Attention isn’t all we need; we need ownership too

Key takeaways

Illia Posushin started programming at an early age, which shaped his career path in AI and machine learning.
The Transformer model represented a significant leap in machine learning, offering new capabilities in natural language processing that continue to evolve.
User-owned AI is a critical concept to prevent the concentration of control over AI technologies in a few corporations, posing risks to innovation and individual interests.
NEAR Protocol exemplifies how blockchain can facilitate smoother, faster transactions necessary for the development and deployment of decentralized AI.
The vision for decentralized AI labs involves community collaboration and infrastructure that prioritizes user rights and data privacy.

Summary

Illia Posushin’s Journey in Tech

Illia reflects on his beginnings in programming at age ten, starting with Pascal and Delphi, leading him into gaming and machine learning. He shares his experience working at a well-established machine learning company in San Diego and his journey to Google Research, where he contributed to natural language understanding and the development of the Transformer architecture. His realization of the potential for deep learning in language processing sparked his desire to work on questions in AI.

The Impact of the Transformer Architecture

The discussion shifts to the revolutionary impact of the Transformer model introduced in the paper co-authored by Illia in 2017. He reflects on how it has changed machine learning, causing a shift in approaches in various applications. Despite expectations for rapid advancements in AI following the paper, Illia notes the evolution has been slower than anticipated, highlighting the gradual improvements and the emergence of AI applications like ChatGPT.

User-Owned AI and Its Future

As the conversation develops, Illia introduces the concept of user-owned AI, emphasizing the need for AI that serves individual users rather than corporate entities. He discusses the challenges and risks of centralized AI control by a few companies and presents his work on NEAR Protocol, which enables decentralized, user-centric AI transactions using blockchain technology.

Innovations in Payment and Infrastructure for AI

Illia outlines how NEAR Protocol facilitates international payments and microtransactions through blockchain technology. He discusses its potential applications in various domains, such as enabling seamless transactions for users across different currencies, facilitating remittances, and providing a platform for crowdsourced AI development.

Creating Decentralized AI Labs

Towards the end of the episode, the focus shifts to Illia’s vision for decentralized AI labs supported by blockchain infrastructure. He explains the process of building AI models with contributions from a global community, leveraging confidential computing to ensure data security and confidentiality in training and inference.

Show notes

NEAR is the blockchain for AI, enabling AI agents to transact freely across networks.

Connect with Illia on LinkedIn and X, and read the original Transformers paper that Illia co-authored in 2017.

Today’s shoutout goes to Populous badge winner Adi Lester for answering the question DataTable - foreach Row, EXCEPT FIRST ONE.

See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.