In this online exchange called "Trench Talk", many pioneers from the AI and blockchain industries had an in-depth discussion on the current status of artificial intelligence (AI) agent technology, the combination of DeFi and AI, and the needs and development prospects of different B2B and B2C application scenarios. The main participants of this dialogue include:
Kunal: From Bangalore, India, focusing on providing marketing and AI agency services for Web3 projects;
NFTs (also known as NFT2): With multiple experiences in AI and crypto industries, it is currently committed to building a decentralized data training platform to provide large-scale annotation and model training for AI companies;
Abdul: works at DecaSonic Fund, focusing on investing in early-stage projects at the intersection of Web3, blockchain, and AI;
Other participants also shared their insights on topics such as 'AI agents', 'DeFi finance', 'B2B and B2C applications'.
This article will summarize the dialogue as comprehensively and thoroughly as possible around several core issues.
One, the current capabilities and future evolution of AI agents
1. The 'ceiling' and pain points of existing AI agents
Participants generally believe that AI agents are still in a rapid iteration phase, with specific capabilities mainly in dialogue, text generation, some visual tasks, and relatively simple decision support. Compared to traditional 'scripted' or 'decision tree' algorithms, modern AI agents have stronger natural language understanding and autonomous reasoning capabilities. However:
Creativity and abstraction ability: At the level of fully autonomous 'deep creation', AI tends to easily fall into 'self-circulation' or repetitive application of existing training data, making it difficult to truly break through human 'inspiration' or 'soul'-like creation.
Multi-source data processing: To enable AI agents to truly excel in complex decision-making or strategy execution, it is necessary to connect and comprehend complex information from multiple domains (including real-time API calls, compatibility with different industry data, etc.), which still faces technical and computational barriers.
Factual and 'hallucination' issues: AI often exhibits 'hallucinations' in text generation or dialogue, fabricating non-existent cases, data, or literature. This type of output, which is highly persuasive yet potentially severely distorted, harbors risks in real-world applications.
2. Moving towards AGI: Full of imagination yet full of challenges
The pursuit of AGI (Artificial General Intelligence) has been repeatedly mentioned:
User expectations: Many users hope that one day, AI agents can act like an 'omniscient and omnipotent' human brain, excelling not only in limited fields but also integrating knowledge across fields, and even deriving new ideas and theories.
Technology and ethics: The path to AGI not only requires solving computational power, data, and algorithm iterations but also must consider moral and safety issues, such as how to keep AI agents controllable and traceable in complex social environments.
Two, AI applications in B2B and B2C: Differences in functions and pain points
1. B2B: Enterprises desire to improve efficiency and automation
In enterprise (B2B) scenarios, AI agents mainly focus on:
Process automation: Assigning the most cumbersome or rule-based processes within enterprises to AI agents, such as project management, financial settlement, document review, etc., greatly improves office efficiency and reduces human errors.
Professional data integration: Many enterprises possess vast amounts of historical data and proprietary information, and if deep training and excavation can be conducted using AI agents, more accurate decision support or predictive capabilities can be obtained, gaining a competitive edge in the global market.
Innovative business scenarios: such as empowering gaming companies to use AI to generate more realistic NPC dialogues, or assisting marketing agencies in quickly producing vast amounts of high-quality content. Some companies are also embedding AI agents into systems like CRM and ERP, allowing for highly collaborative internal information within enterprises.
2. B2C: Consumers need a simple, user-friendly, and 'accompaniment-style' experience
Compared to the complex demands on the enterprise side, ordinary consumers engage with AI in a more direct and emotional way:
Information acquisition and search replacement: Many people have gradually started using LLM dialogue models like ChatGPT and Claude to replace traditional search engines. The instant feedback of 'question and answer' eliminates the need for individuals to search multiple sources.
Creative assistance and entertainment: AI-generated images, videos, and music creation tools allow non-professional users to quickly generate relatively exquisite works; some users also experience the potential of AI in education and companionship.
Emotional companionship: In the context of the evolving 'AI large model + multi-modal interaction + metaverse', AI may become human's 'virtual friend' and 'digital companion', especially as feelings of loneliness and stress rise in contemporary society, AI may fulfill part of the psychological companionship needs.
Three, the rise of DeFAI: The significance of the integration of AI and DeFi
1. From traditional 'quantitative trading' to 'agent-based financial management'
When discussing the combination of AI and decentralized finance (DeFi), several guests mentioned:
Trading and arbitrage: By training AI agents to execute smart matching, cross-chain arbitrage, liquidity mining, and other operations on various decentralized exchanges (DEX), significantly enhancing yield and capital efficiency.
Automated financial management: Ordinary investors who are not familiar with complex contracts and liquidity pools can entrust their funds to 'DeFAI agents', simply setting risk preferences or target returns, and AI can continuously 'self-optimize' to diversify or concentrate investments to match user expectations as closely as possible.
Diversified portfolios and risk control: AI agents can track various indicators (macroeconomic, on-chain data, etc.) in real-time to make dynamic allocations. They can also discover previously unnoticed investment opportunities or potential loopholes through backtesting models.
2. Opportunities and challenges coexist
Inclusive finance: If every user can obtain professional financial strategies through AI, DeFi may truly reach the masses, breaking down traditional financial barriers;
Compliance and safety: If the 'hallucination' of AI decision-making leads to significant financial losses, the responsibility and consequences remain unclear; regulatory measures and legal boundaries for AI economic behavior still need to be explored.
Four, to be replaced? Or embrace the 'super suit'?
1. 'Concern about unemployment' VS 'Tool upgrade'
Many guests mentioned that the fear of AI replacing human labor exists across various industries. However, similar to how the computer revolution replaced some clerical jobs but also created more software and internet-related new jobs, AI will also bring new modes of 'human-machine collaboration'.
Simplifying repetitive tasks: A large number of mechanical processes such as administration, personnel, data labeling or screening, basic customer service, etc., may gradually be taken over by AI.
Releasing creativity: For individuals with technical or artistic talents who know how to use AI to enhance productivity, AI acts like a 'super suit', enabling personal efficiency and creativity to multiply by several times or even dozens of times.
2. Do humans retain the 'emotional and ethical' advantage?
Participants believe that AI still struggles to truly understand 'human emotions' and 'deep values', especially in situations requiring high empathy, critical decision-making, or facing unforeseen circumstances, the intuition, ethics, and experience of human decision-makers are indispensable.
Fields that cannot be easily replaced: healthcare, early childhood education, specific artistic fields, creative planning, etc., may still require more 'humanity' participation.
Reliance on AI may lead to the degradation of human skills: Just as reliance on GPS has caused people to lose their sense of direction, excessive dependence on AI for various skill executions may also weaken corresponding human capabilities.
Five, how to keep up with industry trends? The 'fast track' of knowledge updating
Given the astonishing speed of updates and iterations in AI, how to keep up with cutting-edge information has become another major topic:
Focus on key individuals and research teams: Whether it’s OpenAI, Anthropic, DeepMind, or newly emerging entities like DeepSeek, it's essential to continuously track their teams, paper releases, and community activities.
Make good use of social media and industry circles: X (formerly Twitter) is considered by several guests to be a treasure trove for obtaining the latest industry dynamics, where many researchers and developers share inspirations or achievements immediately; platforms like Telegram, Discord, and GitHub are also great places for in-depth exchanges.
Check VC investment trends: Tracking capital market preferences can provide insights into the emergence of new tracks and competitive dynamics in hot fields, indirectly confirming which AI concepts or projects have greater potential.
Six, conclusion: New opportunities and unknowns in the AI wave
This dialogue started with the current technology status and evolution potential of AI agents, expanding to the multi-dimensional needs of enterprises (B2B) and individuals (B2C), and then to the 'DeFAI' imaginative space brought by the deep integration of DeFi and AI.
The next breakthrough in AI agent technology: Beyond basic large models, how agents obtain 'specialization' and 'multi-data source' support will be the key competition point, while also addressing core challenges such as 'hallucination'.
B2B and B2C resonance: When enterprises introduce AI on a large scale, they need to handle vast amounts of proprietary data; individuals focus more on 'simple interactions' and 'personalized experiences'. Although their application scenarios may differ, they highly share computing power, technological breakthroughs, and talent resources.
Prospects of DeFAI: Leveraging the advantages of AI automation and intelligent decision-making to provide new growth points and risk control means for DeFi projects, allowing ordinary people to access more convenient, safe, and efficient on-chain financial services.
The era of human-machine collaboration has arrived: Concerns about being replaced are not unfounded, but as guests have said, 'those who can use AI' are the irreplaceable labor force and innovative power of the future; mastering the use of 'super suits' will allow them to maintain a competitive edge.
Currently, we are at a critical moment of deep integration between AI and Web3. Breakthroughs in AI agent technology may not only disrupt traditional search models and efficiency management but also redefine 'role distribution' in the digital economy. As more AI projects, decentralized infrastructures, and enterprise demands join the 'trench', the wave of AI and blockchain may trigger unprecedented social changes. For ordinary people, adapting to this trend, continuously learning, and building self-core competitiveness may be the best posture to face this astonishing transformation.
More information:
Game public chain Xai: Binance Square database
