The New Frontier: Generative AI Enhancing Blockchain | by Deepa Ramachandra | October 2023

Generative AI improving the blockchain. Image source: iStock

The generative AI wave that started with the arrival of OpenAI’s ChatGPT shows no signs of abating. Every business, big or small, is looking for GenAI use cases like the proverbial hammer looking for a nail. Every technology – new or old – has been upgraded with the technology that powers ChatGPT. Each is hoping to transform their business and incorporate this seminal invention into their existing technology line. That is why we often come across blog posts, articles and research papers about using ChatGPT or GenAI or Generative AI with _______ technology to revolutionize the world of ________. One of the technologies that people are trying to incorporate GenAI into is blockchain.

Most often, when we hear the words generative AI and blockchain together, it’s usually in the context of how blockchain can benefit genAI by adding data security, transparency, and authenticity. So is it a crooked relationship? Or could we build a more symbiotic connection between them? This article examines this assumption.

Make “smart contracts” smarter with GenAI

When the word “smart contract” was introduced, no one thought of the irony it would create in the future. The code that is written to perform blockchain transactions is often exploited for profit. Critical bugs in smart contract code, which sometimes even penetrated auditor networks, led to bad actors draining $480 million from DeFi platforms in the first half of 2023 alone. None of this suggests anything “smart” about these contracts. In addition, those who understand the laws of business often can’t code, and those who don’t understand the nuances of businesses very well. As a result, the “smart contract” codex is neither wise nor legally binding, with many loopholes to exploit. So could we improve these contracts with generative artificial intelligence?

For starters, generative AI is good at one thing – learning. These algorithms can learn from previously documented vulnerabilities in smart contract code and turn code almost free of bugs and vulnerabilities. The “abuse proof” of the code would improve with each iteration. It can also be used by people who understand business but are not able to code

Assistants like ChatGPT. Non-coders can explain to the bot what a legally binding contract should look like and give it specific instructions on what the contract code should contain. The chatbot would then combine its previous knowledge with what is required and create an effective coded contract. And for those who find it cumbersome to write repetitive code across contracts, genAI copilots can generate standard code so human coders can focus on what makes each code different. Our friend genAI can perform the necessary but overlooked craft of code documentation. It can also review smart contract code and recommend modifications, fixes, and improvements. Critical aspects such as debugging errors, explaining bugs and vulnerabilities in code, and providing solutions are tasks well-suited to generative AI.

Generating tokenized artworks

Generating works of art – images, music, videos, etc. is a common use case for generative artificial intelligence. Tokenizing and tracking digital art on the blockchain is what NFTs are for. What if we can bring genAI to the NFT space?

While artwork mining, generative AI algorithms can insert a unique code into the generated digital art. The code serves as a token that uniquely identifies the artwork and is recorded on the blockchain during the mining process. Digital art can be downloaded or used by anyone, but simply scanning the embedded code would prove who the rightful owner is, preventing its misuse. Creators can earn their royalties every time their artwork changes hands by effectively tracking their digital collectible on the blockchain.

Image source: iStock

Where is My Crypto?

Answering routine questions like these is another job for bots like ChatGPT. Similar to digital assistants for taking notes or ordering food, robots can be used query for information about your crypto account to get an idea of ​​your NFT and crypto holdings. Generative AI can add value by providing up-to-date information on the crypto markets and detailed information about the NFTs you are applying to buy, which helps in decision making. While text bots may not be attractive here (why would I engage in a time-consuming text conversation with a chatbot to find out my wallet balance when I can open the wallet app and get the information instantly), a virtual voice assistant might save time (after all, I can listen to my cryptocurrency goes up or down while my hand is busy making breakfast).

Gain descriptive and prescriptive insights

Today we have traceable and verifiable data captured but distributed across many blockchain networks. With the proliferation of web3 platforms and the growing number of digital coins and NFTs, it is increasingly difficult to assimilate on-chain and off-chain data to gain meaningful insights. This is where multimodal generative AI models can help. But how? Generative models can determine patterns and relationships in massive data sets by collecting all available discrete blockchain data sources. Multimodal generative AI — models that can process different types of data — images, text, audio, etc. — can combine non-blockchain data with on-chain data to create broader insights. GenAI could then create meaningful, personalized and contextual data visualizations. Algorithms can also learn from these insights and prescribe actions to be taken proactively. If we already have existing dashboards, these models can help interpret complex reports and visuals and provide textual summaries of what they mean to us. It would save time that we would otherwise spend wrapping our heads around what the numbers, columns and pies on the screen mean.

Personal tutor Web3

Distributed Ledgers (DLT), Blockchain, Decentralized Exchanges, Smart Contracts, Liquidity Pools, Zero-Knowledge Proofs – these jargons are enough to scare off anyone who wants to dive into the waters of the web3. Therefore, it is imperative that anyone who intends to work or invest in the crypto sector arm themselves with the knowledge of DLT. Besides, some people would understand the intricate concepts of blockchain before the rest of us. What we could use here is education about blockchain and crypto at a pace that follows the student. How can large language models help?

Large language models, or LLMs, thoroughly trained in blockchains, DLTs, and cryptocurrencies could serves as a comprehensive library for web3. On top of the LLM sits a ChatGPT-like tutor who could create a bespoke curriculum – a web3 crash course or lessons spanning four years. This a personalized teacher would create the right level and the right kind of content and deliver it to us in the way that best suits our needs.

Image source: iStock

Create virtual worlds with GenAI

Up until now, we have been talking about the use of generative artificial intelligence in the physical world we inhabit today. But what about the myriad digital realms we create and virtually live in – the metaverse? People buy houses with cryptocurrencies in these virtual worlds. Companies run marketing campaigns in their digital space. We can even buy goods virtually with digital currencies and then pick up its physical counterpart at a nearby brick-and-mortar store. Generative AI could significantly boost the economy of the metaverse. But how?

GenAI could create meta-versions at a reduced cost. He can create hyperrealistic characters, plots, avatars, digital stores, campaigns, and anything else people ask for. Using existing photos, maps and building specifications, generative artificial intelligence could one day create an interactive and immersive world that feels hyper-realistic and has the physical properties of a real environment. The further into the future we go, the better genAI will mesh with the meta version. Imagine the day when we have a theory of mind-level AI—an emotionally intuitive AI that can better understand people by discerning their emotions, beliefs, and thoughts. As our AI friend begins to learn about us on a deeper level, it can then create virtual worlds that adapt to our real-world mood and personality. Imagine that we received a significant amount thanks to the airdrop of a crypto scheme. We nobly (and feel a little too generous) dress up our avatar and walk around in our metaverse. Much to our surprise, we still stumble upon party spots or stumble upon malls, theaters, food parks, and other ways to spend our digital supplies. How?

Or we just had a fight in real life and logged into our virtual world to cool off. Suddenly, our avatar finds himself walking along a forest path and hears the blissful sounds of nature. We also see an offer to rent a cottage tailor-made for us. Wouldn’t you spend more time improving your mood with a small crypto payment? In this way, genAI could create virtual worlds that adapt to our whims and moods so well that it starts to feel mystical.

Generative AI has the potential to democratize the arcane blockchain and web3 space. Still, it’s important to keep in mind that these large language models don’t produce text and images because they’re sensible. What they do is generate the most likely word or image that has the highest probability of occurrence based on the data they were trained on. Since models are only as good as the data behind them, it is important that they are trained with exhaustive data that is specific to the use case.

On the other hand, blockchain is also not a panacea for all data problems. Simply adding data to the blockchain does not implicitly guarantee its integrity. Conversely, if the data entered into the string is incorrect, there is no way to erase it due to the immutable nature of the technology. As they say, garbage in is garbage out.

However, if used carefully, generative AI would add enormous value to blockchain and web3 technologies, and vice versa.

Leave a Comment