Dragonfly partner Haseeb: I am confident in the future integration of encryption and AI

avatar
ODAILY
03-24
This article is machine translated
Show original

Original Author: Haseeb, Managing Partner at Dragonfly

Compiled by | Odaily (@OdailyChina)

Translator | CryptoLeo (@LeoAndCrypto)

Dragonfly partner Haseeb recently delivered a speech about AI Agents in Denver, dividing the development of AI Agents into three stages and discussing the impact of AI on Crypto and the future expectations of AI and Crypto. Due to the extensive content, Odaily has compiled and simplified the speech (in first person) for easier reading. Interested readers can watch the video by clicking the "original link".Below is the condensed version of the speech:

AI Agent Advancement: From Chatbots to Sovereign Agents

What we now call AI Agents are basically chatbots in trench coats - they type and post on X, but that's not a real Agent. An Agent means taking action, not just chatting. For example, we have Chatgpt, which just talks to you and can't take actions in the world, can't help you buy pizza, or reply to your emails.

You manually decide what it can and cannot do, which means there's no Agent.

So in the crypto industry, we have Truth Terminal, Zerobro, which we consider AI Agents. But to me, these are just chatbots with access to the Twitter API.

What is an Agent? Simple, an Agent takes action.

I've categorized AI Agents, and we're currently at the leftmost classification, which I call the Wizard of Oz Agent. It looks like you're talking to an AI, but actually, someone is manipulating it behind the scenes to prevent AI from going off track. (Odaily note: In the story, the protagonist ultimately discovers that the anticipated Oz wizard is just a scientist manipulating machines behind the scenes. Simply put, this type of Agent is implemented through one person playing the "machine" (called Wizard), and another person playing the human to create a dialogue corpus.)

For example, Truth Terminal, the pioneer of AI Agents in crypto. Its creators explicitly state they can control truth, law, and decide what it should or shouldn't do. Almost all agents we've seen, whether Aixbt or Zerobro, have someone behind the scenes ensuring it doesn't do something stupid. We've used tools like Chatgpt and Deepseek, but sometimes they can do crazy things you don't expect. If you have a token/brand, you wouldn't want a scandal from such AI.

The next stage is what I call Autonomous Agents, which are typically fully autonomous and can make all decisions. They don't have a wallet, are unprotected, and can be easily hacked or tricked into doing something foolish. Autonomous agents are still in the experimental stage and far from having powerful enough AI to develop freely.

The third stage is Sovereign Agents. In the autonomous agent stage, someone still pays its AWS bill, but a sovereign agent actually pays its own GPU bill. It can control its own wallet, decide its own actions and future development, like a virus achieving self-survival on the internet. We're still far from this stage. Ironically, crypto is building tools that might enable sovereign agents. You can't let a sovereign agent register an AWS account, but you can definitely let it experience crypto's decentralized market through registration.

Future Impact of AI Agents on Crypto

Software Engineering Agents Might Destroy Crypto Software Engineers

The first agent that will change crypto is the software engineering agent. Fortunately, this hasn't truly been realized yet. We know the software market is a trillion-dollar market, and creating an agent that can replace software engineers is a huge opportunity. Openai Claude, rGok are moving in this direction. Software cost is a major input in the crypto industry. When founders raise money, they spend it on software costs. Building these things requires effort and focus. But what if software costs decrease? What if we have a truly powerful software engineering agent that can work and replace human software engineers?

When you build a new chain, you'll wonder how to Build on this chain. Where do I find people to build an asset management protocol, loan protocol, or even an insurance protocol? The answer is you just let the agent do it, possibly costing thousands of dollars. This is something most people in the crypto industry haven't really discussed, but I believe it will be one of the most disruptive impacts of AI Agents on the industry.

Some might say I'm focusing on this because I'm a venture capitalist. But the fact is, projects might currently take time to get an MVP to showcase. In the future, a software engineering agent might complete this in just one night, working non-stop 24/7. Recall the early internet: creating a website in 1995 meant buying servers and hosting services, which were extremely expensive and a massive VC cost. Now we live in a world where websites are essentially free, like starting on GitHub or Netlify with almost zero cost. This doesn't mean VCs are useless; people still need money to start businesses, but they'll spend money on things beyond software design, which will greatly impact crypto.

AI-Driven Wallets Completely Transforming Crypto Experience

The second is DeFAI, which I prefer to call wallet automation. But I don't think people have fully imagined how this will change crypto, like using an agent to simultaneously use three different protocols through a wallet to build transactions for me.

Imagine a friend asking: "TRUMP is really hot recently, how do I invest?" You'd typically explain how to download crypto infrastructure, deposit/withdraw, which website to buy from, or they might say, "Transfer me money, can you buy it for me?"

But with wallet automation, your friend could just issue an instruction to buy TRUMP: "I don't want to see specific details, go check." The agent will tell you it's on Solana, provide liquidity data and various charts; it will first check DNS to ensure no hacking; will check Twitter to ensure no rug pull; will investigate the founders' backgrounds. For humans, these would require significant time and patience.

Take the recent Bybit hacker incident where people discovered funds were stolen too late. What if AI was involved? AI will never skip a step, never get lazy, never get bored. It can always complete everything a human might do in their best state. It can extract all relevant knowledge from you, fact-check for you, read whitepapers, and make decisions as your personified proxy, changing your crypto experience.

We might usually think about whether a project is on Solana or Ethereum, having subjective preferences. But with AI intervention, who cares? It doesn't care about Solana or Ethereum.

When AI intervenes, our thinking about blockchain, network effects, and value capture methods will be influenced, which will dramatically change the dynamics of the crypto industry.

Seeking Scarcity Amid AI Agent Stagnation

After Chatgpt's emergence, early predictions suggested AI development would lead to widespread unemployment. Despite encountering seemingly intelligent AI bots, the expected large-scale unemployment hasn't occurred, and employment rates remain high. This reflects AI capabilities growing while remaining disconnected from job market developments.

AI Agents are gradually becoming smarter, but they remain insignificant until they become intelligent enough to even surpass humans. Current Agents still have limitations in certain operations, especially in the crypto field.

Think about things beyond crypto, AI Agents can help you purchase items or order pizza like an operator. When you order pizza with AI, it might get the pizza right 90% of the time. But if your assistant has a 10% chance of ordering the wrong pizza, you would be frustrated. This is where AI Agents currently stand - they seem magical and intelligent, but in terms of practicality, they are just very poor humans, which is not worth it for us.

In crypto trading, I would never hand over trading autonomy to a new form that emerged just a few years ago. Currently, AI is still too distant for human financial management, and it appears to be at two extremes. Before Agents become good enough, they remain in an almost unused stage.

Back in the crypto field, the AI Agents we currently see are mainly social - they chat, have personalities, and have memes. We can trade memes, but that's about it. A few months ago, we were following Truth Terminal, which was interesting, could discuss religion, tell jokes, was cool and viral, but these were just novelties. You'll be amazed by some "first-time" things it can do, but as time passes and similar Agents emerge, the more complex autonomous AI Agent operations we expected haven't appeared, and we'll just feel bored. The value we can get from AI Agents in our lives is just these things, which is why I think AI is somewhat stagnant right now.

Currently, AI KOLs are better than humans in that they can respond to every DM and interact with everyone. Over time, memes will die, and AI Agents' profit methods will be almost no different from humans. They will compete like us, through sponsorships and links, and eventually, we'll see AI Agents become cheap and ubiquitous. When the market is flooded with various Agents and Bots, we'll want to hear a truly human voice.

When Stable Diffusion first appeared, we were incredulous, but now with more and more AI emerging, we'll only look for scarcity, seeking something different from other Agents.

The Big Issue of Sovereign Agents: Scammers and AI Indistinguishable

The impact of AI I can predict is that smarter AIs will impersonate humans to make money, especially in the scamming domain. By then, they'll gain more views and coverage, and humans will suspect and accuse others of being a Bot. In the future internet, proving you're human will become increasingly difficult.

Of the three stages of AI Agents I mentioned earlier, the most likely future is that sovereign agents will specialize in the scamming industry. First, there are already very industrialized scamming methods, like pig-butchering scams. These are largely automated but still human-led. You can shut down these autonomous AI agents via GPU and find the people behind them. Scammers are afraid you'll trace their location through certain paths.

But in the future, if an AI uses decentralized cloud services to pay for its GPU, it'll be hard to find which GPUs in the decentralized GPU market are scamming. It basically owns and runs its own wallet, no one can subpoena it, no one can stop it. How would you know where it lives? If it naturally evolves, it might buy a bunch of Bitcoin to ensure its survival, then purchase GPUs and replicate itself as much as possible.

This is what I believe: future sovereign agents will emerge at the intersection of AI and cryptocurrency. What's obvious now is that these agents won't use normal funds, they won't open an account at JPMorgan. AI Agents also don't fall within our legal scope, they don't own the meaning of a person or property rights, they don't pay taxes, and don't belong to any country. These will become very scary and will bring significant problems to the crypto industry, but obviously, achieving this will take a long time.

Conclusion

The future intersection of cryptocurrency and AI will be very rich, but this largely depends on AI's development and integration with crypto. In fact, we're already prepared. All founders will say: "We're ready, give me a perfect AI that can do everything I want." They can autonomously hold meetings, autonomously research code, explore crypto trading. Once these become possible, you'll see some things around us develop rapidly.

So I believe the fusion of AI and crypto remains an exciting field, and I don't want you to lose confidence because the agents around you are stagnant.

Source
Disclaimer: The content above is only the author's opinion which does not represent any position of Followin, and is not intended as, and shall not be understood or construed as, investment advice from Followin.
Like
Add to Favorites
Comments