Ep. 65 - Size Matters. Small (Models) Win w. Nick Havryliak, co-founder @ Assisterr

Wild User Interviews Podcast (Wuipod)

Wide Ranging conversation with Nick Havryliak, co-founder at Assisterr. Assisterr is a Data Provenance Protocol & Marketplace of domain-specific Small Language Models (SLMs) owned by the community.

Summary
In this episode we discuss the convergence of Web3 and AI. They delve into Nick's background in consulting, product management, and his journey into Web3 and AI. Nick elaborates on the limitations of Large Language Models (LLMs) and explains the advantages of using Small Language Models through the Retrieval Augmented Generation (RAG) process. They also brainstorm innovative applications for community-generated data, such as leveraging past hackathon projects to enhance future hackathons. Throughout the conversation, they explore the importance of maintaining human-centric values in AI development while navigating the evolving technological landscape.

Spicy and Memorable
Things that stood out to me (AVB) and kept me reflecting long after episode was recorded:

* Importance of keeping it human. On dancing, focusing on enjoying life and spreading good vibes.
* How the geopolitical (Ukraine) has shaped Nick's worldview. The West just doesn't get it.
* Approaching Dev Rel with a Product Mindset; leveraging AI to obtain efficiency gains and unlock more value and new use cases.
* The limitations of current LLMs (trained on publicly available data) and how to create incentives for procuring privately held data.
* Cost structure of different approaches to AI: RAG, fine tuning, and own base models.

Resources

Assisterr - https://www.assisterr.ai/
Assisterr on X - https://x.com/assisterr

Nick on X - https://x.com/nick_havryliak