Two Voice Devs

Allen Firstenberg and Linda Lawton dive deep into the power of Google's Imagen 3 Editing API. Discover how to effortlessly edit and enhance images, opening up a world of creative possibilities for developers!

* Learn how the In-Painting/In-filling feature can quickly remove wires from an image, add highlights, and correct shading on images that the AI generated, and more.

* Explore how to create your own 3D-printed objects from scratch using AI.

* Discover how you can reference images to put models or products into a specific scene.

* Learn how to use the Out-Painting feature to extend images beyond their original boundaries, transforming portraits into landscapes and beyond.

Also, be prepared for some unexpected and hilarious AI hallucinations along the way as Allen tries to zoom out from an image multiple times! Plus, the duo discusses the ethical implications of AI-generated content and how creatives can leverage these tools to enhance their own artwork.

Don't miss this exciting exploration of Imagen 3 and its potential to revolutionize image manipulation for developers and creators alike!

Timestamps:

00:00:00 Introduction

00:00:55 Imagen 3 Editing API

00:04:36 In-Painting/In-Filling

00:04:52 Generating 3D Models

00:09:00 Vertex AI Studio

00:10:15 Imagen and Gemini Together

00:13:14 Generating Images with Reference Images

00:20:11 Out-Painting

00:31:00 Ethical Implications

#Imagen3 #AI #ImageEditing #GoogleAI #VertexAI #VertexAISprint #MachineLearning #DeveloperTools #GenerativeAI #GenAI #3DPrinting #AIArt

Are you building AI models and systems? Then you need to understand AI ethics! In this episode of Two Voice Devs, Allen Firstenberg welcomes Parul, a Senior Production Engineer at Meta, to dive deep into the world of AI ethics. Learn why fairness and bias are critical considerations for developers, and discover practical techniques to mitigate bias in your AI systems.

Parul shares her experiences and passion for AI ethics, detailing how biases in training data and system design can lead to unfair or even harmful outcomes. This episode provides concrete examples, actionable advice, and valuable resources for developers who want to build more ethical and equitable AI.

More Info:

* Fairlearn: https://fairlearn.org/

* AIF360: https://aif360.readthedocs.io/en/stable/

* what-if tool: https://pair-code.github.io/what-if-tool/

Timestamps:

00:00:00 Introduction

00:00:20 Guest Introduction: Parul, Meta

00:02:22 What is AI Ethics?

00:06:13 Why is AI Ethics Important?

00:08:15 AI Systems vs. AI Models

00:09:52 Examples of Bias in AI Systems

00:12:23 Minimizing Biases: Developer Responsibility

00:14:53 Tips for Minimizing Unfairness and Biases

00:19:40 Fairness Constraints: Demographic Parity

00:23:17 The Bigger Picture: Roles & Responsibilities

00:29:23 Monitoring: Bias Benchmarks

00:32:00 Open Source Frameworks for AI Ethics

00:34:02 Call to Action & Closing

#AIethics #Fairness #Bias #MachineLearning #ArtificialIntelligence #Developers #OpenSource #EthicalAI #TwoVoiceDevs #TechPodcast #DataScience #AIdevelopment

Are you overwhelmed by the sheer number of Large Language Models (LLMs) available? Choosing the right LLM for your project isn't about picking the most popular one – it's about understanding your specific needs and rigorously evaluating your options.

In this episode of Two Voice Devs, Allen Firstenberg and guest host Brad Nemer, a seasoned product manager, dive deep into the world of LLM evaluation. They go beyond the marketing buzz and explore practical tools and strategies for making informed decisions.

Whether you're a developer, a product manager, or just curious about the practical applications of LLMs, this episode provides invaluable insights into making the right choices for your projects. Don't get caught up in the hype – learn how to evaluate LLMs effectively!

More Info:

https://www.udacity.com/blog/2025/01/how-to-choose-the-right-ai-model-for-your-product.html

[00:00:00] Introduction: Meet Brad Niemer

[00:00:38] Brad's Journey to Product Management & AI

[00:03:12] Collaboration with Noble Ackerson and the LLM Evaluation Challenge

[00:05:23] The Role of a Product Manager.

[00:07:43] Product manager relation to engineering.

[00:13:46] Exploring Evaluation Tools: Hugging Face

[00:16:58] Exploring Evaluation Tools: Chatbot Arena (Human Evaluation)

[00:20:30] Chatbot Arena: Code Generation Evaluation

[00:24:43] Evaluating LLMs: Beyond Chatbots and Truth

[00:26:11] Exploring Evaluation Tools: Artificial Analysis (Quality, Speed, Price)

[00:28:47] Exploring Evaluation Tools: Galileo (Hallucination Report)

[00:31:16] Case Study: DeepSeek and the Importance of Contextual Evaluation

[00:34:53] The Future of LLM Testing and Quality Assurance

[00:37:49] Wrap Up contact information.

#LLM #LargeLanguageModels #AIEvaluation #ProductManagement #TechTalk #TwoVoiceDevs #HuggingFace #GenAI #GenerativeAI #ChatbotArena #ArtificialAnalysis #Galileo #DeepSeek #ChatGPT #Gemini #Mistral #Claude #ModelSelection #AIdevelopment #SoftwareDevelopment #Testing #QA #RAG #MachineLearning #NLP #Coding #TechPodcast #YouTubeTech #Developers

Google's white paper on AI Agents has sparked debate – are they truly the next leap in AI, or just large language models dressed up with new terminology? Join Allen and Mark of Two Voice Devs as they dive into the details, exploring the potential of Google's framework while also critically examining its shortcomings. They analyze the core components of agents – models, tools, and orchestration – highlighting the value of defining tools as capable of taking actions. But they also raise key questions about the blurry line between models and agents, the confusing definitions of extensions and functions, and the critical omission of authentication and identity considerations. This episode is a balanced take on a fascinating and complex topic, offering developers valuable insights into the evolution of AI systems.

Key Moments:

[00:00:20] The core definition of agents: A promising start, or too broad?

[00:05:08] Model vs. Orchestration: Understanding the decision-making layers.

[00:17:33] "Tools" Unpacked: Exploring actions, extensions, and functions

[00:25:14] The crucial gap: Authentication, Identity, and User context.

[00:29:36] Reasoning techniques: React, Chain, and Tree of Thought explained.

[00:35:41] The model-agent debate: Where is the boundary line?

[00:37:45] Setting the stage for Gemini 2.0?

[00:39:06] A valuable discussion starter, but with room to grow.

Hashtags:

#AIAgents #GoogleAI #LLM #GenerativeAI #AIInnovation #TechAnalysis #TwoVoiceDevs #AIDevelopment #AIArchitecture #SoftwareEngineering #DeveloperPodcast #GeminiAI #MachineLearning #DeepLearning #AITools #Authentication #TechDiscussion #BalancedTech

Join Allen Firstenberg as he welcomes Lee Mallon, a first-time guest host, for an in-depth discussion about the future of development, user experiences, and the exciting potential of AI-driven personalization! Lee shares his journey from coding on a Toshiba MX 128k to becoming CTO of Yalpop, a company reinventing learning through personalized experiences. This isn't just another AI hype-cast – it's a deep dive into how we can shift our mindset to truly put users at the center of our development process, leveraging new tech to create delightful and efficient experiences.

Lee and Allen discuss everything from the limitations of current recommendation engines to the emerging potential of AI agents and just-in-time interfaces. This is a must-watch for any developer looking to stay ahead of the curve and build truly impactful applications.

#AI #ArtificialIntelligence #GenAI #GenerativeAI #Personalization #UserExperience #UX #Development #WebDev #FutureOfTech #LLMs #LargeLanguageModels #AIagents #MachineLearning #SoftwareDevelopment #Programming #WebDevelopment #TwoVoiceDevs #Podcast #TechPodcast #Innovation #Code #Coding #Developer #TechTrends #UserCentricDesign #Web4 #NoCode #LowCode #DigitalTransformation

Hold onto your keyboards, folks! AI is shaking up the software engineering world, and in this electrifying episode of Two Voice Devs, Allen and Mark are diving headfirst into the chaos. We're not just talking about the theory – we're getting real about how AI coding tools are actually impacting developers right now. Is this the end of coding as we know it, or the dawn of a new era of software creation?

More Info:

* https://newsletter.pragmaticengineer.com/p/how-ai-will-change-software-engineering

* https://addyo.substack.com/p/the-70-problem-hard-truths-about

[00:00:00] Introduction: Meet Allen and Mark and hear about their busy start to the year.

[00:00:39] The Trigger: Discover the article from The Pragmatic Engineer that sparked this conversation about the role of AI in software engineering.

[00:02:16] Addressing the Panic: We discuss the common fear: is AI going to steal developer jobs?

[00:03:34] Key Article Points: Allen breaks down the seven key areas of the article: how developers are using AI, the "70% Problem," and more.

[00:04:43] Design Patterns & Craftsmanship: Mark discusses how AI-driven development relates to established software patterns and developer craftsmanship.

[00:07:44] The Knowledge Paradox: Unpack the key difference in how senior and junior developers use AI and the potential issues it raises.

[00:10:06] AI vs. Stack Overflow: We explore the differences between getting code from AI and from community platforms like Stack Overflow.

[00:12:49] Personal Experiences: Allen and Mark share how they're actually using AI tools in their coding workflows.

[00:17:09] AI Usage Patterns: Discussing the "constant conversation", "trust but verify", and "AI first draft" patterns.

[00:20:55] The 70% Problem Revisited: Is AI just getting us part way there?

[00:23:24] AI as a Team Member: Exploring the idea of AI as a pair programming partner and whether it's actually helping.

[00:24:41] Trusting your Experience: the importance of listening to the gut feeling of an experienced developer when AI-generated code "feels" wrong.

[00:26:06] Programming Languages are Easy for AI: The simplicity and consistency of programing grammars.

[00:27:47] Is English the New Programming Language?: We debate the idea that natural language is becoming as important as coding and discuss what "programming" really means.

[00:30:36] The Problem with Trying to Make Programming Easy: Historical attempts to make programming easier are revisited.

[00:32:37] Programming vs the Rest of the Job: The core job of a software developer is more than just programming and writing code.

[00:37:21] Quality & Craftsmanship in the Age of AI: We explore what will make software stand out in the future and how crafting great software still matters.

[00:40:27] AI for Personal Software: Could AI drive a renaissance in personal software, similar to the spreadsheet?

[00:42:53] The Importance of AI Literacy: Mastering AI development is the new skill to make developers even more valuable.

[00:43:47] Closing Thoughts: The essential skills of developers remain crucial as we move into the future of AI driven coding.

[00:44:59] Call to Action: We encourage you to join the conversation and share your thoughts on AI and software development.

This isn't just another tech discussion – it's a high-stakes debate about the so

ul of software engineering. Will AI become our greatest ally, or our ultimate replacement? Tune in to find out!

#AIApocalypse #CodeRevolution #SoftwareEngineering #ArtificialIntelligence #Coding #Programming #Developers #TechPodcast #TwoVoiceDevs #MachineLearning #AICoding #FutureofCode #TechDebate #DeveloperSkills #CodeCraft #AIvsHuman #CodeNewbie #SeniorDev #JuniorDev #TechTrends

Join Mark and Allen, your favorite Two Voice Devs, as they explore the exciting (and sometimes frustrating!) world of Gemini 2.0's search grounding capabilities and how to use it with LangChainJS! Allen shares his recent holiday project: a deep dive into Google's latest AI tools, including the revamped search grounding feature, and how he made it work seamlessly across Gemini 1.5 and 2.0. We'll show you the code and demonstrate the differences between using search grounding and not, using real-world examples. Learn how to build your own powerful, grounded AI applications and stay ahead of the curve in the rapidly changing AI landscape!

In this episode, you'll discover:

[00:00:00] Introduction to Two Voice Devs and what we've been up to

[00:00:24] Allen discusses tackling bug fixes and updates with Gemini 2.0 and LangChain

[00:00:51] The new Gemini 2.0 Search Grounding Tool: what's new? What does it mean to be "agentic"?

[00:02:13] Allen dives into the Google Search Tool, understanding the differences between 1.5 and 2.0, and building a layer for easy use in LangChain

[00:03:06] Allen walks us through the code! The magic of setting up a model with or without search capabilities in LangChainJS

[00:04:48] Using output parsers and annotating your results in LangChainJS

[00:05:53] Similarities between Perplexity's results, and how LangChainJS handles output

[00:06:46] Running the same query with and without grounding, and the dramatic difference in the response (Who won the Nobel Prize for Physics in 2024?)

[00:08:26] A closer look at how LangChainJS presents its source references and how to use them in your projects.

[00:12:55] Taking advantage of tools that Google is providing

[00:13:20] The goal of keeping backward compatibility for developers

[00:15:39] Exploring how this is a version of RAG and how that compares to using external data sources

[00:16:50] What are data sources in VertexAI and how they relate?

[00:19:14] What is the cost? How is Google pricing the search capability?

[00:20:59] More to come soon from Allen with LangChainJS!

Don't miss this deep dive into cutting-edge AI development! Like, subscribe, and share if you find this information helpful!

#Gemini #LangChain #LangChainJS #AI #ArtificialIntelligence #GoogleAI #VertexAI #SearchGrounding #RAG #RetrievalAugmentedGeneration #LLM #LargeLanguageModels #OpenSource #TwoVoiceDevs #Programming #Coding #GoogleSearch #DataScience #MachineLearning #Innovation #TechPodcast #TechVideo

Happy New Year from Two Voice Devs! Join Allen Firstenberg and Mark Tucker as they recap the whirlwind that was 2024 in AI and tech, and make some bold predictions for 2025. We're diving deep into the biggest players, the most exciting innovations, and the developer challenges that lie ahead. From OpenAI's O3 and Google's Gemini 2.0 to the rise of Anthropic and the resurgence of head-mounted wearables, we preview the stories we'll be talking about this year. Plus, we discuss the important questions around the cost, both financial and environmental, of this new AI landscape.

This episode is packed with insights for any developer looking to navigate the rapidly evolving world of artificial intelligence. Don't miss our discussion on what "agents" actually mean, and what the future holds for voice assistants.

Timestamps:

[00:00:00] Intro & Happy New Year! - Welcome to Two Voice Devs and a fun fact to kick things off!

[00:01:14] Looking Back at 2024: A recap of the biggest AI movers of the past year.

[00:01:32] OpenAI & Google Dominate: Analyzing the impact of OpenAI and Google's announcements, including O3, Gemini 2.0 Thinking, Sora, Veo, and Imagen.

[00:04:02] OpenAI's Internal Turmoil and Google's Notebook LM: A look into the organizational chaos at OpenAI and the impressive upgrade of Notebook LM using Gemini 2.0.

[00:05:24] Apple Intelligence, Amazon & the Catch-Up Game: Discussions around the progress of Apple and the challenges Amazon is facing, along with Anthropic.

[00:08:15] Meta's LLAMA Models and Ray Bans: Exploring the surprising impact of

Meta's AI models and the resurgence of head-mounted wearables.

[00:10:04] Developer Realities: Fine-Tuned Models & DevOps: Mark discusses the importance of smaller, fine-tuned models and DevOps practices for language models.

[00:11:54] The Environmental & Ethical Concerns of AI: A critical discussion about the environmental impact, ethical concerns, and privacy considerations of large language models.

[00:13:15] Allen's 2024 Contributions: Langchain.js and GDE Presentations: Allen shares his work with open-source projects, LangChain, and traveling as a GDE

[00:16:56] 2025: Predictions and "Agents": The focus is on the emergence of "agents" and the uncertainty surrounding their definition.

[00:19:18] Defining "Agents": Allen lays out his predictions of what makes an agent.

[00:20:58] The Resurgence of Voice Assistants: Discussing the future of voice assistants and the potential revival with the emergence of new technologies.

[00:23:59] Google's Project Astra and Android XR: Exploring the new integrations in the voice and AI spaces.

[00:24:46] Home Assistant: An Open Source Alternative: A deep dive into this lesser discussed project and it's voice hardware offering.

[00:27:09] Amazon's Catch-Up: Is Amazon ready to get back into the AI and voice assistant game?

[00:28:01] Looking forward into the future of LLMs: Predictions on LLMs and where they're going.

[00:29:20] Outro: Thank you for joining Two Voice Devs.

#AI #GenAI #ArtificialIntelligence #MachineLearning #LargeLanguageModels #LLM #OpenAI #Google #Gemini #Sora #Anthropic #Meta #LLAMA #AppleIntelligence #Amazon #Alexa #VoiceAssistant #Wearables #MetaRayBans #AndroidXR #GoogleGlass #Agents #Developer #Tech #Innovation #LangChain #GDE #TwoVoiceDevs #Podcast #YouTube #2024Recap #2025Predictions #TechTrends #Programming #Coding #OpenSource

It's the holiday season, and the AI world has been showering us with gifts! Join Mark and Allen on Two Voice Devs as they unwrap a mountain of new announcements and releases from Amazon, Meta, Google, and OpenAI. From groundbreaking new models to developer-friendly tools, this episode is packed with insights on the latest advancements in AI. We'll explore the features and potential of each new "present" and discuss what it means for you, the developer.

[00:00:00] Intro and Holiday Greetings: Mark and Allen kick off the show, reflecting on the recent flurry of AI releases.

[00:00:15] The AI Gift Giving Season: A lighthearted introduction on the sheer volume of new AI tools being released.

[00:01:41] Amazon Nova Models: Amazon's surprising release of multiple new models, including Micro, Lite, and Pro, with a peek at Canvas (image generation) and Reel (video generation).

[00:04:42] Meta's Llama 3.3: The focus on multilingual capabilities and open-source nature of Llama 3.3.

[00:05:38] OpenAI vs. Google Announcement Showdown: The back-and-forth between Google and OpenAI with a focus on developer-related announcements.

[00:06:40] Google's Imagen 3 & Veo: Google's new advancements in image and video generation available on Vertex AI, including image editing via prompting.

[00:07:28] OpenAI's Sora Release: OpenAI makes their impressive video generation model available, but notably, not yet via API.

[00:08:34] OpenAI's Canvas for Code: Explore how you can interact with code as a chatbot on a virtual canvas.

[00:09:21] Microsoft's Expanded Copilot Free Tier: A note about Microsoft expanding access to their code tool.

[00:09:38] Google's Jules: The AI Bug Detective: An introduction to Google's automated bug-fixing system which proposes fixes in a version control branch.

[00:11:09] OpenAI's O1 Model: The official release of the O1 model with function calling, structured output, and image input capabilities.

[00:11:42] Gemini 2.0 API: Google's improved Gemini API, now in public preview, offering better performance with optimized tools.

[00:14:01] OpenAI's Real-Time API & WebRTC: Details about real time APIs, including WebRTC support for simplified browser-to-server connections.

[00:16:15] Google's Gemini 2.0 Live API: Real-time streaming API using WebSockets for multimodal input and output, with demos available on AI Studio.

[00:17:01] Google's New SDKs: A deep dive into the unified libraries for AI Studio and Vertex AI, simplifying things for developers.

[00:18:10] OpenAI's new Java and Go Libraries: OpenAI ups their game by adding libraries to match Google's supported development platforms.

[00:19:49] Google's PaliGemma 2 and Android XR: Vision-enabled open model, and a new Android platform for headsets and smart glasses.

[00:22:04] Wrapping Up: Mark and Allen discuss which tools they're most excited about for the break and what's in store for the future.

Let us know in the comments what you're most excited about, or if you noticed anything we missed. We’ll discuss it on future episodes.

#AI #ArtificialIntelligence #MachineLearning #GenerativeAI #LLM #LargeLanguageModels #AmazonNova #Llama3 #Gemini2 #OpenAI #GoogleAI #VertexAI #AIStudio #ChatGPT #GPT #O1 #Reasoning #ImageGeneration #VideoGeneration #DeveloperTools #Coding #Programming #WebRTC #AndroidXR #TechNews #TwoVoiceDevs

Ever felt like your tech presentations, tutorials, or even code explanations are falling flat? You're not alone! In this episode of Two Voice Devs, Allen and Mark dive deep into the art of effective communication in tech, exploring how to move beyond just listing facts to building a compelling narrative that actually helps people understand.

Inspired by a recent presentation that Allen felt was "just okay," they tackle the challenge of how to present information in a way that resonates, whether you're on stage, creating content, or mentoring new developers.

[00:00:00] Introduction to Two Voice Devs

[00:00:16] End-of-year craziness and the inspiration for the episode

[00:00:40] Allen's experience with a presentation that felt flat despite positive feedback.

[00:01:31] The realization of a missing narrative in the presentation.

[00:02:27] Discussion of building narrative into different types of content.

[00:02:33] Deep dive into the structure and content of Allen's Gemini presentation.

[00:04:04] The real message Allen was hoping to convey, and where the presentation fell short.

[00:05:34] The importance of the "why" behind the "what" when presenting new features and concepts.

[00:05:50] Exploring the concept of "telling a story" to make technical concepts easier to understand.

[00:06:29] How individual learning experiences influence the way that you present material.

[00:07:51] Balancing the desire to include all the information, while also keeping a succinct message.

[00:08:50] Pivoting to talking about other ways of imparting information.

[00:09:07] Mark's method of learning and creating diagrams, which then turn into a video.

[00:11:08] The challenge of jumping into code without sufficient background.

[00:12:10] Presenting information in the order that makes sense to you and why.

[00:12:59] Learning by creating and being willing to share even when you are still learning.

[00:13:39] Why committing to a presentation helps you learn a subject.

[00:14:44] Using social media to get information out there quickly, and also, sample projects.

[00:15:27] How starting with small chunks of code can help with understanding

[00:16:31] Using AI tools to explain code.

[00:17:13] How developers need to understand why code works, and not just that it works.

[00:18:58] Why it's important to make learning a conversation and asking questions.

[00:19:29] Mentoring and understanding where students are starting from.

[00:20:54] How in-person feedback is both a benefit and a challenge.

[00:22:12] Creating a safe space for collaborating and learning together.

[00:23:38] Working together to get a level of understanding.

[00:24:13] Call to action for audience to share their techniques.

#TechContent #TechTutorials #DeveloperPresentations #Mentoring #SoftwareDevelopment #CodeTutorial #DevTips #TechNarrative #CommunicationSkills #TwoVoiceDevs #Coding #SoftwareEngineering #Teaching #Learning #AI #Gemini #Storytelling #MadeToStick #TechnicalCommunication #DevFest #Programming