# Road to Autarky: a gamified exploration of dynamics of technical progress

- Channel: [ZuBerlin](https://streameth.org/zuberlin)
- Date: 2024-06-19
- Duration: 22:41
- Topics: AI, blockchain, machine learning, web3, cryptocurrency
- Watch: https://streameth.org/watch/6672d7b207f92b086c3d98be
- Download: https://vod-cdn.lp-playback.studio/raw/jxf4iblf6wlsyor6526t4tcmtmqa/catalyst-vod-com/hls/959dn7tw28jysonb/1080p0.mp4

## Description

Clip The video features a speaker discussing the importance of model composability and composable systems in the context of agentic AI and blockchain infrastructure. The speaker emphasizes that while many people view AI models as simple query-response tools, real-world applications often involve more complex systems composed of specialized models. These models can be combined to perform complex functions and improve accuracy through ensemble techniques.

The talk also covers the inefficiencies of running computations on blockchains due to their replicated state machine nature. To address this, a novel approach called Validation Computation Separation (VCS) is introduced, involving inference nodes that perform stateless computations and transmit proofs to the network for validation, balancing scalability with decentralization.

The speaker further explores the use of machine learning in quoting dynamic fees for automated market maker (AMM) swaps, potentially reducing losses for liquidity providers (LPs). They highlight how traditional finance uses various risk and forecasting models that are not yet common in Web3 but could become foundational with further development.

Looking ahead, the focus will be on creating a collaborative research environment to develop models specifically designed for Web3 protocols. The infrastructure will support various model architectures and hardware acceleration for inference and proof generation. Additionally, an ERC standard is being developed to define agent interactions on EVM-based chains.

Lastly, questions