All Journal Club members can pull one episode from the archive per day. If you find something on this list that interests you, email us to have it pulled and sent to you. To get the PDF of any paper, click on its image below.
Release Date: November 20th, 2024
Before we talk about today’s paper, we need to pause, jump in our time machines, and go back to talk about a paper from 26 years ago. In 1998, Yann Lecun, a professor at NYU, published a seminal paper on computer vision: Gradient-Based Learning Applied to Document Recognition. In that paper, the authors outlined the application of convolutional neural networks (CNNs) for recognizing handwritten characters. But more importantly for our purposes today, that paper was also the vehicle through which they first published the MNIST dataset. If you’ve worked...
Watch the Trailer
Release Date: November 19th, 2024
If you’re working as a network engineer for a high traffic application, like a website or an API your mandate is likely pretty broad. You need to not only accommodate, serve, log and analyze today’s traffic, but you need to preemptively build the systems that will allow you to continue doing those things tomorrow. If the site scales up, if the traffic peaks, if there’s a flash-sale or a news article that changes the traffic patterns you need to be on top of it…before it happens. And you need to do all of this without a crystal ball. How?...
Request this Episode
Release Date: November 18th, 2024
The system employs a one-stage object detection model, YOLOx, as its foundation. Unlike two-stage detectors that rely on region proposal networks (RPNs) to generate candidate object regions—an inherently computationally intensive process—YOLOx bypasses this stage by directly predicting bounding boxes and classifications in a single pass through the network. This design allows for faster inference speeds, a critical feature for real-time applications. The YOLOx model serves as a lightweight yet robust baseline, leveraging CSPDarkNet as its backbone for efficient feature extraction. This backbone strikes a balance between computation and feature representation, ensuring sufficient granularity to detect objects under varying conditions of motion and occlusion....
Watch the Trailer
Release Date: November 17th, 2024
They started with YOLOv8, a single-stage object detection model designed for efficient and fast object recognition. Its architecture consists of three primary components: The backbone, the neck, and the detection head. The backbone serves as the feature extractor, processing raw image data to produce rich feature maps that emphasize important aspects of the input, such as edges, shapes, and textures. The neck is responsible for feature fusion, taking information from different levels of the backbone to combine spatial and semantic details into more refined representations. Lastly, the detection head predicts bounding boxes and class labels based on these fused features, outputting the final detection results....
Watch the Trailer
Release Date: November 16th, 2024
China is producing nearly 40 million tons of apples a year, and there’s simply not enough people to pick them. So, Chinese producers are increasingly turning to apple-picking robots to do the job. These can take several different forms, but the one we’re looking at today is a single arm on a platform with an effector attached that can grab apples, twist them off the branch, and plunk them into a basket....
Watch the Trailer
Release Date: November 15th, 2024
This paper is about two very different, but related things. On one hand, yes, it is about using Machine Learning to predict or diagnose the existence of Polycystic Ovary Syndrome. But on the other hand, it’s really about explainable A.I. You see, researchers have been developing ML models to predict or flag PCOS in patients for years. It’s such a popular area of research that there are openly available Kaggle datasets on the subject that allow anyone to easily train a model to do that. It’s been done, and it’s being done. That’s really not the point here....
Watch the Trailer
Release Date: November 14th, 2024
In 1916 Albert Einstein predicted the existence of Gravitational Waves: disturbances or ripples in spacetime. Transient displacements in a gravitational field. Gravitational Waves (GWs), would be caused, he said, by massive, accelerated objects like colliding black holes or neutron stars. But, he also predicted that GWs would be extremely difficult to measure, requiring incredibly sensitive instruments....
Watch the Trailer
Release Date: November 13th, 2024
In 1998, NASA launched the Mars Climate Orbiter. The spacecraft was designed to orbit Mars and relay vital atmospheric data back to Earth. For months, the orbiter traveled through space, with mission-control monitoring its progress and making minor adjustments as needed. Anticipation built as the team prepared for the critical moment when the orbiter would enter Mars' orbit. On the scheduled day, engineers gathered together, awaiting confirmation that the spacecraft had successfully positioned itself around Mars. Instead, they were met with an unsettling silence. Attempts to contact the orbiter failed, and it became clear that the Mars Climate Orbiter was lost, its mission ending abruptly as it vanished into the Martian atmosphere....
Watch the Trailer
Release Date: November 12th, 2024
Caching is, in a word, complex. And in today's paper, it gets significantly more complex. Some would say it becomes ridiculously complex, but I’ll leave that for you to decide. This paper takes the traditional concept of a "cache-hit" or "cache-miss" and adds a third option: a "similarity-hit". A similarity-hit means that the system found cached content that is (in some way) close to what the user is requesting, but not exactly it. Instead, it’s something similar-enough that it might still satisfy the user, and it can be retrieved from the cache rather than the original source....
Watch the Trailer
Release Date: November 11th, 2024
Today, November 11th 2024, marks the first day of COP29, the 29th meeting of the Conference of the Parties. COP was created by the UNFCCC, the United Nations Framework Convention on Climate Change, to bring countries together to negotiate and set international climate action goals. COP29 is happening in Baku, Azerbaijan, and will last nearly two weeks. During this time, member countries will set their climate change policies for...
Watch the Trailer
Release Date: November 10th, 2024
In the early days of the 20th century, AT&T had a problem. They wanted to build the first transcontinental phone line, connecting San Francisco and New York City. But, their engineers told them it was practically impossible. Voices in a phone line couldn’t travel anywhere near that far. Why? Signal attenuation. As signals flowed through the phone lines over distance, they got weaker and weaker. Even if they had the physical wires connecting the two...
Request this Episode
Release Date: November 9th, 2024
Imagine that you're a bank, and you've got a fraud problem. People are signing up for your credit-cards with stolen identities, racking up charges, then disappearing. And it's costing you a fortune. It turns out, the fraudsters are doing this all over town, all over the state, all over the country. It's not just a problem for your bank, it's a problem for all the others as well. So one day you and the other bankers decide to get ahead of it....
Watch the Trailer
Release Date: November 8th, 2024
Let’s say you’re a software engineer at a major video-streaming company. This could be Netflix, Max, YouTube, or even Twitch or Kick. Either way, your company’s core competency is video: ingesting it, processing it, encoding and decoding it, hosting it, and streaming it out to consumers. And you’re good at it. But a new feature request just came down from on-high, and it’s, well, as we like to say: "non-trivial." What feature? Search....
Watch the Trailer
Release Date: November 7th, 2024
If you’re a lover of fine art, I have a little bit of a spoiler for you today. Some of the most iconic pieces are not actually fully original anymore. If you go see the Mona Lisa, The Last Supper, Guernica, Starry Night, or even the ceiling of the Sistine Chapel, you’re not seeing that art as the painter left it however many decades or centuries ago. You’re seeing a version of the art that curators and historians have attempted to revive and restore to reflect what it looked like back when it was first painted....
Watch the Trailer
Release Date: November 6th, 2024
If you ever rent an Airbnb out in the desert somewhere, like Joshua Tree, for example, your host is probably going to spend quite a bit of time talking to you about your shoes. Where to put them, how to put them there, where not to put them, what to do before you put them on, and what to do if you ignored the first instructions and accidentally left them somewhere you weren’t supposed to leave them....
Watch the Trailer
Release Date: November 5th, 2024
What makes good code? If you have two snippets that solve the same problem without throwing errors, what makes one of those snippets better than the other? More advanced than the other? More maintainable than the other?...
Watch the Trailer
Release Date: November 4th, 2024
Think back for a moment to the first time you ever drove a car. It was you, it was whoever was teaching you, the steering wheel, gas, brake, clutch, and the shifter. If you’re a bit younger, then maybe no clutch. But as far as advanced technology, that was probably it. You may have had a radio, or some way to play music, but if you're like me you didn't have a phone, GPS, lidar, cameras, motion sensors, traffic data, bluetooth, or anything else fancy....
Watch the Trailer
Release Date: November 3rd, 2024
Late on the night of March 12, 1928, the St. Francis Dam, a crucial piece of Los Angeles’s water infrastructure, failed without warning. Built just a few years earlier, the dam had been a symbol of progress, an ambitious project to secure water for a growing city. But at 11:57 p.m., a crack turned into a rupture, and within seconds, the dam broke apart, releasing billions of gallons of water into the San Francisquito Canyon. A wall of water, over 100 feet high, surged downstream, taking homes, farms, and hundreds of lives with it. By the time it reached the Pacific Ocean, over 50 miles away, the flood had left a trail of destruction across the valley....
Watch the Trailer
Release Date: November 2nd, 2024
A couple of months ago, a video went viral on TikTok of a patient named Jessica explaining the lengths she has to go to just to get her various medical providers on the same page. Jessica carries a huge binder to her medical appointments, and from the sounds of it, that binder contains virtually every x-ray, procedure, medication, lab result, and exam they've ever had. It includes everything that any doctor who has seen them has ever documented, prescribed, ordered, or said about Jessica. Some people might watch that video and think: Why on earth is Jessica doing this? Why is this necessary? But for the millions of people who have watched and rewatched this video, for the numerous people in the comments who have asked Jessica how to structure their own patient binder, what information to include, and whether to use a cover page, Jessica really strikes a chord....
Request this Episode
Release Date: November 1st, 2024
If you go to Google Trends or the Keyword Planner and look up the worldwide search traffic for the word “Kafka,” you might be forgiven for assuming that Metamorphosis must be on the bestseller list. Every month, there are almost 700,000 searches for that term. What gives? Did every high schooler in the world get assigned the same book report? No, searches for Kafka have been rising for a decade and started spiking a year ago not because of Franz Kafka but because of Apache Kafka, the open-source distributed event streaming platform....
Watch the Trailer
Release Date: October 31st, 2024
This paper has a whole lot going on, and from the title, it might not be entirely clear what it’s focusing on. At the highest level, this paper is about creating a better system for Software Fault Prediction (SFP). That is: shoring up the quality of the software you deliver by using machine learning model that can look at your codebase and predict where issues will arise. SFP isn’t by itself a holistic Quality Assurance process, but it can be a meaningful part of a larger QA/QC regime. So everything we’re going to talk about in this paper does in some way tie back to that: creating better SFP tools so that developers can ship better code....
Watch the Trailer
Release Date: October 30th, 2024
Just outside of Geneva, right near the border between Switzerland and France is the charming little municipality of Meyrin. Population about 26,000. They’ve got some apartment buildings, some low-rise office parks, a few hotels, some restaurants, cafes, a train station, and oh yeah… they’ve got the world’s largest and highest-energy particle accelerator: the LHC. CERN’s Large Hadron Collider. If you’re driving around, you won’t notice the LHC at all because it’s actually buried about 100 meters underground. This 17-mile track has been home to some of the most momentous experiments in modern physics. If you remember the race to identify the Higgs boson subatomic particle a few years ago...that was here. They did that. And that’s just one of the types of experiments that the LHC does....
Watch the Trailer
Release Date: October 29th, 2024
Imagine it’s 2003, and your friend brings a thumb-drive to school. On it: a pirated version of Outkast’s new album, Speakerboxxx/The Love Below. He was up all night on Limewire finding and downloading a decent copy. But he did it. This copy is solid, he listened to the whole thing on his PC before he copied it onto the drive. You borrow the thumb-drive and take it home, and plug it into your Mac. Error. You try again. Error again. Your Mac can’t read the drive. Something about NTFS. The computer suggests you reformat the drive, a process that will erase all the files. You stare at the screen, no Roses for you today....
Watch the Trailer
Release Date: October 28th, 2024
Let’s say I am a despot at the head of an oppressive totalitarian government. You are one of the millions of people who live in the country under my control. One way I can retain my position and stay in power is through mass surveillance, keeping an eye on what everyone is saying, what they’re doing, where they’re going, and who they’re associating with. A big part of that is monitoring what they’re doing online: what sites they’re visiting, how long they’re staying on those sites, and how much information they’re sending to and receiving from those sites. If I can figure out every website that everyone is visiting and every app they’re using, I can use that data to crack down on journalists, dissidents, activists—you name it. Even if I can’t see the information they’re sending to or receiving from these sites, just knowing the sites they’re connecting to is enough for me to act....
Watch the Trailer
Release Date: October 27th, 2024
If you’ve ever tried to train a deep neural network, you’ve probably spent a lot of time waiting. Waiting for training, waiting for boosting, waiting for validation runs, waiting waiting waiting. Why is that? Well, remember in school when you learned matrix multiplication, and you had to do it by hand? Remember how long that took? How tedious it is? Well yeah, your computer feels the same way. The crux of the problem here are MACs: Multiply-Accumulate Operations, in which matrix multiplication is an integral part. MACs are what the machine is doing, and what’s taking so long....
Watch the Trailer
Release Date: October 26th, 2024
If you’ve been listening to Journal Club for a while, you’ve probably noticed that a lot of the Machine-Learning, Deep Learning and Computer Vision episodes have a similar story arc:The researchers have a problem. They collect data. They clean and normalize that data. They extract and select the features they care about. They pick a few algorithms to test against each other. They run the training. They choose a few metrics to use to validate the models and benchmark them against each other. Sometimes they also stack models together or use some kind of ensemble learning. That storyline exists in Journal Club so frequently because that is the most common flow for how this research happens. But what our storylines don’t necessarily capture, is that many of these steps are highly manual, tedious, and repetitive. I might, for example, casually mention that a researcher trained models on KNN, Random Forest and SVM, but that doesn’t mean that those three training cycles looked anything like each other. They might have been in totally separate programs or notebooks, with different libraries, happening at different times on different machines, with a number of discrete manual idiosyncratic steps for each one. Getting different models trained with different algorithms for a given set of data isn’t generally a push-button thing....
Watch the Trailer
Release Date: October 25th, 2024
When I go on vacation, I like to do: nothing. I want to fly to wherever I’m going, sleep all day, chill all night, lay around the room, lay around the pool, order room service, be a sloth. For me, the idea of having to wake up early, put on actual clothes, and go do something sounds completely insane. Why anyone would jam-pack their vacation time with an itinerary full of stuff they have to do is beyond me. That just sounds like work....
Watch the Trailer
Release Date: October 24th, 2024
We all use search engines every day, for a variety of tasks. And our use of them can be broadly categorized into two types: standard searches and exploratory searches. In a standard search, a user looks up information when they already have some understanding of the subject. Even if they don't know the specific answer, they know enough to craft a coherent query that leads to relevant results. For example, searching for "what is the capital of France?" is straightforward because the user knows the general structure of the answer (a single city name) and can phrase the question clearly. This is the type of search traditional engines like Google excel at—they deliver concise, accurate results based on well-defined, unambiguous queries....
Watch the Trailer
Release Date: October 23rd, 2024
Let’s say you’re in charge of network security at a decent sized eCommerce site. Your company’s big enough that they don’t use AWS or GCP or any other cloud provider, they just run their own machines, maybe in a datacenter or colo-center, or maybe even in a server closet right next to your desk. Either way, that network is a real physical thing that your company relies on, and everyone’s counting on you to protect it. Everything’s going fine for a while. But then, a week before your biggest sale of the year—a sale that accounts for a huge percentage of your company’s annual profits-–-a week before that, it happens....
Watch the Trailer
Release Date: October 22nd, 2024
Toe-to-toe, a little honeybee doesn’t stand a chance against a hornet. Most hornets are 2-3 times the size of a bee, plus they have a thick exoskeleton like a lobster. And they’re mean. When a hornet enters a honeybee’s nest, it’s not there to make friends. It’s there to kill bees and steal food. But honeybees have a secret weapon: a technique called "balling". The bees latch onto the intruder one by one, eventually surrounding it in a blanket of bees. As they hold on, they vibrate their flight muscles, generating heat. Lots of heat. The hornet gets hotter, and hotter, and hotter, and then it dies. Victory: bees....
Watch the Trailer
Release Date: October 21st, 2024
Let’s say you and I are Oompa Loompas in a chocolate factory. We have various jobs, and we do all the meaningful work, while the crazy guy in the hat gets all the credit. But that’s fine, it’s what we signed up for. Your job is to take caramel squares and dip them in chocolate. Then you hand them to me. I sprinkle a little salt on top and carefully wrap each chocolate-covered caramel in a cellophane wrapper, twist the ends, and then place that carefully in a box lined with tissue paper. You dip, I wrap....
Watch the Trailer
Release Date: October 20th, 2024
In September 2023, Neuralink received approval to begin human clinical trials for a neural implant called the N1. Their study, nicknamed “Prime,” builds on years of animal trials conducted on pigs and monkeys. These studies were so revolutionary and buzzworthy that even the participants, the animals themselves, became internet celebrities. Millions of people watched as Pager, the monkey, used his Neuralink implant to play ping pong with his mind....
Watch the Trailer
Release Date: October 19th, 2024
In late October 1997, Atlanta Georgia played host to a 4-day technology conference: the 1997 International Conference on Network Protocols. At that event, a paper was presented by a 3-person team from AT&T Research Laboratories, Lucent Technologies Bell Labs, and Fujitsu Laboratory of America. AT&T, or “Ma Bell,” had been broken up a decade before in antitrust proceedings, but these ostensibly now-independent entities still collaborated on research. The paper they presented had all the hallmarks of Bell Labs innovation, namely being decades ahead of its time. The paper was called “A cooperative packet recovery protocol for multicast video,” and in it, they outlined a system in which packet loss between a sender and a receiver could be mitigated by a 3rd party server that steps in to replace packets as they’re lost....
Watch the Trailer
Release Date: October 18th, 2024
In the 2017 mayoral race in Jackson Mississippi, an unusual thing happened. The democratic incumbent, Mayor Tony Yarber, was challenged by members of his own party. Eight of them. They attacked him on his record, on his policies, and on his plans. But most of all, they attacked him on potholes. You see, potholes are endemic in Jackson. About 200 miles north of New Orleans, Jackson is often hot and humid. The moisture penetrates the road surface, seeping into small cracks, and weakening the foundation. Over time, the soil is less able to support the road surface, and as traffic passes over it, deformities begin to form. First rutting, then larger dips and cracks. Meanwhile, scorching midday temperatures soften the road, then colder nights cause it to contract again, and the cracks and dips get worse. Water fills the deformities and the vicious cycle continues. In Jackson that meant pothole after pothole after pothole. The residents had seen enough, and were ready for someone to step in and fix the problem....
Watch the Trailer
Release Date: October 17th, 2024
In 2020, Apple revealed the first release in their “M” series of computers, their foray into designing and building their own chips. No longer would they depend on Intel's processors to power their machines; the future of Apple would be vertically-integrated chip development. And it all started with The M1. While most of the headlines of the time were about the sheer processing power of the computer, many articles buried the real lead. The incredible part of the announcement was less about Apple bringing chip-manufacturing in-house, and more about the fact that they had switched from x86 to ARM. This wasn’t just a coup d’etat for ARM, it was a win for RISC: The Reduced Instruction Set Computer architecture....
Watch the Trailer
Release Date: October 16th, 2024
The past decade has seen nothing short of a revolution in privacy protection. From GDPR in Europe to CCPA in California, to third-party cookie restrictions built into iOS, and stronger browser defaults around HTTPS. Compared to the wild-west days of the early and mid-2000s, the internet consumer today has far more protections and avenues for recourse than ever before. The operators of web applications, mobile apps, APIs, newsletters, and other internet services are governed by strict compliance regimes that dictate what they can and can’t do with your personal information. Importantly, these regimes also specify how they must care for, handle, and protect your personal data—securing it from bad actors, encrypting it, and ensuring its safety....
Watch the Trailer
Release Date: October 15th, 2024
In the fall of 2021, after nearly six months in the halls of Congress, the Infrastructure Investment and Jobs Act (IIJA) was finally signed into law. This package, also known as the Bipartisan Infrastructure Law (BIL), allocated $1.2 trillion of funding, with $65 billion of that going specifically towards modernizing America’s power grid. Of that $65 billion, around $13 billion (20%) was set aside just specifically for power grid security. That’s thirteen Billion dollars to secure a system that, if you’re like me, you might’ve thought was already secure. Turns out, we were wrong. Not only is the grid vulnerable to a wide range of attacks from large state actors, but as the grid modernizes into a "smart grid," the attack surface actually grows...
Watch the Trailer
Release Date: October 14th, 2024
Okay, so this almost never happens. 99% of the time, on Journal Club the papers we are covering are advancements in computer science. That’s our bread and butter. But once in a blue moon, an article comes out that is the complete opposite. An article that says, in effect: "Hey everyone: when it comes to this one particular technology, we have all gotten out over our skis. We as an industry are making claims that a technology is capable of something it’s not actually capable of." These kinds of articles are rare, but I think they contribute just as much to our understanding of the state of the industry, as the other articles do. So it’s in that spirit that I bring you today’s paper:...
Watch the Trailer
Release Date: October 13th, 2024
Let’s pretend that your name is Janet and you are the CEO of Janet’s Bicycle Company. You sell bicycles to distributors, and you make those bicycles in a factory. The raw materials and third-party components come into the shipping-dock, and you’ve got a production/assembly line where you transform that pile of metal and rubber into a finished bike. From the start of the industrial revolution all the way up to the 1960s, the individual tasks involved in a production-line like yours were manually done by a human being. That person would likely have been using tools, and those tools might have provided leverage, but at its core industrial production was still a human-centered manual task. Bicycles, toasters, cars, shoes, and tape-measures, these were all made by a real person on an assembly line....
Watch the Trailer
Release Date: October 12th, 2024
In May 2021, the White House issued Executive Order 14028, the "Executive Order on Improving the Nation’s Cybersecurity". This order, among other things, directed NIST (the National Institute of Standards and Technology) to develop best-practices around software supply-chain security. Before we go on, I want to disambiguate that phrase. We are not talking about supply-chain software security. That is the security of the software that manages a business' supply chain. No. We’re talking about software supply-chain security: the security of the supply chain of the software we build....
Watch the Trailer
Release Date: October 11th, 2024
Sweet Mama’s Tomato Mix is a tomato-paste produced by the Weddi Africa Tomato Processing and Agro Farm. Weddi is a vertically integrated wholesaler: it operates the farm, it runs the processing plant, and it owns the distribution centers. It delivers five products (pastes in various-sized containers) directly to retailers in its own fleet of trucks, driven by its own drivers. The company operates in Ghana, in or around the city of Kumasi, a metropolitan area that is home to nearly 4 million residents. Weddi's main factory is about a hundred miles outside of the city, but its distribution hubs are closer to the city limits....
Watch the Trailer
Release Date: October 10th, 2024
A few days ago it was a friend's birthday. So at some point during the day I pulled out my phone, went to my message threads, found that person, and typed “Happy”. But as I was typing B-I-R-T-H-D-, the auto-suggest on my phone popped up with an emoji of a birthday cake 🎂. So I clicked on the suggestion, and it replaced the word I was typing with the cake. So what I was left with read as “Happy [Cake]”. I thought to myself, "Why would it think I wanted that?" Why would it think I was replacing the word Birthday with an emoji of a cake? Clearly I wanted to complete the word “Birthday” and then put a cake afterwards. Isn’t that how everyone uses emojis? As decorative accents after the text?...
Watch the Trailer
Release Date: October 9th, 2024
Here’s an uncomfortable thought about your ISP. (Your Internet Service Provider). When it comes to privacy, your interests and their interests are not aligned. Why? Because from their perspective, they need to provide you with a certain quality of service. In order to do that, they need to perform load balancing, routing, traffic prediction and forecasting. They need to identify malicious traffic flows and DDOS attacks. And if you live in a jurisdiction which doesn’t practice net neutrality, then part of their service may include application-specific metering. All of these things are predicated on their ability to classify the traffic that is passing through their network....
Watch the Trailer
Release Date: October 8th, 2024
Imagine for a second that you are an IoT device. You’re a smart watch, or a smart speaker, or a smart lamp, or a smart fridge. Either way you’re smart, you’ve got a processor onboard, and you can handle a lot of the computational workload yourself. But for some tasks, there are just so many calculations to do (so quickly) that your onboard chip isn’t up to the job. For those kinds of workloads, you offload the processing to a remote server, let that machine do the heavy lifting, and then return the result. But there’s a problem. When it comes to choosing the type of server that you send your workload to, you are spoiled for choice. You’ve got three main classes of options: Cloud Servers, Edge Servers, and Fog Nodes....
Watch the Trailer
Release Date: October 7th, 2024
In 1948, less than 90 years after Darwin published "On the Origin of Species", Alan Turing applied the concept of natural selection to Computer Science. In an unpublished essay titled “Intelligent Machinery,” he wrote: “There is the genetical or evolutionary search by which a combination of genes is looked for, the criterion being the survival value.” This line appears to be the first mention by any person, anywhere, of what is now referred to as an evolutionary algorithm. And now, 76 years later, this niche concept which has existed on the fringes of computer science for decades, is poised to finally have its time in the spotlight. Specifically, one technique within this field: Genetic Programming is coming to the forefront. Why? Because the same hardware advances, the same distributed systems and cloud computing advances that have enabled the proliferation of LLMs can also enable the widespread adoption of Genetic Programming....
Watch the Trailer
Release Date: October 6th, 2024
About a month ago, on September 7th, Héctor Garzó rounded the final turn at San Marino, crossed the finish line, and became the 2024 MotoE World Champion. He wasn’t riding any normal motorcycle. He was riding a fully-electric Ducati superbike, with a top speed of about 171 miles per hour. In fact, everyone in that race was riding a Ducati, because the storied Italian manufacturer has signed a deal to be the exclusive provider of electric motorcycles for the entire field through 2026. You see, Ducati is going long on electric motorcycles, and doing everything they can to push the technology (and the acceptance of it) to higher and higher levels. So far, they haven’t seen much success. The MotoE race where Garzo was crowned champion was streamed live on Youtube. As of this writing, the video only has about 11,000 views. Total. For a world championship....
Watch the Trailer
Release Date: October 5th, 2024
In the immortal words of Frankenstein’s monster: FIRE BAD! And unfortunately, we often don’t spot fires until they’re really bad. We’ll miss the small brush fires and only take notice when they’re large and out of control. This happens for a few reasons: Fires can start anywhere, even away from the cities. In the middle of open land, or in a forest where nobody will notice. If someone does notice, there can be a “bystander-effect”: everyone thinks that someone else must have called 911 to report the fire, or that someone else must already be handling it. So nobody ends up calling at all, and the fire grows and grows. Where I live in California, this is a big problem. Large parts of the state are sparsely populated open areas with lots of fuel for wildfires (dry grass and the like). So every summer a sad ritual repeats itself here. Loose cigarette butts, mismanaged campfires, downed power lines, or any of several other causes can be the trigger. And once it starts, the fire goes unspotted for a while, It grows, it spreads and eventually causes real damage. These fires cost California billions of dollars, every single year....
Watch the Trailer
Release Date: October 4th, 2024
NPM is a fantastic package manager in a number of ways, but it's really bad at search. To be fair, every other package manager is bad at it too: PIP, Crates, Maven, Homebrew, RPM, etc. They all struggle to give meaningful search-results for a query. When you have a problem that you need a library to solve, your chances of finding a relevant package might hinge on your ability to guess what that package might be named. The search-bar in NPM returns such irrelevant results, that you'll likely spend your day jumping from search engines, to Github, to blog posts, to youtube videos, to forum discussions. You'll go back and forth to NPM over and over again, trying out different packages to see what fits. You'll spend time reading their docs, checking their Github issues, checking if they're stable, well maintained and recently updated. It’s a pain, to say the least. I’ve always thought that non-programmers would be shocked to see how much of a Software Engineers day is spent trying to find and use new packages. Out of exasperation, many developers just choose the most popular package that seems like it might possibly do the job. Not the package best suited to the problem, not the package with highest test coverage, not the package with the best documentation, or the lowest number of open issues, just the one with the most impressive download graph....
Watch the Trailer
Release Date: October 3rd, 2024
Shenzhen is a bustling metropolis. 17.5 million people, millions of homes, millions of cars, millions of workplaces. But it wasn’t always this way; in fact it wasn’t even this way recently. Thirty years ago Shenzhen was a 10th of its current size, forty years ago it was a quarter of that. The growth experienced in this region over the last half-century has been incredible, and the city is still growing so quickly that local municipalities are having a hard time even tracking the growth. This makes the day-to-day practice of governing somewhat difficult. Common tasks like land-use classification, population density estimation, and even housing-price tracking have become non-trivial. The census is only conducted in China once every 10 years, so how are they supposed to keep track of the population in the meantime?...
Watch the Trailer
Release Date: October 2nd, 2024
First, we need to step back in time for a moment. In 2017, Vaswani et al published a seminal paper that would send shockwaves through the A.I. community: “Attention Is All You Need”. In it, the authors described a novel concept: the Transformer. Transformers (and their building-blocks called “transformer-blocks”) were a new type of a Neural-Network architecture based on the concept of self-attention. Self attention allowed for two key benefits: 1. The ability for tokens to maintain relationships with long-range dependencies. 2. The ability to parallelize both your training and inference. In other words, transformers made Large Language Models possible, set the stage for what would become the Foundation Models, ChatGPT, and arguably the whole AI boom of the last few years. I’m telling you this because the paper we’re about to dive-into builds on top of the concept of a transformer, utilizing an even newer concept: a Conformer. You see, after a few years working with transformers, researchers started to realize that they had a few limitations. Namely, transformers were great at determining the long-range relationships between tokens, but they struggled with short-range relationships or “local dependencies”. So the conformer was born, which utilizes a transformer but adds on convolutions for local patterns. Thus the name “con” from convolution and “former” from transformer. Conformers are designed to be great at both the big picture and the tiny details....
Request this Episode
Release Date: October 1st, 2024
In 2014 Masayoshi Son, the founder of SoftBank, revealed Pepper: A four foot tall humanoid robot that could talk, answer questions, and perform a variety of pre-programmed actions. Pepper had a tablet sticking out of its chest to display visual information to the person it was communicating with. Over the last decade thousands of these little robots were sold around the world and reprogrammed to become everything from front-desk receptionists, to tour guides, to waiters. Most of these installations were experiments: cute demos and a sneak-peak of what might be possible in the future....
Request this Episode
Release Date: September 30th, 2024
Cotton farmers in Pakistan have a problem: The Silverleaf Whitefly. This little insect is a carrier of the Cotton Leaf Curl Virus (CLCuV), which causes Cotton Leaf Curl Disease (CLCuD). Cotton Leaf Curl is serious. Once infected, the virus causes photosynthate blockage in the veins of a cotton plant. Photosynthates are the compounds produced during photosynthesis, so having a photosynthate blockage is really really bad. The tertiary veins on the leaves turn yellow and thicken, the leaves start to curl, then secondary veins get blocked, which reduces surface area of the leaves. This means less photosynthesis, and the downward spiral continues. Infected plants end up shorter, with curly scaly leaves, and most importantly: the stress on the plant can cause cotton yields to drop 80%....
Request this Episode
Release Date: September 29th, 2024
Remember a year or two ago when “passwordless” authentication for websites was all the rage? News articles from the time proclaimed that passwords were dead and that passkeys (and related technologies) were clearly the future. Well…what happened? From what I can tell, a lot of people gave passkeys a try, hated them, and went right back to using passwords. To be fair, passwordless authentication for mobile lock-screens and mobile payments did become a reality (in the form of biometrics) but passwordless authentication on the web never really caught-on (outside of Oauth). It turns out, if you want to disrupt something as ubiquitous as passwords you need to offer a solution that is waaaay better. Not just a slight improvement, but a full step-function better. And in the minds of many users, passkeys didn't reach that bar....
Watch the Trailer
Release Date: September 28th, 2024
Worldwide, approximately 1/5th of children under the age of five suffer from some sort of physical growth-impairment. That figure has improved virtually every year for decades, but there’s still a lot of work left to do. Growth-impairment has traditionally been viewed strictly as a function of malnutrition, but there is an increasing awareness that there are a multitude of other factors that correlate strongly with growth-impairment. Identifying these risk factors could help clinicians intervene earlier and more effectively. In this research the authors attempt to build a predictive model for growth-impairment: a model that can analyze other household and familial risk factors, and inform medical professionals of any elevated risk....
Watch the Trailer
Release Date: September 27th, 2024
In the world of virtual reality, developers and designers care a lot about something called "embodiment". Embodiment is the feeling or sensation that you’re "in" the character/avatar that you're playing. You are the avatar. You have agency, you have ownership. You have what’s referred to as "visual-motor congruence", and you feel accurately represented by the avatar in the virtual universe. Remember that scene in the Matrix when Neo wants to learn Kung Fu? They plug him into the simulation, and suddenly he’s in a room with Morpheus training and fighting. Well, in that moment Neo knows that what he's experiencing is a simulation, but that's counteracted by the fact that he has “high embodiment” in his avatar. It’s not his actual body, but in a way, it is. He’s not really in that room, but in a way, he is. That's embodiment....
Watch the Trailer
Release Date: September 26th, 2024
You've probably heard of both deforestation and reforestation. But what on earth is afforestation? Afforestation is the process of planting and growing trees where none originally grew (or they haven’t grown in a long time). Afforestation is not replanting a recently-cut forest, it’s creating a forest where there was previously grassland, desert, or even ice, for example. Though it sounds a bit odd, afforestation is a key practice in sustainable timber production. Rather than chopping down existing forests, a company will grow a brand new forest, chop it down, and repeat. But, this technique does have its drawbacks. Commercial afforestation is extremely water-hungry. And as a result, many countries regulate it tightly. In South Africa, where today’s research takes place, you can’t grow a new forest without first being issued a water license. The licenses are granted based on the government's statistical models that show how that proposed crop would affect the overall energy balance, environment and water table. In order to create those models, the government needs to know a key metric for the species of tree you want to grow: Evapotranspiration (ET)....
Watch the Trailer
Release Date: September 25th, 2024
Now, if you’re like me: alarm bells are already ringing. These shootouts happen all the time, and in many cases the experimental design is seriously flawed. In my opinion, this paper is no exception. So rather than take this article at face value (and report their findings as fact), we’re going to take a more critical lens. I am going to present their research and their findings, but I’ll spend much more time than usual on history and context, and then I'll use that to inform counter-arguments to their analysis. This study is imperfect, but it’s not without value. They did unearth some interesting findings, I just don’t think those findings should be taken at face value; so we won’t. The way we'll get the most out of this paper is through a critical, interrogative lens, so let's do that....
Watch the Trailer
Release Date: September 24th, 2024
This year over 90% of the new cars sold in the United States have some kind of driver-assistance technology built in. That means some kind of system that can aide in the steering, braking, or both. This basket of technologies is broadly referred to as autonomous driving, semi-autonomous driving, self-driving or full-self driving etc. There’s a lot of names for it, and a lot of implementations. To keep everyone on the same page, self-driving is classified on the SAE Driving Automation Scale, from Level 0 (minimal or no self-driving) all the way up to Level 5....
Request this Episode
Release Date: September 23rd, 2024
Serverless is great. In a number of ways it's truly a transformative technology. And it's spreading: every year more hosts jump into this space to offer their own Function as a Service (FaaS) platforms: from AWS, GCP and Azure, to the edge offerings available on Cloudflare, Fastly and Akamai, to the open-source options like OpenFaaS. But as with anything, serverless is not without its drawbacks. From the "uncapped costs" problem, to the "how do I run this locally?" problem, there are obviously a few areas where the developer-experience could be improved. And that's to be expected given how nascent the technology still is. But there is one issue that's a deal-breaker for many devs. An issue so off-putting that it actually prevents them from adopting serverless at all. That issue is the "cold start" problem....
Request this Episode
Release Date: September 22nd, 2024
A half-hour drive south west of Rome sits an archaeological site called Ostia Antica. It’s the ruins of an ancient city from two millennia ago. Right in the center of the site is a brick house called “Casa Di Diana” (Diana’s House). It’s believed to originally have been a five-story building used for both residential and commercial activities. Now it’s a series of half-standing walls, chambers and pillars. From an archaeological perspective, Casa Di Diana is important. It’s one of the many treasures left from the Roman empire. A site that the archaeologists who have been studying it would love to share with the world. But, in its current form, it’s simply too fragile to let anyone but trained professionals visit....
Request this Episode
Release Date: September 21st, 2024
Right now, in 2024, writing programs for quantum computers feels like you’re writing code for IBM mainframes in the 1960s. Here’s what it's like: You write your code. You find a vendor with an available quantum computer that has a compiler compatible with your code. You send your code to that vendor in the hopes that they will compile and run your code sometime soon. You wait and you wait and you wait (for minutes or hours or days) You eventually get a readout back of the result of your program and/or your errors. The whole process is like stepping into a time machine!...And not in a good way....
Request this Episode
Release Date: September 20th, 2024
Way back in 2014 Google released an open-source version of Borg, their in-house cluster management system. Borg was immense and highly specific to Google’s systems and servers, but this new open-source version would be different: It would be smaller, portable to different types of systems and hardware, focused narrowly on container orchestration, and generally useful for a number of workloads. They called the new project: Kubernetes. Kubernetes came out at exactly the right time. Adoption of Docker had spiked in the previous years, microservices were all the rage, and companies of all sizes were grappling with a non-trivial task: container orchestration. With no robust turnkey solutions available, developers were either building one-off systems from scratch, or struggling to port their existing CI/CD systems to work with containers. Google was uniquely positioned to step in. They had been early adopters of containers, and had spent a decade learning how to run them at scale. Kubernetes (or k8s as it came to be known) was the distillation of everything they’d learned...just scaled down....
Watch the Trailer
Release Date: September 19th, 2024
For as long as there have been classrooms, there have been students in those classrooms figuring out ways to cheat. And as long as students have been cheating, there have been teachers and administrators trying to catch them in the act. Nothing about this cat-and-mouse game is new. What is new however, is the power of the AI tools that students now have available to them. Foundation Models exposed through interfaces like ChatGPT can produce a decent term-paper for a student in seconds, and this has drastically changed the playing field. So what’s a teacher or professor to do? How can they be expected to identify which papers were actually written by their students, and which were copied-and-pasted from a GenAI tool?...
Watch the Trailer
Release Date: September 18th, 2024
The researchers in this study wanted to compile a database of police-reports, and then group those reports by the type of crime-scene described within each document. Seems simple enough, no? Unfortunately the authors immediately hit a wall: It turns out that police reports contain so much sensitive and private information that they’re rarely made public. So the authors couldn't get their hands on any of them and had to figure out a workaround: Rather than deal directly with police reports, they would use public court documents instead. Their theory was that court documents often have the narrative-areas of the relevant police-reports embedded within them (literally copied and pasted from the police reports right into the court docs). So the researchers could still build a database of police-writing, they'd just have to get that writing from court documents instead of directly from the police reports....
Watch the Trailer
Release Date: September 17th, 2024
If you find yourself working with IoT devices, you’ll probably spend a lot of time thinking about light. No matter how elegantly you program your device, and no matter how cleanly you construct the backend, the interactions between your device and its server will only ever be as quick and responsive as light allows. Light travels at around 300,000 kilometers per second. That's around 186 miles per millisecond. So for tasks that require a round-trip to the server, the ultimate latency and lagginess of your device has nothing to do with your programming abilities; it has to do with how physically far the IoT device is from the nearest data center. Let's say you're building a device with strict latency requirements, and you want a roundtrip baseline of 6 milliseconds. If that’s the case, your server can’t be more than around 560 miles away from the device. Any further and light won’t have enough time to get to the server and come back....
Watch the Trailer
Release Date: September 16th, 2024
One of my earliest memories is of the Loma Prieta earthquake. It hit California on a Tuesday afternoon in 1989. I was three years old and I was at day-care, playing in the sandbox. All I remember is that my universe went from still and quiet to violent and loud very suddenly. In just twenty seconds 63 people died, nearly 4,000 were injured and over 12,000 displaced. A piece of the upper deck of the Bay Bridge collapsed, the Nimitz freeway in Oakland fell down, and local homes and businesses suffered billions of dollars in damages. There was no warning. None. Back then Earthquake Early Warning systems (EEWs) had been invented, but they weren't installed in California until a few years later. So when a big earthquake hit, nobody had a chance to prepare or take cover. Authorities had no time to slow freeway traffic, turn stop-lights red, or close bridges. Now, thankfully that’s all changed. And for good reason! The "big one" (a megathrust earthquake) could happen in California anytime, and is projected to happen sometime in the next few decades. This time, when it hits, we'll at least have a few moments of warning....
Watch the Trailer
Release Date: September 15th, 2024
Picture this: It’s some time in the not too distant future, and you’re out with your friends one night having a few drinks. Maybe a few too many drinks. At some point you decide to leave, jump in your car and start driving down the road. Within a few blocks your car becomes aware that you’re driving drunk and it takes action. What happens next is anyone’s guess: maybe the car takes your control away and goes into full self-driving mode. Maybe it just pulls over and turns itself off. Maybe it calls you a cab, maybe it calls the police. Who knows. The more pertinent question in front of us today is: How on earth did the car know you were driving drunk in the first place?...
Watch the Trailer
Release Date: September 14th, 2024
As lithium-ion batteries age they get worse and worse and eventually reach the end of their useful life. This is a big issue for companies who operate fleets of UAVs (drones). If they replace the batteries in their drones too early, they’re wasting money. If they replace them too late their missions and deliveries get compromised. So they need a way to predict exactly when a drone’s battery is going to reach the end of its useful life. This paper is using AI to do exactly that. Let's take a look....
Watch the Trailer
Release Date: September 13th, 2024
If you’ve been around computer science for a while you’ve undoubtedly heard of the Bin-Packing problem. While it's surely difficult, imagine how much worse it would be if you had to optimize for several variables at the same time. That’s the situation faced by high volume SaaS companies who are trying to allocate Virtual Machines (or containers) onto their bare metal compute. They’re not just trying to minimize resource waste and cost, they need to balance power consumption, failure rates, application responsiveness, uptime, and more. Doing that is NP-hard and incredibly complex. In this paper the authors conceive an elegant solution: A framework called MoVPAAC....
Watch the Trailer
Release Date: September 12th, 2024
In the tech industry we’ve gotten pretty used to disruption. It seems like every few months there’s a new technology that’s "going to change everything". But very few technologies have the potential to be as truly disruptive as autonomous driving. Why? Because in the United States, it’s estimated that around a third of all civilian jobs involve driving in some way. And around 3% of all jobs are full-time driving. So it’s not hyperbole to say that a huge number of American workers can and will have their daily lives and employment significantly impacted by autonomous driving. Now, you may think autonomous driving is a bad thing (there’s an argument for that), or you may think it’s a great thing (there’s an argument for that as well). But what I think we can all agree that technology this disruptive can't and shouldn't be solely in the hands of a few giant companies with closed-sourced systems....
Watch the Trailer
Release Date: September 11th, 2024
Here's a fun fact: your credit score isn't what it used to be. Now, I don't mean that your personal credit score has gone down, I mean that the days when banks relied solely or even mostly on your credit score to determine credit worthiness are gone. Long gone. These days: When a bank needs to determine your credit worthiness, it might look briefly at your score, but then its going to use its own proprietary A.I. to review all of your credit and banking history, as well as all the publicly available data that it can find on you. Then its model will spit out a credit worthiness decision. And lest we forget: credit worthiness determines a lot. It doesn't just determine whether or not you're approved for a loan or credit card, but the interest rate you get, the size of the credit limit, the rewards you qualify for, and more....
Watch the Trailer
Release Date: September 10th, 2024
The field of Psychiatry doesn't always lend itself well to data-science. The field of Computational Psychiatry, on the other hand, thinks we can do better. In this new study, researchers built a proof-of-concept model that should be able to take a new patient (who is admitted for depression, mania or psychosis), review their history, and then project forward the ultimate progression and possible treatment of their condition. Far fetched? Let's dive in and find out....
Watch the Trailer
Release Date: September 9th, 2024
A few companies are planning on putting their datacenters in the stratosphere. Yes, the stratosphere! Their theory is that the sheer amount of free cooling would justify the costs and headache. For a moment let's put aside how they're going to get the machines all the way up there, and then let's put aside how they're going to keep them afloat. There's still one big problem: TCP isn't going to work very well in the stratosphere. Not well at all. That's according to a new study in which the researchers are proposing an alternative new variant of TCP for use exclusively in the stratosphere. They're calling TCP Stratos....
Watch the Trailer
Release Date: September 8th, 2024
In Computer Science we don't talk about cashews very often, but maybe we should. A team of researchers in Vietnam just trained a computer-vision classifier on top of YoloV8 that can grade cashew nuts as they're coming down a conveyor belt. It can perform the grading in milliseconds, with less than a 3% error rate, running entirely on cheap off-the-shelf components. Impressive!...
Watch the TrailerThat's it. You've reached the end of the archive.