Two Voice Devs

Join Allen Firstenberg and Roger Kibbe as they delve into the exciting world of local, embedded LLMs. We navigate some technical gremlins along the way, but that doesn't stop us from exploring the reasons behind this shift, the potential benefits for consumers and vendors, and the challenges developers will face in this new landscape. We discuss the "killer features" needed to drive adoption, the role of fine-tuning and LoRA adapters, and the potential impact on autonomous agents and an appless future.

Resources:

* https://developer.android.com/ai/aicore

* https://machinelearning.apple.com/research/introducing-apple-foundation-models

Timestamps:

00:20: Why are vendors embedding LLMs into operating systems?

04:40: What are the benefits for consumers?

09:40: What opportunities will this open up for app developers?

14:10: The power of LoRA adapters and fine-tuning for smaller models.

17:40: A discussion about Apple, Microsoft, and Google's approaches to local LLMs.

20:10: The challenge of multiple LLM models in a single browser.

23:40: How might developers handle browser compatibility with local LLMs?

24:10: The "three-tiered" system for local, cloud, and third-party LLMs.

27:10: The potential for an "appless" future dominated by browsers and local AI.

28:50: The implications of local LLMs for autonomous agents.

Join us on Two Voice Devs as we welcome back Roger Kibbe. Fresh off emceeing the developer track at the Unparsed Conference in London, Roger shares his insights on the biggest takeaways, trends, and challenges facing #GenAI, #VoiceFirst and #ConversationalAI developers today.

Get ready for a dose of reality as Roger emphasizes the need to view LLMs as powerful tools – think hammers – rather than magical solutions. We dive deep into:

Timestamps:

* 0:00 - Intro

* 1:56 - Exploring the Unparsed Conference

* 4:47 - LLMs: The hype vs. the reality for developers

* 6:37 - The underappreciated power of LLMs for "understanding", not just generating

* 11:03 - The right tool for the job: Why a toolbox approach is essential for conversational AI

* 13:52 - Beyond the chatbot: Detecting emotion and the future of human communication

* 20:28 - Hackathon highlights and the need for more realistic QA approaches

* 28:55 - Navigating the shift from deterministic to stochastic systems

* 31:59 - Will AI replace junior developers?

* 36:30 - How senior developers can (and can't) benefit from AI coding assistants

* 39:04 - Final thoughts: The value of cutting through the hype

Don't miss this insightful conversation about the future of conversational AI development – grab your toolbox and hit play!

What should people developing with LLMs learn from a decade of experience building Alexa skills? How will Alexa skill developers leverage the latest #GenerativeAI and #CoversationalAI tools as they continue to build #VoiceFirst and multimodal skills?

Join Allen and Mark on Two Voice Devs as they delve into the evolving landscape of Alexa skill development in the era of large language models (LLMs). Sparked by a thought-provoking discussion on the Alexa forums, they explore the potential benefits and challenges of integrating LLMs into skills.

Key topics and timestamps:

(0:00:00) Introduction

(0:02:00) LLMs and the Future of Alexa Skills

(0:04:00) Limitations of Current Alexa Skill Model with LLMs

(0:07:00) Benefits and Drawbacks of Developing for Alexa

(0:10:30) Overlooked Potential of Multimodality with LLMs

(0:14:50) Lessons from Early Voice Experiences

(0:17:00) Intents vs. Tool/Function Calling

(0:21:30) Handling Hallucinations and Off-Topic Requests

(0:22:00) LLMs' Ability to Handle Nuanced Intents

(0:28:00) Cost Considerations of LLMs

(0:32:00) Monetizing LLM-Powered Alexa Skills

(0:39:40) The Future of Alexa Skill Development: A Hybrid Approach?

(0:40:00) Outro

Tune in as they discuss the need for hybrid models, the importance of conversation design, and the uncertain future of monetization in this rapidly changing landscape. Don't forget to join the conversation on the Alexa Slack channel or leave your thoughts in the comments below!

OpenAI's ChatGPT 4o and GPT 4o announcements have sent shockwaves through the developer community! In this episode of Two Voice Devs, Mark and Allen dive into the implications of these new models, comparing them to Google's Gemini.

We discuss:

[00:00:10] Initial takeaways from the OpenAI presentations.

[00:02:29] The impressive voice capabilities of ChatGPT 4o.

[00:04:49] Concerns about OpenAI's ambitions for conversational AI.

[00:07:30] The difference between "doing" and "knowing" AI systems.

[00:14:15] A detailed breakdown of GPT 4o, including its strengths and weaknesses.

[00:17:43] Comparison with Gemini and implications for developers.

[00:19:41] The importance of competition in driving innovation and lowering prices.

[00:21:48] The future of AI assistants and the role of developers.

Let us know what you think about GPT 4o and Gemini! Have you used them? Share your experiences and thoughts in the comments below.

Allen Firstenberg chats with fellow Google Developer Expert (GDE) Mike Wolfson about his career, the evolution of Android, and his new interest in generative AI. Mike shares his thoughts on the future of AI with agents, Large Action Models (LAMs), and the potential of the "Rabbit," a new AI-powered device. Does the Rabbit live up to its promise? If not - what could?

Timestamps:

00:00:00 - Introduction

00:01:32 - Mike's career journey

00:04:15 - Transition from enterprise Java to Android development

00:05:04 - Creating "Droid of the Day" app

00:06:49 - Becoming an Android developer and Google Developer Expert

00:09:23 - Shift in focus from Android to generative AI

00:10:57 - Generative AI as a platform

00:11:47 - The Rabbit and its potential

00:14:59 - Mike's take on the Rabbit as a developer

00:17:31 - Current integrations with the Rabbit

00:19:52 - The future of AI and the Rabbit

00:24:46 - Edge AI and its potential

00:27:16 - The capabilities of the Rabbit and its future

00:32:17 - The Rabbit vs. other devices like meta glasses

00:34:28 - Conclusion and call to action

Join Allen and Roya as they dissect the major AI announcements from Google I/O 2024. From Gemini updates and new models to responsible AI and groundbreaking projects like ASTRA, this episode dives into the future of AI development.

Timestamps:

[00:00:00] Introduction and Google I/O Overview

[00:02:00] Gemini 1.5 Flash & Gemini 1.5 Pro: New Models and Features

[00:04:30] AI Studio Access Expansion for Europe, UK & Switzerland

[00:06:20] Choosing the Right AI Model for Your Project

[00:06:50] Gemini Nano in Google Chrome: Bringing AI to the Browser

[00:08:00] Pali Gemma: Open Source Model with Image & Text Input

[00:08:50] AI Red Teaming & Model Safety Tools

[00:09:50] Parallel Function Calling for Developers

[00:10:30] Video Frame Extraction: Easier Multimodal Development

[00:11:20] GenKit: Firebase's Generative AI Integration

[00:12:00] Gems: Customizable Gemini for Developers

[00:12:50] Semantic Embeddings: Understanding & Creating Images

[00:13:50] Imogen 3: API Access for Image Generation

[00:14:20] Veo: Video Generation with Lumiere Architecture

[00:14:50] SynthID: Watermarking & Identifying Generated Content

[00:16:30] Responsible AI & Inclusivity

[00:18:00] Gemini Developer Competition: Win a DeLorean & Cash Prizes!

[00:19:30] Project ASTRA: Multimodal AI with Contextual Memory

[00:21:00] Google Glasses & Project ASTRA Integration

[00:22:00] Closing Thoughts: AI for Everyone

Join Allen and Mark as they delve into Voiceflow's groundbreaking new feature: intent classification using a hybrid of LLMs and classic NLU models. Discover how this innovative approach leverages the strengths of both technologies to achieve greater accuracy and flexibility in understanding user intent. How they're doing it just may blow your mind! 🤯

Timestamps:

0:00:00 - Introduction

0:00:33 - Exploring the concept of intents and slots in conversational UI

0:05:11 - Understanding Natural Language Understanding (NLU) and its role in intent classification

0:06:02 - Voiceflow's hybrid approach: Combining classic NLU with LLMs

0:08:36 - Deep dive into Voiceflow's documentation on intent classification using LLMs

0:13:43 - Understanding the hybrid approach and its components: intent descriptions, prompt wrappers, and training data

0:24:31 - How the classic NLU model pre-filters intents for the LLM, improving efficiency and accuracy

0:27:27 - Exploring the user experience and the flow of intent classification with the hybrid model

0:32:53 - Voiceflow's commitment to open research and sharing knowledge with the developer community

0:35:52 - The value of benchmarking and analyzing different LLM models for intent classification

0:39:12 - Call to action: Share your thoughts and experiences with Voiceflow's hybrid approach

Join Allen Firstenberg and guest host Stefania Pecore on Two Voice Devs as they delve into the exciting announcements and highlights from Google Cloud Next 2024! This episode focuses on the latest advancements in AI and their impact on the healthcare industry, providing valuable insights for developers and tech enthusiasts.

Learn more:

* https://cloud.google.com/blog/topics/google-cloud-next/google-cloud-next-2024-wrap-up

Timestamps:

00:00:00: Introduction

00:01:02: Stefania's background and journey into AI

00:07:20: Stefania's overall experience at Google Cloud Next

00:11:59: Focus on Healthcare and AI applications, including Mayo Clinic's Solution Studio

00:15:38: Exploring the new Gemini product suite and its features like code assistance and data analysis

00:20:44: Discussing Gemini API updates, including the 1.5 public preview with 1M token context window and grounding tools

00:26:06: Vertex AI Agent Builder and its no-code approach to chatbot developmen

t

00:33:02: Hardware announcements, including the A3 VM with NVIDIA H100 GPUs

00:35:24: Stefania's reflections on Cloud Next and the value of attending

Tune in to discover the future of AI and its transformative potential, especially in the healthcare sector. Share your thoughts on the Google Cloud Next announcements in the comments below!

This episode of Two Voice Devs takes a closer look at BERT, a powerful language model with applications beyond the typical hype surrounding large language models (LLMs). We delve into the specifics of BERT, its strengths in understanding and classifying text, and how developers can utilize it for tasks like sentiment analysis, entity recognition, and more.

Timestamps:

0:00:00: Introduction

0:01:04: What is BERT and how does it differ from LLMs?

0:02:16: Exploring Hugging Face and the BERT base uncased model.

0:04:17: BERT's pre-training process and tasks: Masked Language Modeling and Next Sentence Prediction.

0:11:11: Understanding the concept of masked language modeling and next sentence prediction.

0:19:45: Diving into the original BERT research paper.

0:27:55: Fine-tuning BERT for specific tasks: Sentiment Analysis example.

0:32:11: Building upon BERT: Exploring the Roberta model and its applications.

0:39:27: Discussion on BERT's limitations and its role in the NLP landscape.

Join us as we explore the practical side of BERT and discover how this model can be a valuable tool for developers working with text-based data. We'll discuss i

ts capabilities, limitations, and potential use cases to provide a comprehensive understanding of this foundational NLP model.

Embark on a wild race with Gemma as we explore the exciting (and sometimes slow) world of running Google's open-source large language model! We'll test drive different methods, from the leisurely pace of Ollama on a local machine to the speedier Groq platform. Join us as we compare these approaches, analyzing performance, costs, and ease of use for developers working with LLMs. Will the tortoise or the hare win this race?

Learn more:

* Model card: https://console.cloud.google.com/vertex-ai/publishers/google/model-garden/335

* Ollama: https://ollama.com/

* LangChain.js with Ollama: https://js.langchain.com/docs/integrations/llms/ollama

* Groq: https://groq.com/

Timestamps:

0:00:00 - Introduction

0:03:05 - Getting to Know Gemma: Exploring the Model Card

0:05:30 - Vertex AI Endpoint: Fast Deployment, But at What Cost?

0:13:40 - Ollama: The Tortoise of Local LLM Hosting

0:17:40 - LangChain Integration: Adding Functionality to Ollama

0:21:44 - Groq: The Hare of LLM Hardware

0:26:06 - Comparing Approaches: Speed vs. Cost vs. Control

0:27:35 - Future of Open LLMs and Google Cloud Next

#GemmaSprint

This project was supported, in part, by Cloud Credits from Google