Dominic Williams: Why AI is a smart contract

Author: Dominic Williams, Founder and Chief Scientist of DFINITY Foundation

First of all, I want to thank HashKey and Wanxiang for inviting me back to this great event. It is a great honor to be here today. I want to talk about AI running on the blockchain. As a smart contract, we hear a lot of discussions about Web3 blockchain, blockchain for AI, tokenization for AI, which is a completely different topic. We are talking about AI running on the blockchain.

If you are not familiar with our company, it is a non-profit foundation based in Switzerland. In the past many years, we have operated the largest technology and R&D in the blockchain field. As blockchain maximalists, we try to promote the singularization of blockchain and play an important role in the cloud through the third generation of blockchain. We want to move the world's computing to the blockchain. We want to use the blockchain to run the world's online services, enterprise systems, etc.

The DFINITY Foundation was established in October 2016, and the Internet computer network, the network that you see today, is the product of more than 1,000 years of scientific research efforts per person per year. We have many famous crypto scientists, computer science researchers and engineers involved in R&D, so we have produced a lot of research results and academic publications. The Internet of Things Computer is a huge innovation and the world's first third-generation blockchain. It is also the embodiment of large-scale blockchain.

If you go to the website Dashboard.internet.Computer.Org, which is actually the dashboard of the Internet Computer network, you will see the transaction figures. Today, the Internet Computer blockchain processes the equivalent of 330,000 Ethereum transactions per second. If you look at the low-level smart contracts that are executed by the blockchain, this is a very large number. Maybe in the past, Bitcoin could only do 10-20 transactions per second, but now it can do 300,000 equivalent Ethereum transactions per second, which is a huge improvement. So when we talk about artificial intelligence, we may soon enter the equivalent of 1 billion Ethereum transactions per second. By the end of the year, our plan may be to reach a trillion Ethereum transactions per second, which is also the world's first and only third-party infinitely scalable blockchain network.

What exactly drives the calculation of smart contracts? For this kind of small smart contract calculation, developers can build some small contract software on the Internet Computer. For example, they can deploy 100% on social networks, build social networks, social media, ledgers, etc. on the blockchain, and even exchange DEX looks very similar to traditional exchanges. They can do SWOT transactions, just like Nasdaq can be expanded to the entire blockchain. These small contract software can better serve the user experience and let us enter a new era, running artificial intelligence on small contracts and running artificial intelligence on blockchain networks.

Let's talk about how artificial intelligence operates on small contracts, and how these small contract software runs artificial intelligence on Internet computers? The answer is that it actually has different capabilities on each small contract. It can store 400GB of data. There is also stable memory on the Internet Computer, with 4GB of storage. Of course, it will increase soon, which is very important for AI.

Web3 is actually a very efficient way to operate. The co-founder of WebAssembly is also the founder of the Internet Computer. Guess (sound) games can be used as small contracts to pay for their calculations. This seems to be the application of traditional blockchain in the cloud, and Ethereum is the same. Small contracts also have to make corresponding payments for their own calculations. They can also perform calculations in parallel and have large-scale calculations on the network. They are like multi-block transactions.

Traditional blockchains, such as Ethereum, may trigger a small contract, and this calculation can be completed in a single block. Transactions in the Internet Computer can extend the calculation to multiple blocks, which is very important for AI. These small smart contract processes of the Internet Computer can serve the interactive Web3 experience. There may be thousands of valid small contracts running in traditional blockchains, and they are multiple blockchains interacting. You can also automate the Internet Computer and use more contracts to make it compatible with AI in the end.

If you haven't seen it yet, you can go and see a recent presentation of mine, a fully decentralized AI presentation posted on Twitter. Recently, I showed a neural network running a small contract on the blockchain of the Internet Computer. This neural network classified an image. In this presentation, I also showed that if the neural network sees an image, it will see if the neural network can equate it to an Americano or a coffee cup. So we believe that artificial intelligence is the "heartbeat" of Web3, so we want to invite everyone to dream with me and imagine how artificial intelligence can run on the blockchain and how it can be used.

Here is an example of Ethereum security, which is also a common problem. You can use the blockchain's Explore as an Either skin (sound) to briefly explain that as a small contract of DeFi, you will see the original code of these smart contracts on Ethereum. BlockChain Explorer can match these hashes in the Get-hepy (sound) original code to display this code. Most people are not software engineers and cannot read these codes, and software engineers do not have time to audit other people's codes, so there will be some vulnerabilities at this time. There may be some malware that uses these small contracts to enter other smart contracts. This is a huge problem of traditional blockchains. One solution is to use artificial intelligence and AI smart contracts to run on Internet computers. As a small contract, you can browse the small contracts of Ethereum and then verify whether it has no vulnerabilities, no backdoors, and no malware. At the same time, we can publish a proof that will be displayed on the website saying that this is the code. You may not want to read it or understand it, but OK, because artificial intelligence has checked it for you, which provides more guarantees and allows people to interact with DeFi.

The next example is automated KYC, which is about knowing your customers. This is a financial process to combat money laundering. A common problem is that decentralized exchanges on DEXs want to list certain security tokens. Security tokens represent a DEX of stocks, real estate or gold, and the securities space is actually highly monitored. If a DEX wants to list a security token, you need to conduct KYC to understand your customers and find out who is trading securities. This requires some small contracts to do an identity verification through KYC and verify the small contracts. KYC identity verification, small contracts can provide a solution.

Just now I mentioned that the image classification of the neural network shows that there is a performance video on Twitter. The neural network can also be retrained to do face recognition, which can create a small contract that allows users to interact through videos to verify that the person in the video matches the photo in the identity document he provides. This can make identity verification KYC completely automated and integrated into DEX, so that smart contracts will not have KYC problems, allowing Internet identities to be verified and allowing traders to enter DEX.

If you are interested, you can go to ICDex.io (sound) to have a look. It is built on the blockchain and its user experience is like, under the control of Dao.

The last example is organizational intelligence. If we imagine that a large language model is running in a startup company or government department, and the purpose of the large language model is to improve the effectiveness of the organization, it may be to bring a lot of information together. The idea is to fill the large language model with relevant data, such as chat, email, calendar, etc. Some technical organizations, such as Get-hepy or Zero (sound), or some of them will interact with AI. For example, give them some information. My team has such an idea, or is hindered by such things. People may say artificial intelligence, my CTO has a free time this afternoon to hold a meeting. Does anyone have any ideas? You can solve this problem. This problem, if AI runs on traditional IT infrastructure, if it is hacked, someone hacks the infrastructure that runs AI, they can obtain this private information, because it is stored in one place, so traditional IT infrastructure is not secure.

Second, what if artificial intelligence stops? In the future, if we rely on artificial intelligence, it will be like relying on satellites, just like city navigation. Traditional IT infrastructure is not resilient or strong enough, so the solution is to use organizational-level AI and large language models as a glue to run on the Internet and blockchain, so that large language models can be used more efficiently as a smart contract. With the power of small contracts, like other small contracts, it is safe, it is also composed of corresponding security data and code, it cannot be stopped, they will not crash, they have very strong resilience and independent operation, and only by distributing it to the Internet computer, just like if there is a central organization in Hong Kong, you can use this network. If you want to contact me, you can scan this QR code, or get to know us through these contact information.