 4 for processing image data and computing positions, and the two robots communicate via Bluetooth to coordinate their efforts.

It’s an interesting technique that could have some real applications in swarm robotics, and in operations in areas where satellite navigation and other typical localization techniques are not practical. If you’re looking for more information, you can find the paper here. We’ve seen some other neat localization techniques for small robots before, too. Video after the break.

Continue reading “Robots Collaborate To Localize Themselves Precisely” →

50-Year-Old Program Gets Speed Boost

November 7, 2023 by 19 Comments

At first glance, getting a computer program to run faster than the first electronic computers might seem trivial. After all, most of us carry enormously powerful processors in our pockets every day as if that’s normal. But [Mark] isn’t trying to beat computers like the ENIAC with a mobile ARM processor or other modern device. He’s now programming with the successor to the original Intel integrated circuit processor, the 4040, but beating the ENIAC is still little more complicated than you might think with a processor from 1974.

For this project, the goal was to best the 70-hour time set by ENIAC for computing the first 2035 digits of pi. There are a number of algorithms for performing this calculation, but using a 4-bit processor and an extremely limited memory of only 1280 bytes makes a number of these methods impossible, especially with the self-imposed time limit. The limited instruction set is a potential bottleneck as well with these early processors. [Mark] decided to use [Fabrice Bellard]’s algorithm given these limitations. He goes into great detail about the mathematics behind this method before coding it in JavaScript. Generating assembly language from a working JavaScript was found to be fairly straightforward.

[Mark] is also doing a lot of work on the 4040 to get this program running as well, including upgrades to the 40xx tool stack, the compiler and linker, and an emulator he’s using to test his program before sending it to physical hardware. The project is remarkably well-documented, including all of the optimizations needed to get these antique processors running fast enough to beat the ENIAC. We won’t spoil the results for you, but as a hint to how it worked out, he started this project using the 4040 since his original attempt using a 4004 wasn’t quite fast enough.