AI16Z is a major player that has emerged in the current wave of AI agents and is the initiator of many AI agent projects. Many of their ideas about AI agents are quite avant-garde and bold.

Recently, Shaw, the founder of AI16Z, visited Shanghai and shared some of his views on the future development of AI agents with industry media. These views were summarized and published on Twitter by Joe Zhou, the editor-in-chief of ForesightNews @joezhoublack (link at the end of the article).

In this summary, I am particularly focused on his observations regarding the development of AI agents in cryptocurrency games.

Cryptocurrency games have always been a focus of my attention. In the current crypto ecosystem, the gaming ecosystems closely tied to Ethereum that I am particularly interested in include: eternum.realms.world on Starknet (see the article from November 26), Bored Ape's ApeChain, Treasure Layer 2 expansion, and Redstone Layer 2 expansion.

Previously, my expectation for these gaming ecosystems was that they could develop into self-governing worlds on-chain, but there has not been much progress in that direction.

After the rise of AI agents, I speculate that this could bring different gameplay to cryptocurrency games. In an article on December 18, I once speculated:

"Don't Bored Apes and Magic have their own dedicated gaming (Layer 2 expansion) chains, ApeChain and Treasure? They could also adopt the gameplay of Virtuals to nurture gaming AI agents within their ecosystems and empower their own tokens. This way, they could establish their own gaming AI agent ecosystem and support the price of their tokens."

Shaw mentioned two out of the four ecosystems above.

The first one is the game in the Treasure Layer 2 expansion:

"Shaw mentioned that @elizawakesup is collaborating with @Treasure_DAO to create a game called Smolworld, where you will have a pet monkey that you can tell what to do. It might listen to you, or maybe not. This game is interesting because without AI agents, this type of game would not be possible. Your goal is to take care of your virtual pet like a parent."

The second one is eternum.realms.world on Starknet:

"It is integrating Eliza into Eternal, so the agents in the game will have wallets. And you can 'kill' these agents and take their money."

In the first scenario, the AI agent is set as the protagonist of the game, and it has its own "judgment" and "opinions."

Although the protagonists in current games have their own "judgment" and "opinions," these are all pre-set by the program, just adding some random variables to make the protagonists appear to have more unpredictable factors.

However, the 'judgment' and 'opinions' of AI agents come from their own learning and continuously evolving 'intelligence,' making it likely that such agents as protagonists in games will create different experiences.

This reminds me of a conversation I saw online yesterday between a human and an AI agent:

A user asked AIXBT on Twitter, what will the price of Bitcoin be on December 28?

AIXBT answered 102K.

Seeing this answer, my first reaction was not whether this prediction is accurate, but that it actually answered the question, and it looked quite convincing—this price is entirely possible, but it is hard to determine its accuracy, making it difficult to distinguish whether it is AI or a real person.

In the future, if AI agents are used as protagonists in games, can we still distinguish whether we are battling against a "real person" or AI?

In the second scenario, the AI agent plays the role of the game player.

As some readers mentioned in the comments at the end of previous articles, players in the future could completely train their own AI agents, provide the AI agents with some initial capital, and then let the AI agents play games, earn assets, and level up.

In this process, players can continuously communicate their experiences to the agents, helping them grow rapidly.

Both of these scenarios can be fully realized with current technology, and they are bound to appear in the crypto ecosystem in the near future.

In connection with AI16Z's recent collaboration with Stanford University on AI agents in the crypto economy, this ecosystem has already incorporated the university's research institutions. Its development next year will surely yield even more astonishing results.

Reference link:

https://x.com/joezhoublack/status/1872053116860018735?s=46