The Prompt Desk

Note! This is an episode that we found in our archive unreleased. It is one of the earliest episodes we ever recorded, but has so far gone not been listened to. It's not quite the same style as our current episodes, but we thought we'd release it anyway as a special treat for interested listeners!

The podcast episode explores the realm of large language models, prompt engineering, and best practices, with a emphasis on discu...

In this insightful episode, hosts Bradley and Justin tackle the challenge of managing expectations when using large language models (LLMs) and AI agents for tasks like content creation and website interactions. They dispel the common misconception that these systems can produce perfect outputs right out of the gate without any human guidance.

Instead, the hosts emphasize the importance of breaking down complex workflows into...

In this episode, we discuss, how we might protect prompt-based applications and LLMs from prompt injection. We discuss how data validation was done in the 1960s and modern libraries and techniques that can successfully act as a first line of defense against prompt injection. We touch on the idea that using other types of models, such as decision trees, conventional NLP pipelines, embedding models, or neural networks trained on data...

If you are anything like us, for 36 hours your Twitter and LinkedIn feeds were set ablaze by claims that Llama 3 is the greatest LLM since sliced bread. But how do we know that is actually the case? How do we separate the hype and noise from the reality? On what basis can we credibly claim that one LLM is better than another?

On this episode of The Prompt Desk, Justin Macorin and I brought on LLM expert Maxime Labonne, who h...

As AI language models become increasingly powerful, more and more people are discovering the potential of well-crafted prompts to boost productivity and solve complex problems. Justin and Brad share their expertise on designing effective prompts that adhere to rigorous engineering principles. They highlight real-world examples of prompts that have yielded impressive results and provide valuable insights into the art and science of ...

What is LLM and ML Ops? How has the field of MLOps changed with the introduction of LLM models? What is the emerging field of Prompt Ops? How should we measure the effectiveness of our models in production? To answer these questions, we invited MLOps expert Andrew McMahon to chat.

Check out Andrew's LinkedIn, Twitter, and book using these links:
Linkedin - https://www.linkedin.com/in/andrew-p-mcmahon/
Twitter/X - @ele...

Dive into the exciting world of prompt experimentation and uncover the secrets to crafting the perfect prompts! In this episode, we share our tried-and-true techniques for iterative prompt engineering, from running prompts on realistic inputs to making creative modifications that can transform a prompt's performance. We also explore the tantalizing possibilities of using datasets to generate prompts and the challenges of integratin...

Prompt Injections might be the single biggest risk to the deployment of LLM systems. In this episode of The Prompt Desk, your host Bradley Arsenault sits down with researcher Jonas Geiping to talk about his research on LLM security.

Jonas and his colleagues have found optimization techniques that allow them to find prompts that they can put into any LLM model that can make it do anything. The open-ended nature of the AI secu...

Explore the six essential components that define an AI agent in 2024, including input/output, memory, loops and recursion, third-party actions, and comprehension. Join Brad and Justin as they discuss the criteria needed for a sophisticated AI system to be considered an agent, and how these requirements contribute to realizing the sci-fi vision of AI depicted in movies.

—

Continue listening to The Prompt Desk Podcast for everything LL...

Are you wondering how LLM Agents can be applied in large enterprises? Are we all going to become managers to teams of AI systems, cobbled together like a rag-tag group of robot butlers? In the latest episode of The Prompt Desk, hosts Justin Macorin and Bradley Arsenault sit down with Alan Wunsche to talk about the deployment and rollout of AI agents in enterprise business environments.

Check out Alans' Book Agentware here: https...

In this episode of The Prompt Desk, hosts Justin Macorin and Bradley Arsenault discuss the various ways they have been using coding assistants, such as Github Copilot, in order to boost their productivity as software developers. The two exchange ideas, thoughts and techniques on how to better use these tools.

—

Continue listening to The Prompt Desk Podcast for everything LLM & GPT, Prompt Engineering, Generative AI, and LLM Secur...

Join us with Sachin Kamath as we explore how to use AI in creative projects effectively and how we may effectively integrate AI into existing workflows, significantly enhancing productivity and creative output. We discuss frameworks for advanced prompting, ethical considerations, and best practices in AI.

Consistent Character GPT:

https://chat.openai.com/g/g-a9JivI0y2-consistent-character-gpt-fast-high-quality

AI Bootcamp for ...

Does Voodo in prompt engineering exist? The evolution from manual prompt crafting to automated refinement and optimization in AI interactions marks a transition from art to science. Let's discuss how initial natural language inputs transform, improving communication with large language models for more precise and effective outputs. Join us as we explore the fascinating process of refining prompts for AI, highlighting the blend of c...

This episode discusses improving code generation with LLMs by dynamically extracting functions and context, using techniques like Python's inspect modules and JavaScript's toString. The goal is to produce code that integrates well with existing projects, enhancing customization and practicality.

—

Continue listening to The Prompt Desk Podcast for everything LLM & GPT, Prompt Engineering, Generative AI, and LLM Security.

Check out ...

Join us as we explore the advancements and challenges in implementing Retriever-Augmented Generation (RAG) systems for document retrieval and summarization, examining selective information inclusion, re-ranking strategies, and the importance of document parsing and query pre-processing. We discuss with Denys Linkov the practical applications, benefits, and considerations of RAG-based solutions in improving information retrieval acc...

Don't miss out on our discussion about the critical factors of expert domain knowledge, intellectual property, and the impact of first-mover advantage. These factors remain the most important competitive advantages, even in today's rapidly evolving technological landscape.

—

Continue listening to The Prompt Desk Podcast for everything LLM & GPT, Prompt Engineering, Generative AI, and LLM Security.

Check out PromptDesk.ai for an op...

In today's world, application security is more important than ever before. One of the most significant risks to application security is prompt injection. It can cause irreparable damage to your business. That's why it's critical to understand the various injection techniques, their impact on your business, and how you can mitigate the risks.

—

Continue listening to The Prompt Desk Podcast for everything LLM & GPT, Prompt Engineer...

Let's discover the rise of local Small Language Models (SLMs), contrasting them with their larger, internet-dependent counterparts - LLMs. We discuss real-world applications like Google Maps, in-browser neural networks, and Apple Photos, examining the benefits of local installation for speed, reliability, privacy, and cost, and address the challenges and opportunities for engineers in this evolving landscape.

—

Continue listening to ...

In this episode, we interview Charles H Martin about his open-source Weight Watcher project ( found here https://weightwatcher.ai/ ), which provides ways to test the quality and fit of deep neural networks without having to rely upon a validation dataset. Given the scarcity of high-quality data and the complexity of modern multi-stage ML training and deployment pipelines, this technique could prove to be extremely valuable to any A...