Hackathon challenge: AI Integration in Bisq

[HenrikJannsen]

Translation Tool

Bisq uses AI-powered tools to translate application strings into supported languages. This process is highly automated, ensuring that translations stay up to date with each release.

This work has been done by @hiciefte.

---

AI Support Agent

@hiciefte has also begun working on an AI-based support agent trained on both static Bisq content (e.g., the wiki) and dynamic content (e.g., Bisq 2 support communications). This agent could be integrated into Bisq 2 as an additional support channel alongside human support agents.

It is important to clearly communicate that this is an AI assistant, not a replacement for human support. Its goal is to complement the support team by handling repetitive or common queries, thereby reducing workload.

---

Ideas for Further AI Integration in Bisq

Here are some potential directions for using AI to improve the Bisq user experience:

• AI-Powered Trade Wizard
  Users describe what they want to trade in natural language, and the wizard suggests optimal trade offers or market paths.

• Real-Time Translation of Chat Messages
  Automatically translate chat messages to overcome language barriers.

• Risk Assessment
  Use AI to assess a trader's risk level based on behavioral and historical data.
  Note: This comes with challenges—particularly how to implement it in a privacy-preserving and decentralized way.

• Enhanced Reputation Scoring
  AI could generate more nuanced and accurate reputation scores by analyzing more complex user behavior patterns.

• Trade Bots
  Automate trading using bots—especially useful for altcoin trading. For fiat trades it would require that fiat payment methods provide APIs.

---

Novel Use Cases

In the early days of the internet, most applications were simply digital versions of real-world services—like email replicating postal mail. The truly transformative, internet-native applications came later, with innovations that had no direct analog in the physical world—think Bittorrent, Twitter, or Bitcoin.

Similarly, in the context of Bisq and AI integration, we should ask:

• What are the "AI-native" use cases that go beyond mimicking traditional features?
• Which ideas could be the next killer app—far beyond mere "nice-to-have" features?

---

Challenges

As mentioned above, privacy protection remains a significant challenge for many potential use cases.

Another key issue is integration within a decentralized system. Current AI models are too large to be embedded directly in the app. At present, there are two viable approaches to handle the resource demands of AI systems:

• Remote hosting via a server (accessible through Tor). This service could be operated under a bonded role if trust minimization is needed.
• Local AI integration, connecting to a user's own toolchain. I’m not familiar enough with the current landscape to assess how feasible this option is. It would likely be limited to users who are both capable and willing to run a local AI setup. However, similar to running a full Bitcoin Core node, this approach empowers users who prioritize privacy and decentralization—while the hosted model remains a more convenient fallback for others.

---

Developer-Focused AI Use

AI-assisted development (e.g., code completion, code review) is already being used and explored by Bisq developers. However, these are outside the scope of this discussion, which focuses on user-facing AI applications.