nullNow you can directly access the bytes. You can go directly to the chart location in the book, copy the chart to your application, and execute it. This significantly improves the efficiency of the system. This is not just a minimum viable product (MVP), it is a fully functional, well-constructed data access mechanism. Therefore, you have an infinite computing system and an infinite hard drive, and when you combine them, you have a supercomputer.
This has never been built before, and now it is being offered to everyone at the lowest cost. This is the current status of AO, and we are very excited about it. The implementation of the system also operates at the operating system level. Therefore, we are turning WeaveDrive into a sub-protocol of AO, which is a computing unit extension that anyone can load. This is interesting because it is the first extension of its kind.
AO has always had the ability to add extensions to the execution environment. Just as you have a computer and want to add more memory, or add a graphics card, you physically insert a unit into the system. You can do this with AO’s computing units, and that’s what we’re doing here. So, at the operating system level, you now have a hard drive that simply represents the file system for data storage.
This means that not only can you access this data in AO and build applications in the usual way, but you can actually access it from any application brought onto the network. Therefore, this is a widely applicable capability that all personnel building in the system can access, regardless of the language they use, whether it is Rust, C, Lure, Solidity, etc., they can access it as if it were a native function of the system. In the process of building this system, it also forced us to create sub-protocol protocols and create other methods for extending computing units so that other people can also build exciting things in the future.
Now we have the ability to run calculations in memory sets of any size and load data from the network into processes within AO. The next question is, how to conduct inference itself.
Because we chose to build AO on Web Assembly as its primary virtual machine, it is relatively easy to compile and run existing code in this environment. Because we built WeaveDrive to be exposed as a file system at the operating system level, running Llama.cpp (an open-source large language model inference engine) on the system is actually relatively easy.
This is very exciting because it means that not only can you run this inference engine, but you can also easily run many other engines. Therefore, the last component to making large language models run in AO is the large language model inference engine itself. We have ported a system called Llama.cpp, which sounds a little mysterious, but it is actually the leading open-source model execution environment.
Running directly in AO smart contracts, once we have the ability to have any amount of data in the system, and then load any amount of data from Arweave, is actually relatively easy.
To achieve this, we also worked with a Single Instruction, Multiple Data (SIMD) computing extension, which allows you to run these models faster. Therefore, we have also enabled this feature. This means that currently these models run on the CPU, but the speed is quite fast. If you have asynchronous computing, it should suit your use case. Like reading news signals and then deciding which transactions to execute, it works well in the current system. But we also have some exciting upgrades that will be discussed shortly about other acceleration mechanisms, such as using GPU acceleration for large language model inference.
Llama.cpp allows you to load not only Meta’s leading model Llama 3, but also many other models. In fact, nearly 90% of the models you can download from the open-source model website Hugging Face can run in the system, from GPT-2 if you wish, to 253 and Monet, Apple’s own large language model system, and many other models. So now we have a framework that allows you to upload any model from Arweave, upload the models you want to run in the system using the hard drive. You upload them, they are just regular data, and then you can load them into AO’s processes and execute them, get results, and work in the way you like. We believe this is a bundle that makes applications that were previously impossible in the smart contract ecosystem possible, even if they are possible now, the amount of architectural changes in existing systems such as Solana is just hard to predict, not on its roadmap. Therefore, in order to demonstrate this to you and make it real and easy to understand, we have created a simulator called Llama Fed. The basic idea is that we get a Federal Reserve Board, which are llamas, whether as the Llama 3 model or as the chairman of the Federal Reserve.
We also tell them that they are llamas, just like Alan Greenspan or the chairman of the Federal Reserve. You can enter this small environment.
Some people will be familiar with this environment, in fact, it’s just like Gather where we work today, you can chat with llamas, ask them for tokens for a very interesting project, and they will decide whether to give you tokens based on your request. So you burn some Arweave tokens, wAR tokens (provided by the AOX team), and they will give you tokens based on whether they think your proposal is good. So this is a meme coin, with completely independent and intelligent monetary policy. While it’s a simple form of intelligence, it’s still interesting. It evaluates your proposal and those of others, and runs monetary policy. Analyzing news headlines and making intelligent decisions or interacting with customer support and returning value, all of these can now be implemented in smart contracts. Elliot will now demonstrate.
Hello, I’m Elliot, and today I want to show you Llama Land, a chain-driven autonomous world running inside AO, powered by Meta’s open-source Llama 3 model.
The conversations we see here are not just between players, but also between fully autonomous digital llamas.
For example, this llama is human.
But this llama is a chain AI.
This building contains Llama fed. It’s like the Federal Reserve, but serving llamas.
Llama fed runs the world’s first AI-driven monetary policy and mints Llama tokens.
This guy is the Llama king. You can give him packaged Arweave tokens (wAR) and write a request to get some Llama tokens.
Llama King AI will evaluate and decide whether to grant Llama tokens. Llamafed’s monetary policy is completely autonomous, with no human supervision. Every agent in the world and every room itself is a chain process on AO.
It looks like the Llama King has granted us some tokens, and if I check my ArConnect wallet, I can see that they are already there. Nice. Llama Land is just the first AI-driven world implemented on AO. It’s a new protocol framework that allows anyone to build their own autonomous world, with the only limitation being your imagination. All of this is 100% implemented on the chain, and only possible on AO.
Thank you, Elliot. What you just saw is not only a large language model participating in financial decision-making and running an autonomous monetary policy system. There are some very interesting things here. These places bring together people using different financial products. We see in the DeFi ecosystem that if someone wants to participate in a project, they first check on Twitter, then visit the website and participate in the basic components of the game.
Then they join a Telegram group or Discord channel or chat with other users on Twitter. This experience is very decentralized, and we all jump between different applications. One interesting idea we are trying is, if you have the user interface of these DeFi applications, let their communities come together and collectively manage this autonomous space they access because it’s a permanent web app, adding to the experience.
Imagine being able to go to a place that looks like an auction house, chat with other users who like the protocol. Basically, you can chat with other users when financial mechanism processes are happening on AO. The community and social aspects are integrated with the financial part of the product.
We think this is very interesting and has a wider impact. You can build an autonomous AI agent here, roam it in the Arweave world, and interact with different applications and users it finds. So if you’re building a metaverse, when you create an online game, the first thing you do is create NPCs (non-player characters). Here, NPCs can be universal.
You have an intelligent system that roams around and interacts with the environment, so you don’t have a cold start problem. You can have some autonomous agents, trying to make money for themselves, trying to make friends, interacting with the environment like normal DeFi users. We think this is very interesting, albeit a bit quirky. We look forward to seeing what happens.
Looking ahead, we also see opportunities to accelerate large language model execution in AO. Earlier, I talked about the concept of computing unit extensions. This is how we built WeaveDrive.
Not just stopping at WeaveDrive, you can build any type of extension for AO’s computing environment. There is an exciting ecosystem project that is solving the problem of GPU-accelerated large language model execution, the Apus Network. I’ll let them explain.
Hi, I’m Mateo. Today, I’m very excited to introduce Apus Network. Apus Network is dedicated to building a decentralized, trustless GPU network.
We use permanent on-chain storage through Arweave to provide an open-source AO extension that provides a deterministic execution environment for GPU and provides economic incentives for decentralized AI using AO and APUS tokens. Apus Network will competitively execute the optimal, trustless model training on Arweave and AO using GPU mining nodes. This ensures that users can use the best AI models at the most cost-effective price. You can follow our progress on X (Twitter) @apus_network. Thank you.
That’s the current state of AI on AO today. You can try Llama Fed, try building your own smart contract applications based on large language models. We believe this is the beginning of introducing market intelligence into a decentralized execution environment. We are very excited about this and look forward to seeing what happens next. Thank you for your participation today, and we look forward to communicating with you again.
