[Podcast] a16z Crypto's Latest Research: Four New Business Models at the Intersection of AI and Blockchain - Summary

title: [Podcast] a16z Crypto Latest Research: Four New Business Models Under the Integration of AI and Blockchain - Summary
published: false
date: 2023-08-27 00:00:00 UTC
tags: 
canonical_url: http://www.evanlin.com/a16z-ai-blockchain/
---

![image-20230829114535672](http://www.evanlin.com/images/2022/image-20230829114535672.png)

## Source of Information

Chinese translation news: [https://news.cnyes.com/news/id/5292495](https://news.cnyes.com/news/id/5292495)

The original source is from a [podcast](https://web3-with-a16z.simplecast.com/episodes/ai-crypto-centralization-decentralization) by Ali Yahya, one of the former core developers of Google Tensorflow. Highly recommended to listen.

## Thoughts

Some of the ideas mentioned are quite interesting, explaining future business opportunities through four aspects: (mutually exclusive, current challenges, how to cooperate to form new business models, AI and social graphs) I'll pick a few interesting ones to present.

1.  By using zk encryption ([ZKML](https://worldcoin.org/blog/engineering/intro-to-zkml), [Chinese article](https://www.blocktempo.com/base-layer2-to-mint-free-nft-path-to-mainnet/)), to enable LLMs to achieve distributed AI training. And it can allow the original talents to get the correct data. (I don't quite understand zk / zkml here, I'll learn more about it when I have time)

Imagine a situation like this: Alice has a model she wants to protect. She wants to send the model to Bob in an encrypted form. Bob now receives the encrypted model and needs to run his own data on this encrypted model. How to do this? Then you need to use what is called fully homomorphic encryption to calculate encrypted data. If the user has an encrypted model and plaintext data, then the encrypted model can be run on the plaintext data, and the encrypted result can be received and obtained. You send the encrypted result back to Alice, and she can decrypt it and see the plaintext result.

2.  There is an incentive mechanism mentioned inside:
    1.  Large LLMs are open for everyone to use.
    2.  Individual users provide better customized data to optimize the model. (Get incentives)
    3.  Make the model better.
3.  In the past, music platforms allowed creators to communicate with everyone through blockchain encryption. Now, through LLM AI, every participant can become a "creator". They also have new communication models.