Alexander Chukovski’s Post

A quick story and tour of the new MotorDoc EMPOWER motor current signature analyzer (EMPOWER - Electric Motor Performance Optimization Waveform Evaluation and Recognition). Extremely portable hardware that works with any laptop or pc at an extremely affordable price. First units ordered in October, 2024 ship on November 1, 2024. It all started with a bet that we could not design a powerful analytic tool using generative AI and Python... well, here it is. We do also discuss the lessons learned, which we will cover in detail in our next podcast. Note: yes, we are open to distributors in the USA and abroad. There are requirements that are mentioned in the podcast. #motordocllc #motordocai #ai #ml #ESA #MCSA #EMPOWER #ECMSE1 #EMPATH #motortesting

It's unreal to be able to speak to folks like the guests I hosted on the SuperDataScience Podcast last month. ICYMI, today's episode highlights the most riveting moments from October. The specific conversation highlights included in today's episode are: 1. UC San Diego neuroscience professor Dr. Bradley Voytek on how data science facilitates breakthroughs in our understanding of the brain. 2. Eloquent Natalie Monbiot on how lifelike, digital versions of ourselves can scale up our public-facing work. 3. Lightning AI CTO Dr. Luca Antiga on where he sees generative A.I. being most useful in our professional lives. 4. Gable CEO Chad Sanderson on how, when we work with data, we always need to think about how downstream users might come to interpret our data... which is why he finds data contracts so important that he's writing an O'Reilly book about it. 5. Polars CEO Ritchie Vink on the incredible specs (e.g., efficiency speedups) of his open-source DataFrame-operations library for Python. Check out today's episode (#834) to hear all these eye-opening conversations. The "Super Data Science Podcast with Jon Krohn" is available on all major podcasting platforms and a video version is on YouTube. #superdatascience #datascience #machinelearning #ai #podcast

Coming soon - The MLSecOps Podcast S2Ep7: 🎙️"ReDoS Vulnerability Reports: Security Relevance vs. Noisy Nuisance" Join us for this episode, inspired by research from the AI/ML bug hunting community, huntr, and a compelling blog post (link in comments) by William Woodruff, Engineering Director at Trail of Bits. William joins the conversation to delve into various topics such as: - The practical utility of ReDoS (regular expression denial of service) reports in security - A high level summary of code reviews - Exploring how LLMs (Large Language Models) could revolutionize code analysis - Assessing the strengths and weaknesses of the CVSS (Common Vulnerability Scoring System) - And much more! Be the first to know when the full episode is released. Subscribe here! ➡️https://bit.ly/3wFoRlB #MLSecOps #aisecurity #MLSec #huntr #threatresearch #ReDoS #llm #ai #podcast #machinelearning #ProtectAI

🧠 In this snippet from Season 5, we explore the intriguing aspects of artificial intelligence and its mimicry of human behavior. 🤔 🎙️ Ready for a deeper understanding? Watch the full episode! https://bit.ly/PSMSpodcast #ChatGPTInsights #PodcastWisdom #PublicSectorMarketingShow

I just finished reading "Everything Is Predictable: How Bayesian Statistics Explain Our World" by Tom Chivers. If you are interested in #machinelearning, I highly recommend adding this to your library. Bayesian techniques are used in machine learning for classification and regression problems, particularly in the development of spam filters, recommendation systems, and in Bayesian networks for probabilistic inference. While you are at it, also check out Chris Voss's podcast on the subject (links in comments).

🧠 In this snippet from Season 5, we explore the intriguing aspects of artificial intelligence and its mimicry of human behavior. 🤔 🎙️ Ready for a deeper understanding? Watch the full episode! https://bit.ly/PSMSpodcast #ChatGPTInsights #PodcastWisdom #PublicSectorMarketingShow

“As important as data is, even as we improve the power of ‘artificial intelligence,’ it means that human beings’ own critical thinking skills become more important” Hear more from Michael Schrage of @mitide on #WDAYPodcasts. #TeamWDAY

“As important as data is, even as we improve the power of ‘artificial intelligence,’ it means that human beings’ own critical thinking skills become more important” Hear more from Michael Schrage of @mitide on #WDAYPodcasts. #TeamWDAY

Are you using LLM's but struggling to get the prompt right? Does the model just refuse to do what you want it to? In the latest episode of The Prompt Desk, the talented Justin Macorin and I discuss our processes for experimenting with our prompts. We cover everything from iterative development, the importance of running the prompt against multiple inputs, and the tantalizing possibility of using datasets to automatically construct our prompts. What we used to do at training, we now do at inference! Get the latest episode of The Prompt Desk in your favorite podcast app, including Spotify, Apple Podcasts, and my personal favorite, Pocket Casts.

I recently tuned into an excellent episode of the Latent Space Podcast featuring Alistair Pullen from Cosine. The topic? Chain-based fine-tuning. This isn’t your typical domain-specific fine-tuning. It’s about teaching models to think, plan, and execute through structured reasoning chains. One standout example: Cosine fine-tuned GPT-4 to reason like a software engineer—not just on domain-specific data, but on structured thought processes. The innovation? Their data pipeline generates human-like reasoning traces, enabling models to tackle tasks with newfound sophistication. This approach aligns closely with OpenAI’s O1 reasoning model, which integrates reasoning chains directly into the model. It’s sparked some big questions for me: 🔹 How could reasoning-based fine-tuning reshape other domains? 🔹 Are we shifting from “what models know” to “how models think”? I’m eager to see how these methods evolve—especially with open-source alternatives in the mix.