The author suggests that AI can help users understand crypto transactions in plain language and protect them from mistakes. However, they caution against using AI directly against malicious misinformers and scammers. They also discuss the potential of using AI as part of the rules of a game, such as in blockchain-based smart contracts or DAOs, but warn about the risks of adversarial machine learning in these applications. The article concludes by suggesting that limiting who can query the model and hiding the training data can help curtail black-box attacks.
Key takeaways:
- The intersections between crypto and AI are becoming more apparent with the rise of powerful AI and crypto technologies, leading to promising applications of AI within blockchain ecosystems.
- AI can be categorized into four major categories in relation to blockchains: AI as a player in a game, AI as an interface to the game, AI as the rules of the game, and AI as the objective of the game.
- While AI can help users understand and navigate the crypto world, it also poses risks, particularly in the form of adversarial machine learning attacks. These attacks can trick AI models into producing incorrect outputs.
- Despite the challenges, there are potential solutions such as limiting who can query the AI model and hiding the training data while ensuring the process used to create the data is not corrupted.