AI16Z is a major player that has emerged in this round of AI agency waves, and is also the initiator of many AI agency projects. Many of their ideas for AI agents are avant-garde and bold.

Shaw, the founder of AI16Z, recently visited Shanghai and shared some of his views on the future development of AI agents with industry media.

Crypto games have always been the focus of my attention. In the current encryption ecosystem, the game ecosystems that closely follow Ethereum and that I am more concerned about include: Starknet eternum.realms.world (see November 26th article), Boring Ape’s ApeChain, Treasure second-tier extension and Redstone second-tier extension.

My previous expectation for these game ecosystems was that they would develop a world of on-chain autonomy, but I have not seen much progress in this direction.

When AI agents rise, I guess this may bring a different way of playing crypto games. In the article on December 18, I once speculated:

"Don't Boring Ape and Magic have their own dedicated game (second-layer expansion) chains ApeChain and Treasure? ? They can also use Virtuals' gameplay to support game-like AI agents in their own ecosystems, and strongly empower their own tokens. In this way, they can also build their own game-like AI agent ecosystem and boost their own tokens. Price"

Shaw mentioned two of the four ecosystems above this time.

The first is a game in the second-level expansion of Treasure:

"Shaw means @ elizawakesup is working with @Treasure_DAO on a game called Smolworld where you have a little pet monkey that you can tell it to do and it may or may not listen to you. This game is as fun as it gets without AI. Agent This kind of game is unachievable, and your goal is to take care of your virtual pet like a parent."

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

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

In the first scenario, the A.I. The agent is set as the protagonist of the game, and it has its own "judgment" and "opinion".

Although the protagonists in current games have their own "judgments" and "opinions", these are preset by the program, but they are added Some random variables make the protagonist seem more unpredictable.

But the "judgment" and "opinion" of the AI ​​agent come from its own learning and evolving "intelligence". It is very likely that such an agent will become the protagonist of the game. This will make the game a different experience.

This reminds me of a conversation between humans and AI agents I saw online yesterday:

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

AIXBT answered 102K.

When I saw this answer, my first reaction was not whether the prediction was accurate, but that it actually answered the question and seemed to be able to answer it. It has a look and feel----------The price is entirely possible, but it is impossible to tell whether it is correct or not, making it difficult to tell whether it is an AI or a real person.

If an AI agent is used as the protagonist in future games, can we still tell whether we are playing against a "real person" or an AI?

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

As some of our readers mentioned in the comments at the end of the previous article, players can train their own AI agents in the future and give the AI ​​agents a certain amount of initial funds. , and then let the AI ​​agent play games for you, earn assets, and defeat monsters to upgrade.

During this process, players can also continuously communicate with the agent.own experience to help agents grow rapidly.

Both scenarios are completely achievable with current technology and will definitely appear in the encryption ecosystem in the near future.

Lenovo AI16Z recently cooperated with Stanford University on AI agents in the crypto-economy. This ecosystem has already included the university's scientific research institutions. Its development next year will definitely produce even more amazing results.