Data Science at Home

A sharp takedown of the AI funding frenzy and why many startups are glorified wrappers around existing models. A critique of investors mistaking pattern matching for technical understanding. Parallels drawn to past hype cycles like blockchain. Discussion of inflated valuations, the raise-build-sell playbook, and where real research gets drowned out by buzzword products.

VortexNet uses actual whirlpools to build neural networks. Seriously. By borrowing equations from fluid dynamics, this new architecture might solve deep learning's toughest problems—from vanishing gradients to long-range dependencies. Today we explain how vortex shedding, the Strouhal number, and turbulent flows might change everything in AI.   Sponsors This episode is brought to you by Statistical Horizons  At Statistical Horizons, you can stay ahead with expert-led livestream seminars that make data analytics and AI methods practical and accessible. Join thousands of researchers and professionals who’ve advanced their careers with Statistical Horizons. Get $200 off any seminar with code DATA25 at https://statisticalhorizons.com     References https://samim.io/p/2025-01-18-vortextnet/  

Also on YouTube   Two AI experts who actually love the technology explain why chasing AGI might be the worst thing for AI's future—and why the current hype cycle could kill the field we're trying to save. Want to dive deeper? Head to datascienceathome.com for detailed show notes, code examples, and exclusive deep-dives into the papers we discuss.   Subscribe to our newsletter for weekly breakdowns of cutting-edge research delivered straight to your inbox—no fluff, just science! 📧 Join the conversation! Our Discord community is full of ML engineers, researchers, and AI enthusiasts discussing papers, sharing projects, and helping each other level up. Whether you're debugging your first neural net or training your tenth transformer, there's a place for you. Link in the show notes! 💬 Newsletter https://datascienceathome.substack.com/subscribe Website https://datascienceathome.com     References CEO is obsolete AI is the new blockchain Dr Eliseo Ferrante NYU https://nyuad.nyu.edu/en/academics/divisions/science/faculty/eliseo-ferrante.html    

Mark Brocato built Mockaroo—the tool that taught millions of developers how to fake data. Now, as Head of Engineering at Tonic.ai, he's building the AI agent that's making his own creation obsolete. In this episode, we explore why static test data can't survive the AI era, what it means to "negotiate" datasets with an agent instead of scripting them, and whether we're heading toward a future where sandbox environments vanish entirely. From the hidden failures of legacy mocks to the security implications of agent-driven synthesis, Mark reveals what happens when data generation becomes a conversation—not a pipeline.   Sponsors   Tonic.ai Synthetic data solutions for software and AI development. Accelerate engineering velocity and ensure compliance with AI-powered data synthesis   This episode is brought to you by Statistical Horizons  At Statistical Horizons, you can stay ahead with expert-led livestream seminars that make data analytics and AI methods practical and accessible. Join thousands of researchers and professionals who’ve advanced their careers with Statistical Horizons. Get $200 off any seminar with code DATA25 at https://statisticalhorizons.com  

In this episode, Matt Lee, founder of Schematical and CloudWarGames, shares nearly 20 years of wisdom on AI and cloud infrastructure. He reveals that many companies struggle with decision-making rather than AI challenges. Matt explains when AI truly adds value, how to cut costs with strategic model choices, and the importance of manual testing before automation. He discusses common pitfalls in AI spending and offers practical advice on framing AI projects in financial terms to win over CTOs and boards.

Dive into the game-changing concept of Continuous Autoregressive Language Models, which compress tokens into vectors for speedier text generation. Discover how this innovation can slash AI costs by 44%. The discussion covers the implications of a likelihood-free training method and introduces the concept of semantic bandwidth for denser information. Also explored are the environmental benefits of more efficient models and the importance of open science versus corporate secrecy in research. Perfect for tech enthusiasts eager for the latest in AI advancements!

Discover how whirlpools can revolutionize neural networks with VortexNet. Fluid dynamics concepts, like vortex shedding and the Strouhal number, are reshaping deep learning solutions. Learn about adaptive damping and how vortex interactions create implicit attention without the typical complexities. This innovative approach tackles deep learning challenges such as vanishing gradients and long-range dependencies. Plus, explore practical applications in fields like finance and weather forecasting!

Fred Jordan, Co-CEO of FinalSpark, takes us inside the radical world of biological computing, where real neurons extracted from human tissue are being trained to solve problems that would require 10 megawatts in silicon. We explore the life support systems keeping these "wetware" processors alive, the ethical quandaries of computation performed by living cells, and why the messiness of biology might be exactly what AI needs next. From training cycles and reproducibility challenges to the surprising behaviors these neural networks display, Jordan paints a picture of 2030 where your devices might be powered by something closer to a brain than a chip.   Sponsors   This episode is proudly sponsored by Amethix Technologies. At the intersection of ethics and engineering, Amethix creates AI systems that don’t just function—they adapt, learn, and serve. With a focus on dual-use innovation, Amethix is shaping a future where intelligent machines extend human capability, not replace it. Discover more at https://amethix.com This episode is brought to you by Intrepid AI. From drones to satellites, Intrepid AI gives engineers and defense innovators the tools to prototype, simulate, and deploy autonomous systems with confidence. Whether it's in the sky, on the ground, or in orbit—if it's intelligent and mobile, Intrepid helps you build it. Learn more at intrepid.ai   References Website: finalspark.com Discord account:   / discord   Newsletter: https://finalspark.com/#newsletter   Topics: Biological computing • Neural engineering • Energy-efficient AI • Wetware vs hardware • The future of computation

Sanjoy Chowdhury reveals AI's hidden weakness: while systems can see objects and hear sounds perfectly, they can't reason across senses like humans do. His research at University of Maryland College Park, including the Meerkat model and AVTrustBench, exposes why AI recognizes worried faces and thunder separately but fails to connect them—and what this means for self-driving cars and medical AI.   Sponsors This episode is proudly sponsored by Amethix Technologies. At the intersection of ethics and engineering, Amethix creates AI systems that don’t just function—they adapt, learn, and serve. With a focus on dual-use innovation, Amethix is shaping a future where intelligent machines extend human capability, not replace it. Discover more at https://amethix.com   This episode is brought to you by Intrepid AI. From drones to satellites, Intrepid AI gives engineers and defense innovators the tools to prototype, simulate, and deploy autonomous systems with confidence. Whether it's in the sky, on the ground, or in orbit—if it's intelligent and mobile, Intrepid helps you build it. Learn more at intrepid.ai   Resources: The first audio-visual LLM with fine-grained understanding: Meerkat: Audio-Visual Large Language Model for Grounding in Space and Time (Accepted at ECCV 2024) Benchmark for evaluating the robustness to adversarial attacks, compositional reasoning: AVTrustBench: Assessing and Enhancing Reliability and Robustness in Audio-Visual LLMs (Accepted at ICCV 2025) First audio-visual reasoning evaluation benchmark and test time reasoning distillation pipeline AURELIA: Test-time Reasoning Distillation in Audio-Visual LLMs Accepted at ICCV 2025 For a detailed list of Sanjoy's work, please visit his webpage: https://schowdhury671.github.io/

This episode exposes the uncomfortable truth: most defense tech startups are just software engineers cosplaying as military innovators, creating fragmented solutions that Pentagon doesn't need. Not now, at least.   References War On The Rocks: https://warontherocks.com/2025/08/ukraine-isnt-the-model-for-winning-the-innovation-war/ LinkedIn: https://www.linkedin.com/in/jonasrsinger/ Spotify: https://tr.ee/Omy_1X8k1U Apple Podcast: https://podcasts.apple.com/us/podcast/defence-innovation-podcast/id1797131332 YouTube: https://youtube.com/@DefenceInnovationpodcast?si=cu2WlnVgL5XKnM0p   Sponsors This episode is proudly sponsored by Amethix Technologies. At the intersection of ethics and engineering, Amethix creates AI systems that don’t just function—they adapt, learn, and serve. With a focus on dual-use innovation, Amethix is shaping a future where intelligent machines extend human capability, not replace it. Discover more at https://amethix.com   This episode is brought to you by Intrepid AI. From drones to satellites, Intrepid AI gives engineers and defense innovators the tools to prototype, simulate, and deploy autonomous systems with confidence. Whether it's in the sky, on the ground, or in orbit—if it's intelligent and mobile, Intrepid helps you build it. Learn more at intrepid.ai   ✨ Connect with us! 📩 Newsletter: https://datascienceathome.substack.com 🎙 Podcast: Available on Spotify, Apple Podcasts, and more. 🐦 Twitter: @DataScienceAtHome 📘 LinkedIn: https://www.linkedin.com/in/fragadaleta/ Instagram: https://www.instagram.com/datascienceathome/