In a world filled with misinformation and oddball theories, it’s inevitable to come across individuals who hold beliefs that defy basic logic and established facts. One such example is actor Terrence Howard, who famously claimed that 1 x 1 = 2. As baffling as this assertion might be, it presents an opportunity to explore how to effectively counter such erroneous ideas. Understand the Claim Before you can counter any claim, it’s essential to fully understand what is being stated. Terrence Howard’s assertion that 1 times 1 equals 2 contradicts fundamental arithmetic principles. Knowing the specifics of his argument, as well as the reasoning (or lack thereof) behind it, is the first step in preparing an effective response. Stay Calm and Collected When confronted with bizarre claims, it can be tempting to react with frustration or ridicule. However, it’s important to stay calm and collected. Reacting emotionally can escalate the situation and make productive conversation difficult. Approach the discussion with patience and a clear mind. Use Clear and Simple Explanations One of the best ways to counter misinformation is through clear and simple explanations. Here’s how you can address Howard’s claim about multiplication: Fundamental Principle: Explain that multiplication is a form...

A wave of AI-powered scams is sweeping across WhatsApp, costing UK families nearly half a million pounds in 2025 alone and it’s only May. Cybercriminals are now combining old tricks with new tech. In the evolving “Hi Mum” scam, fraudsters impersonate a loved one over WhatsApp and ask for emergency cash. The twist They’re now using AI-generated voice messages to mimic children’s voices, making the deception frighteningly convincing. “Scammers are increasingly getting better at manipulating people… cloning any voice is now simple, even in a matter of moments,”says Jake Moore, global cybersecurity advisor at ESET. By the numbers: 506 WhatsApp scams since Jan 2025 Victims lost £490,606 ($651,230) April alone: 135 cases, £127,417 lost How it works: You get a WhatsApp from an unknown number: “Hi Mum, I lost my phone.” They claim they’re locked out of their bank. They send a voice note and it sounds like your child. They ask you to urgently transfer money to a new account. A screen-grab excerpt of the WhatsApp ‘Hi mum’ text scam. Photograph: Santander The danger Scammers scrape social media for voice clips and personal details. Then they use generative AI to clone the voice and craft a believable story. “I...

As autonomous drones become more integral to logistics, emergency response, and industrial applications, the challenge of navigation in GPS-denied environments has driven technological breakthroughs. While GPS remains the dominant method for outdoor navigation, it is ineffective in locations such as warehouses, underground facilities, dense urban landscapes, and disaster zones. Recent research and industry innovations, including a case study from QuData on autonomous drone navigation without GPS, are reshaping how drones operate without relying on satellite signals. Unmanned aerial vehicles (UAVs) navigating without GPS must rely on alternative localization techniques such as visual odometry, LiDAR-based perception, and sensor fusion. By integrating data from cameras, inertial measurement units (IMUs), barometers, and radar systems, drones can build real-time environmental models that enable stable and accurate movement. But while computer vision and LiDAR systems have been used for indoor navigation, they struggle in low-light conditions or spaces with repetitive features. Addressing these limitations, a team at MIT has introduced MiFly, a system that leverages radio frequency (RF) waves and backscattered signals to provide accurate localization with minimal hardware requirements. MiFly operates using a single, low-power tag that reflects millimeter-wave signals transmitted by a drone’s onboard radar. Unlike traditional systems that require multiple markers or...

Summarizing, probability distribution function for Discrete Random Variables is called Probability Mass Function (PMF) and that for Continuous Random Variables is called Probability density function. 3. Derivates and Matrix Calculus I highly recommend these PDFs to brush up Matrix calculus fundamentals. Source – https://atmos.washington.edu/~dennis/MatrixCalculus.pdf Source – https://arxiv.org/pdf/1802.01528.pdf References: https://en.wikipedia.org/wiki/Dot_product Khan Academy – https://www.youtube.com/watch?v=dOr0NKyD31Q

The complexities of managing and deploying ML infrastructure continue to grow and can indeed be daunting. Some time ago, we shared “A Step-by-step Guide To Setting Up MLflow On The Google Cloud Platform,” which was well-received. However, as time has passed, advancements in technology have offered us the potential to simplify and automate this process further. And this is where Terraform comes into play. As a robust Infrastructure as Code (IaC) tool, Terraform enables automation, significantly streamlining your ML infrastructure management.  In this step-by-step guide, we’ll demonstrate how to seamlessly set up MLFlow on the Google Cloud Platform using Terraform, further optimizing the process and incorporating additional automation and scalability into your machine learning infrastructure. Whether you are already leveraging MLFlow’s capabilities to optimize machine learning experiments or just starting to navigate its potential, this guide will deliver a step-by-step tutorial to augment your work. What is Terraform? Before delving into the specifics of utilizing Terraform for our ML infrastructure management, let’s start with an explanation of what Terraform is.  Terraform is an open-source Infrastructure as Code (IaC) software tool created by HashiCorp. It allows developers to define and provide data center infrastructure using a declarative configuration language. Simply articulate...

February 25, 2017 Vasilis Vryniotis . 3 Comments The ALS algorithm introduced by Hu et al., is a very popular technique used in Recommender System problems, especially when we have implicit datasets (for example clicks, likes etc). It can handle large volumes of data reasonably well and we can find many good implementations in various Machine Learning frameworks. Spark includes the algorithm in the MLlib component which has recently been refactored to improve the readability and the architecture of the code. Spark’s implementation requires the Item and User id to be numbers within integer range (either Integer type or Long within integer range), which is reasonable as this can help speed up the operations and reduce memory consumption. One thing I noticed though while reading the code is that those id columns are being casted into Doubles and then into Integers at the beginning of the fit/predict methods. This seems a bit hacky and I’ve seen it put unnecessary strain on the garbage collector. Here are the lines on the ALS code that cast the ids into doubles: To understand why this is done, one needs to read the checkedCast(): This UDF receives a Double and checks its range and...

Chaotic systems, such as fluid dynamics or brain activity, are highly sensitive to initial conditions, making long-term predictions difficult. Even minor errors in modeling these systems can rapidly grow, which limits the effectiveness of many scientific machine learning (SciML) approaches. Traditional forecasting methods rely on models trained on specific time series or broad datasets lacking true dynamical structure. However, recent work has demonstrated the potential for local forecasting models to predict chaotic systems more accurately over longer timeframes by learning the numerical rules governing these systems. The real challenge is achieving out-of-domain generalization—creating models that can adapt and forecast new, previously unseen dynamical systems. This would require integrating prior knowledge with the ability to adapt locally. Still, the need for task-specific data constrains current methods and often overlooks key dynamical system properties such as ergodicity, channel coupling, and conserved quantities. Machine learning for dynamical systems (MLDS) utilizes the unique properties of such systems as inductive biases. These include fixed relationships among system variables and invariant statistical measures, like strange attractors or conserved quantities. MLDS models use these properties to build more accurate and generalizable models, sometimes incorporating probabilistic or latent variable techniques. While datasets of dynamical systems have been curated...