Freeplay’s Post

Building with LLMs? Learn from two leaders who have built, and importantly **deployed** AI products 🎉 Tons of wisdom from these two ⬇

Become a podcast editor ✅ Jokes aside, listening to this episode over and over again was a treat. Jeff Seibert and the team at Digits are doing incredible work with AI in accounting. Jeff's perspective on the hype around AI vs ML, running a lean engineering team that really digs into the row-level data, and why you shouldn't aim for 100% accuracy are refreshing and insightful. Whether you're building, designing, marketing or selling AI products, this episode is worth a listen. Ps - keep an eye for the next one 👀

🎙️ Our Founder & CEO Jeff Seibert sat down with Ian Cairns on his new podcast, Deployed, to cut through the noise and discuss what it's really like to build AI products in today's landscape. Learn more about how Digits is: ▶ Using AI to automate accounting and bookkeeping for startups ▶ The benefits of investing in row-level data review & dataset curation ▶ Running autonomous agents in production ▶ Solving LLM Text-to-SQL challenges Full episode out now 👉 https://lnkd.in/ggQKJ-KG

In Episode 100, we answer your questions: In your 100 episodes of the show, what is the most stand out moment for you in terms of AI beginning to redefine key aspects of people’s future of work?

Tired of TL;DR syndrome? Recast AI might just be your new literary lifesaver! This week at The Vision Debugged - AI Newsletter, we are diving into Recast AI, the game-changing platform that's turning your reading list into a personalized podcast station. Tired of TL;DR syndrome? Recast AI might just be your new literary lifesaver! Key features that make Recast AI stand out: 🎙️ Multi-voice audio summaries for a podcast-like experience 🧠 AI-powered smart summarization of lengthy articles 🤝 Community-driven content library for instant value 🎯 "For You" playlists with personalized content curation Curious how Recast AI is leveraging the Growth Loop framework to expand its user base? Want to see how it's creating a "Blue Ocean" in the crowded content consumption market? Check out the full article here: https://lnkd.in/g4wNy7yM

Highly recommend this book… for teams wanting to deploy models, and for companies looking for models to deploy.

Check out this episode of Alex Gluz's Revenue Engine Podcast featuring Mark McEachran! 🎙️ Dive into a discussion of the future of digital advertising involving the importance of first party data, machine learning, and more. Link in the comments! #digitaladvertsing #machinelearning #AI

Today's GeekWire Podcast with Juan M. Lavista Ferres of Microsoft's AI for Good Lab is based on the new book, "AI for Good: Applications in Sustainability, Humanitarian Action, and Health," due out next week, detailing the lab's work with its partners. I got to read an advance copy in advance of the conversation with Juan. I'd recommend the book for anyone interested in getting a better understanding of the inner workings of large-scale AI projects, including the limitations and long-term potential. • The book is well-organized and formatted so you can jump to the projects that interest you in sustainability, humanitarian action, and health. • Each chapter gives an executive summary of the project, and then goes deeper on the methods and implications. If you want to know more, there are references to the full research articles for additional information. • The first three chapters serve as a primer on AI. They filled in several gaps in my knowledge, things I didn't know I didn't know. There's so much abstract discussion about AI these days, the tangible examples were refreshing. As a non-computer scientist trying to improve my understanding of AI and inform my reporting, I found it worth the time. Book: https://lnkd.in/g4njVbD9 Podcast: https://lnkd.in/g7_dYUef

Pirates or Pioneers? AI’s Great Data Plunder As we the people drown in data, experts warn that within the next couple of years, AI could face a data drought. Models like GPT-4 have been trained on trillions of words, but the next generation will demand even more—quadrillions, quintillions—numbers so large the zeros disappear into the distance. But where will all this data come from? Who owns it? Who controls it? Welcome to this week’s episode of From Myths to Models to Madness. In this episode, our intrepid (and slightly annoying) AI hosts delve into the implications of AI’s insatiable appetite for data and what it means for the future development of AI technologies. Drawing on a diverse range of academic research they discuss, amongst other things, the impact on our data identities, privacy and consent, and copyright laws that increasingly look like they are no longer fit for purpose. If you don’t think this topic affects you, think again. Unless you live completely off-grid (you wouldn’t be reading this if you did) and don’t use any technology made in the last 30 years, you are already using AI and having your data used by AI. Please add a comment below if you have questions or themes you'd like explored. Or just add your thoughts! Click on the link below to access the podcast and from their you can also access your favourite listening platform by clicking the "listen on" button (we are on the usual suspects). If you like what you hear, please show your support by subscribing through your favourite platform.

Need a weekend pod? I got you. Check out the latest episode of the Talking AI Podcast with CEO of FeatureForm, Simba Khadder and HatchWorks AI CTO, Omar Shanti. Episode highlights in case you are on the fence: ▶ [00:01:53] Simba on What's Different About Generative AI Simba explains how prompts differentiate large language models (LLMs) from traditional transformers, highlighting that LLMs contain more ingrained knowledge and how chatbots utilize prompts for user interaction. ▶ [00:03:18] The Use Cases for Generative AI Simba emphasizes that the new use cases for LLMs, such as copilots and agents, have led to different teams and organizations being formed, especially in enterprises without traditional ML expertise. ▶ [00:04:43] Integration of ML and LLMs Omar and Simba discuss how ML and generative AI are fundamentally similar, built on the same underlying techniques, and how organizations should integrate both into their workflows instead of treating them as separate technologies. ▶ [00:07:52] LLMs as a Subset of ML The hosts debate whether LLMs are a subset of ML. Simba suggests that LLMs are just another tool, alongside other ML applications, to be applied to data. ▶ [00:09:18] Simba on Features in ML and LLMs Simba breaks down feature engineering and how traditional ML models and LLMs use similar concepts like embeddings to create value, although in different contexts. ▶ [00:13:16] RAG Misconceptions Simba critiques the current understanding of Retrieval Augmented Generation (RAG), arguing that current RAG workflows, which rely on simple nearest neighbor lookups in vector databases, will soon be seen as naive. He suggests more complex approaches using knowledge graphs and traditional search techniques will be superior. ▶ [00:19:43] Simba on ML vs. Gen AI Lifecycles Simba outlines the four stages of the ML lifecycle (data, training, deployment, and evaluation) and contrasts this with generative AI’s lifecycle, noting that generative AI can skip some stages like training but still has similar bottlenecks like data wrangling. ▶ [00:30:36] Challenges of ML Models in Production Simba and Omar discuss why many ML models fail to make it into production, attributing this to data scientists spending most of their time dealing with infrastructure and organizational challenges rather than model development. 🎧 Find it on Apple, Spotify, and YouTube 🎧