User-Owned AI: The Rise of Small Language Models

Small language models (SLMs) are gaining popularity, driven by the need for privacy and user ownership. In this session you will learn what large language models (LLMs) are, how they internally work and the factors that led to their rise. We will provide a brief overview of AI's history, highlighting the development of Transformers and generative models. Next, we’ll introduce SLMs and explain the importance of ensuring privacy and user control. We will explain techniques that make SLMs more efficient, such as Low-Rank Adaptation (LoRA) and model quantization. We will discover new hardware trends that allow AI models to run on personal devices on CPU and Neural Processing Units (NPUs).
The practical benefits of SLMs will be explained, including enhanced privacy, independence from large tech companies, and greater control over AI applications. You will learn how to host your own language models at home or in your company, using tools like Ollama and LM Studio. We’ll also discuss innovative ways to integrate LLMs and SLMs into your daily workflow, including the latest prompt engineering techniques, and ponder whether prompt engineering will remain relevant as models improve.

Modern Software Engineering in the age of Generative AI

Modern Software Engineering in the Age of Generative AI workshop session is designed to provide participants with the skills to deploy advanced Generative AI models using MLOps platforms without needing DevOps infrastructure management, or any kind of knowledge regarding AI.
In this hands-on session, you will:
Learn to deploy state-of-the-art language models like LLaMA3 and image generation models like Stable Diffusion.
Learn to use these models in your applications to provide new features, as well as how to improve the existing ones

Generative AI and LLMs have many real-world applications and we’ll be building some of them as a part of this workshop:

Chatbot which is provided the documentation of your service or application and based on that provides real-time support to users
LLM which analyses why an ML workflow has failed and provides user with feedback on what went wrong based on the code and console output of the workflow
Text-to-image service which generates image datasets which can be used to train various models like image classification, segmentation, and object detection

Beyond the Hype: Decentralizing AI Using Zero Knowledge Proofs

Explore the intersection of AI and blockchain through the lens of Zero Knowledge (ZK) proofs and their potential to transform both fields. This session will explain how ZK technologies like Fully Homomorphic Encryption (FHE), Multi-Party Computation (MPC), hardware enclaves, chain abstraction, and elliptic curves help ensure secure and trusted computation.
We will discuss the current race to create efficient systems with low proving overhead, inspired by the rise of blockchain technology. Initially aimed at improving scaling and cost, these advancements now have the potential to revolutionize traditional cloud services, ushering in the Web3 global cloud era. This new era envisions a world where everyone can provide compute power and be fairly compensated instantly.
This talk will cover the need for verified computation, the current state of ZK technology, and future prospects, all in accessible language without complex mathematics. Learn how the speaker’s winning idea at the Near 2023 hackathon is contributing to this exciting field. Join us to understand how ZK proofs can make AI more decentralized, secure, and accessible for everyone.

A Smooth Introduction to Blockchain Development for Software Engineers

This session helps software engineers transition smoothly to blockchain development by adapting their existing skills to the Web3 landscape. We'll start with the basics: setting up wallets, understanding gas fees, developing and deploying smart contracts. You’ll learn about developer UX, toolchains, and testing, highlighting the differences and similarities with traditional software development.
You will discover the practical aspects of chain abstraction, governance mechanisms, and collaborating with stakeholders in a decentralized environment. We’ll dive into privacy applications, exploring how technologies like Noir and Mina keep data private on public blockchains. You’ll see how blockchain functions as a different form of database and learn to identify when it’s the right fit for your projects.
Expect code demos in Rust, Solidity, and TypeScript, showing the versatility of blockchain development. By the end of this session, you’ll be equipped to navigate the Web3 world confidently, ready for future projects, and more attractive in the job market. Join us to transform your software engineering skills and unlock new opportunities in blockchain development.

Decentralizing DevOps: Building the Future of IaC with Thetaform.Cloud

In this talk, Igor will explore the evolution of Infrastructure as Code (IaC) and its critical role in modern DevOps. He will explain how tools like Terraform have revolutionized infrastructure management through repeatable, scalable, and automated deployments.
The lecture will introduce Thetaform.cloud, a decentralized IaC solution for deploying infrastructure on the Theta Edge Cloud, aligning with standard DevOps best practices. Igor will also discuss the broader Decentralized Physical Infrastructure Networks (DePIN) landscape, featuring technologies like Zero-Knowledge Virtual Machines (ZK VMs), Trusted Execution Environments (TEEs), and Multi-Party Computation (MPC).
Focusing on the future of cloud computing, Igor will highlight how Theta Network supports decentralized AI workloads through Lavita AI and FedML jobs, positioning it to revolutionize the cloud industry. He will explain how Thetaform.cloud empowers DevOps teams to seamlessly transition from traditional legacy cloud providers to decentralized infrastructure, enabling greater scalability, cost efficiency, and secure execution of AI and federated learning tasks. This session is essential for DevOps professionals and anyone interested in the future of decentralized computing.

Thetacon 2024 - Las Vegas