Garnet S. Heraman’s Post

I've been meaning to write about the Czech-Brazilian philosopher Vilem Flusser for ages because I find his concept of the "apparatus" really useful. I wrote this Frame after having watched the launch of ChatGPT-4o. As always, curious to hear your thoughts. https://lnkd.in/gVTuS5i7 PS: When talking about Flusser, I'd also recommend one of the weirdest books I've ever read and LOVED: "Vampyroteuthis Infernalis."

Here's my review of Stephen Wolfram's "What Is ChatGPT Doing ... and Why Does It Work?" This book delves deeply into the mechanics and principles of ChatGPT and Large Language Models (LLMs), expertly blending technical detail with broader implications in AI. Each chapter methodically explores key concepts like neural networks, machine learning, and the complexities of AI development. Essential for anyone interested in AI, it provides rich insights into both the technical intricacies and the philosophical dimensions of ChatGPT and similar models.

For all of our #msps friends in Florida at #itnation and throughout the country, here is a free tool you can use on the house. Just imagine what can be done when you proprietary information that your #msp can provide to your your teams internally but to your clients as well. This is not just a 'chatbot', this is embedded data within the framework of GPT-4 that has leveraged their vector search to make a more dynamic experience. The book is in the Library of Congress and this is the smartest AI book on this topic in the world. Check it out and bookmark the link! https://lnkd.in/gW2JG7w5 #ai #gpts #vectorsearch #embedded

ChatGPT and generative AI are generally untrustworthy...even for seemingly mundane tasks. I love the prospects, but they really don't seem to be ready for prime time. How can it not be able to distinguish Camus from Sartre? Next thing I know, it will be confusing Foucault with Nabokov. #chatgpt #chatgpt4 #ai #generatieveai #information #hype #hypecycle

Jeremy Khan shows how using ChatGPT and other LLMs can impact the way you think and you won’t even notice. It may strengthen the biases you already have by not challenging you enough or introduce new once intentionally or unintentionally. More in the excerpt below, but the whole book is worth reading.

Verity Fact Checker GPT 🚀 Dive into the world of accurate information with Verity Fact Checker GPT! Prioritize truth with our AI-powered tool that blends academic rigor, verified sources, and an engaging Jordan Peterson-inspired tone. 📚 🔍 Say goodbye to misinformation and hello to clarity! Try it FREE now at https://lnkd.in/gEngJXVv #FactCheckWithVerity #TruthSeeking #OpenAI #InformedWorld

We're less than a week away from our big kick-off of my 4 Books, 5 Days course! We're going to be digging into how to not only plan a series of 4 books, but how to use ChatGPT to make the magic of planning work so much faster! Click here to learn more and enroll 👉 https://lnkd.in/gAyDvqTU #Ai #Authors #Dreams #Series #publishingfaster #IndieAuthor #PrepTober #NaNoWriMo

For my friends who read Korean, I'm so delighted by this translation of my book on regexen and AI code assistants. It's really beautifully produced in the paper edition... nicer than the English version, I think. https://lnkd.in/g89ZqHkS #aiandml #regularexpressions #korean

A colleague sent me this article from The New Yorker that nicely frames how LLM have evolved and why, in some cases, they get things wrong. I've been doing a lot of work with having LLM help me with academic standards and benchmark alignment by playing one LLM off another (more on that later) and this article helped me think about why they are so good at tasks like this and so bad at seemingly simple tasks. I took the article and pasted into ChatGPT-4 and then asked it slightly altered math and poem questions that the writer mentioned to see if reading the article would help its "understanding" of the concepts. One thing I noticed yesterday when working with the standards alignment analysis, it actually displays an analysis as it is "reasoning". This happened today when I slightly altered the math question from the article, but keeping it in line with the original concept. Its response was correct. We've known that giving LLM examples of correct methods will train them, so I was kind of hoping this would happen. Just further reinforces how important we are when developing content that we can't just go with the default responses. I'm afraid that's what many will do without training and support. The poetry question was the same. In the spirit of the article, I asked, "Write a short poem in which the second line uses the same words as the first, but in reverse order." It wrote: Morning's light, embracing night, fades into dreams. Dreams into fades, night embracing, light morning's. Probably better than anything I wrote this week :) Hopefully these kinds of articles help us with new insights on how to interact better with these tools. https://lnkd.in/gQ2pGrxu