Deep Learning

From Tokens to Vectors: The Efficiency Hack That Could Save AI (Ep. 294)

LLMs generate text painfully slow, one low-info token at a time. Researchers just figured out how to compress 4 tokens into smart vectors & cut costs by 44%—with full code & proofs! Meanwhile OpenAI drops product ads, not papers. We explore CALM & why open science matters. 🔥📊
 
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
 

Why AI Researchers Are Suddenly Obsessed With Whirlpools (Ep. 293)

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/
 

Your Favorite AI Startup is Probably Bullshit (Ep. 287)

The brutal truth about why Silicon Valley is blowing billions on glorified autocomplete while pretending it's the next iPhone.
We're diving deep into the AI investment circus where VCs who can't code are funding companies that barely understand their own technology. From blockchain déjà vu to the "ChatGPT wrapper" economy—this episode will make you question every AI valuation you've ever seen.
Fair warning: We're naming names and calling out the hype. Don't listen if you work at a "revolutionary AI startup" that's just OpenAI's API with a pretty interface.
 
#AIBubble #VentureCapital #TechReality #StartupBullshit
 

Brains in the Machine: The Rise of Neuromorphic Computing (Ep. 285)

In this episode of Data Science at Home, we explore the fascinating world of neuromorphic computing — a brain-inspired approach to computation that could reshape the future of AI and robotics. The episode breaks down how neuromorphic systems differ from conventional AI architectures like transformers and LLMs, diving into spiking neural networks (SNNs), their benefits in energy efficiency and real-time processing, and their limitations in training and scalability. Real-world applications are highlighted, including low-power drones, hearing aids, and event-based cameras. Francesco closes with a vision of hybrid systems where neuromorphic chips and LLMs coexist, blending biological inspiration with modern AI.
📚 References
SpikingJelly: https://github.com/fangwei123456/spikingjelly
Norse: https://github.com/norse/norse
IBM TrueNorth: https://research.ibm.com/blog/brain-inspired-chip
Intel Loihi 2: https://www.intel.com/content/www/us/en/research/neuromorphic-computing.html
SpiNNaker: https://apt.cs.manchester.ac.uk/projects/SpiNNaker/
BioRobotics Institute: https://www.santannapisa.it/en/institute/biorobotics
🎙️ Sponsors
MillionaireMatch — The elite platform for high-achieving singles🌐 https://www.millionairematch.com
AGNTCY — The open source collective building the Internet of Agents🌐 https://www.agntcy.org

DSH/Warcoded Kill Chains and Algorithmic Warfare – Autonomy in Targeting and Engagement (Ep. 282)

In this gripping follow-up, we dive into how AI is transforming kinetic operations—from identifying a threat to executing a strike.
🔍 Highlights from this episode:
How AI compresses the OODA loop (Observe, Orient, Decide, Act)
The spectrum of autonomy: human-on-the-loop vs. human-out-of-the-loop
Real-world systems like loitering munitions (Switchblade, Harpy) and Selective Ground Response AI (SGR-AI)
The ethical and legal dimensions of delegating lethal decisions to machines
 
As the lines blur between algorithm and operator, we explore who—or what—is pulling the trigger.
 
 
Sponsors
Warcoded 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 amethix.com
 
Warcoded 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
 
 
#Warcoded #DataScienceAtHome #AI #AutonomousWeapons #MilTech #DefenseTech #KillChain #OODAloop #LAWs #EdgeAI #Podcast

DSH/Warcoded: Eyes and Ears of the Machine – AI Reconnaissance and Surveillance (Ep. 281)

Welcome to DSH/Warcoded
We explore how AI is transforming ISR (Intelligence, Surveillance, Reconnaissance)—from satellite imagery to drone feeds. In this episode:
🔍 Computer vision for target ID📡 Predictive surveillance & pattern-of-life modeling🧠 LLMs for SIGINT & OSINT intelligence briefings🌍 Real-world examples: Ukraine, Gaza & more
Listen now and see how machines are learning to see, predict, and inform at the edge of modern conflict.
 
 
Sponsors
Warcoded 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 amethix.com
 
Warcoded 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
 
#AI #defensetech #ISR #LLM #Warcoded #DataScienceAtHome #OSINT #SIGINT #dronewarfare

Humans vs. Bots: Are You Talking to a Machine Right Now? (Ep. 273)

In this episode of Data Science at Home, host Francesco Gadaleta dives deep into the evolving world of AI-generated content detection with experts Souradip Chakraborty, Ph.D. grad student at the University of Maryland, and Amrit Singh Bedi, CS faculty at the University of Central Florida. 
Together, they explore the growing importance of distinguishing human-written from AI-generated text, discussing real-world examples from social media to news. How reliable are current detection tools like DetectGPT? What are the ethical and technical challenges ahead as AI continues to advance? And is the balance between innovation and regulation tipping in the right direction? 
 
Tune in for insights on the future of AI text detection and the broader implications for media, academia, and policy.
 
Chapters 
 
00:00 - Intro 
00:23 - Guests: Souradip Chakraborty and Amrit Singh Bedi 
01:25 - Distinguish Text Generation By AI 
04:33 - Research on Safety and Alignment of Generative Model 
06:01 - Tools to Detect Generated AI Text  
11:28 - Water Marking
18:27 - Challenges in Detecting Large Documents Generated by AI 
23:34 - Number of Tokens 
26:22 - Adversarial Attack
29:01 - True Positive and False Positive of Detectors 
31:01 - Limit of Technologies 
41:01 - Future of AI Detection Techniques 
46:04 - Closing Thought
 
Subscribe to our new YouTube channel https://www.youtube.com/@DataScienceatHome
 

AI Says It Can Compress Better Than FLAC?! Hold My Entropy 🍿 (Ep. 268)

Can AI really out-compress PNG and FLAC? 🤔 Or is it just another overhyped tech myth? In this episode of Data Science at Home, Frag dives deep into the wild claims that Large Language Models (LLMs) like Chinchilla 70B are beating traditional lossless compression algorithms. 🧠💥
But before you toss out your FLAC collection, let's break down Shannon's Source Coding Theorem and why entropy sets the ultimate limit on lossless compression.
We explore: ⚙️ How LLMs leverage probabilistic patterns for compression 📉 Why compression efficiency doesn’t equal general intelligence 🚀 The practical (and ridiculous) challenges of using AI for compression 💡 Can AI actually BREAK Shannon’s limit—or is it just an illusion?
If you love AI, algorithms, or just enjoy some good old myth-busting, this one’s for you. Don't forget to hit subscribe for more no-nonsense takes on AI, and join the conversation on Discord!
Let’s decode the truth together. Join the discussion on the new Discord channel of the podcast https://discord.gg/4UNKGf3
 
Don't forget to subscribe to our new YouTube channel 
https://www.youtube.com/@DataScienceatHome
 
 
References
Have you met Shannon? https://datascienceathome.com/have-you-met-shannon-conversation-with-jimmy-soni-and-rob-goodman-about-one-of-the-greatest-minds-in-history/
 
 

What Big Tech Isn’t Telling You About AI (Ep. 267)

Are AI giants really building trustworthy systems? A groundbreaking transparency report by Stanford, MIT, and Princeton says no. In this episode, we expose the shocking lack of transparency in AI development and how it impacts bias, safety, and trust in the technology. We’ll break down Gary Marcus’s demands for more openness and what consumers should know about the AI products shaping their lives.
 
Check our new YouTube channel https://www.youtube.com/@DataScienceatHome and Subscribe! 
 
Cool links
https://mitpress.mit.edu/9780262551069/taming-silicon-valley/
http://garymarcus.com/index.html

Kaggle Kommando’s Data Disco: Laughing our Way Through AI Trends (Ep. 265) [RB]

In this episode, join me and the Kaggle Grand Master, Konrad Banachewicz, for a hilarious journey into the zany world of data science trends. From algorithm acrobatics to AI, creativity, Hollywood movies, and music, we just can't get enough. It's the typical episode with a dose of nerdy comedy you didn't know you needed. Buckle up, it's a data disco, and we're breaking down the binary!
 
Sponsors
Intrepid AI is an AI assisted all-in-one platform for robotics teams. Build robotics applications in minutes, not months.
Learn what the new year holds for ransomware as a service, Active Directory, artificial intelligence and more when you download the 2024 Arctic Wolf Labs Predictions Report today at arcticwolf.com/datascience
 
🔗 Links Mentioned in the Episode:
Generative AI for time series: TimeGPT Documentation
Lag-llama: GitHub (Note: The benchmark results on this one are pretty horrible)
Open source LLM: Olmo Blog Post
Quantization for LLM: Hugging Face Guide
And finally, don't miss Konrad's Substack for more nerdy goodness! (If you're there already, be there again! 😄)