Peter Cotoia

The process of digitization, as with any change, comes with its pro and cons, gives and takes. For instance as we discussed in a n earlier post about the Jacquard Loom, the invention itself was a catalyst in the textile industry, paving a new path forward that perfectly joined efficiency, intricacy, and repeatability. Its invention while incredible, displaced long standing jobs of weavers by automating the process. This give and take is the cost of innovation.  Digitization in a more modern sense - through the transmission of radio, electric, or wireless signals - also came with a give and take. While it did create and displace jobs throughout the century with the onset of the next  best technology, the creation of the technology itself demanded incredible amounts of trial and error. Most notably, the joining of analog and digital systems to create the first national communication network: the land line. How was it that someone in New York could hear their relative over the phone...

This week we got the chance to listen to the first chapter of, Genius Maker, by Cade Metz. In this first chapter, Metz transports listeners back to 2012, when Geoffrey Hinton, a professor at the University of Toronto, and two of his graduate students founded a small company called DNNresearch. Their goal was to bring their breakthroughs in neural networks—what we now call “deep learning”—to the forefront of artificial intelligence. At the time, most of the AI community had been focused on symbolic AI, which relied on explicit rules and logic. Generative and learning-based systems were widely dismissed as too complex or computationally unrealistic. Hinton’s work changed that narrative, proving that machines could, in fact, learn patterns and representations from massive amounts of data. Hinton and his students took their company to the NIPS conference in Las Vegas, where they discreetly held an auction for its acquisition. Google, Microsoft, Baidu, and DeepMind all...

In a story that bridges mathematics and artificial intelligence, The Wall Street Journal profiled Axiom Math, a startup founded by 24-year-old Carina Hong. After leaving her PhD program at Stanford, Hong set out to create an “AI mathematician” capable of solving complex mathematical problems, generating detailed proofs, and even discovering new theorems. Axiom’s model aims to translate the language of mathematics—drawn from textbooks, academic papers, and journals—into a system that can produce and verify solutions autonomously. Backed by $64 million in seed funding and valued at $300 million, the company has attracted top AI researchers from Meta’s FAIR lab, uniting expertise in machine learning, mathematics, and computer science under one ambitious goal: to use AI as a tool for mathematical discovery and innovation. Hong’s background in mathematics and neuroscience, combined with her team’s AI experience, gives Axiom a strong technical foundation. The company hopes that its work...

  When we think about the Industrial Revolution, we usually picture steam engines, smoky factories, and mechanical progress. But one of the most important inventions of that time wasn’t just about industry—it changed how people began to understand information itself. In 1804, a French inventor named Joseph Marie Jacquard built a new kind of loom for weaving silk. At first, it looked like just another tool for making fabric faster. But what made it groundbreaking was that the loom could be “programmed” with punched cards. Each hole or blank space carried instructions, telling the loom exactly which threads to raise or lower to form a pattern. For the silk industry in Lyon, this was revolutionary. Complex, intricate designs that once required enormous amounts of time and skill could suddenly be woven more efficiently and consistently. But more importantly, Jacquard’s loom revealed something bigger: that you could take abstract information—like a design or even a picture—and...

Have you ever wondered how your phone can recognize your face, or how apps can tell the difference between a cat and a dog? Behind the scenes, much of this magic comes from something called a neural network — a way for computers to “learn” patterns, a bit like how our brains work. A neural network is made up of layers of tiny units called neurons. You can think of each neuron as a little station of information. Information travels from neuron to neuron in a neural network as it is processed. The first layer - the input layer - takes in the raw data, for instance the color of a pixel on an image. The last layer - the output layer - gives the answer — such as "that's a cat" or "that's a dog". In between are hidden layers, where most of the learning happens. The cool part is that neural networks don’t need us to "tell it" every detail; instead, it figures out what matters by "practicing" on lots of examples. This "practice" that a neural network does is kind of like how people practice....

Welcome to WordPress. This is your first post. Edit or delete it, then start writing!