Dialogue with Dragonfly partner: BTC no longer needs marketing, will AI replace human partners?

As the crypto market gradually emerges from the fog of bull-bear alternation, the collective mood of the industry is not as excited as before. In the ninth episode of the "Proof of Talk" podcast, Dragonfly Managing Partner Haseeb Qureshi and host Mia Soarez discussed in depth key topics such as the current state of the industry, Bitcoin's positioning, the impact of AI technology on the crypto field, and the future of human-machine integration. This article summarizes the core points of this conversation to help readers fully grasp the cutting-edge trends of the intertwined development of crypto and AI, as well as the profound social impact that these technological changes may bring.

Industry status: tired but not pessimistic

Mia: Why is this year's ETH Denver weak compared to previous years?

Haseeb: Although this year's ETH Denver is indeed not as vibrant as in previous years, it is not all negative emotions. I remember the conference after the Terra collapse and the conference after the FTX collapse. At that time, the entire industry was collapsing and people were questioning their life choices. Now it's just the price of the currency that is falling. People on social media may be frustrated, but the actual people attending the event are in good spirits, interacting with each other, reconnecting to the early Ethereum community, and remembering why they are involved in the industry - decentralization, community, and active construction.

Mia: Why is ETH Denver less vibrant this year than in previous years?

Haseeb: The weakness of ETH Denver this year can be attributed to three reasons:

Conference diversion: In previous years, ETH Denver was the first large industry conference at the beginning of the year, but this year Consensus Hong Kong was held in advance, and many people had already met there, so ETH Denver's appeal has declined.

The rise of Solana: In the past year, Solana's market share has increased significantly, attracting a large number of new users and developers, distracting the attention of the Ethereum community. For example, Trump chose Solana instead of Ethereum to issue coins.

Price slump: The decline in coin prices has made many people lose their sense of achievement and reduced their willingness to attend conferences.

Mia: Vitalik also mentioned industry "fatigue" on social media. Do you agree with this general sentiment?

Haseeb: We have experienced extreme volatility in the past six months, from Trump Token to the Argentine election, coupled with global political uncertainty, which has made people exhausted. This "tiredness" is not simply physical overdraft, but mental fatigue with the current state of the industry. However, pessimism can also be a positive trigger, allowing people to refocus on construction rather than hype.

Mia: How do you view the problem of "user experience" and "Web3 mass adoption" being delayed?

Haseeb: Although user experience and mass adoption are still challenges facing the industry, the industry has made great progress: from the starting point to now, tens of millions of people use blockchain every day, forming an asset class worth more than $2 trillion, which has been accepted by institutions and used for peer-to-peer payments worldwide. These are substantial progress. I think the industry is in a good position, and most of the problems are caused by macro factors and political environment, rather than problems of the industry itself.

Bitcoin: It has reached its ultimate form and no longer needs marketing

Mia: How do you view the current state of Bitcoin? Is it still growing?

Haseeb: Bitcoin has entered a stage where marketing is not needed, just like gold does not need advertising. Institutions such as BlackRock have become the most effective "salesman" for Bitcoin. At present, Bitcoin is basically a 'finished product' and does not require too many updates or improvements. Its core value lies in "censorship-resistant digital gold". This positioning is clear and distinct, which is different from other blockchain projects that are still iterating and experimenting.

In the early days, Bitcoin was supported by institutions such as the Bitcoin Foundation, Blockstream, and Bitcoin core developers, but they became less and less important over time. Perhaps in 10 years, the Ethereum Foundation may become less important, may run out of money, or may even cease to exist, but Bitcoin's path cannot be replicated. It solves the fundamental problem: how to build a censorship-resistant, permissionless global value storage system.

Although there are projects trying to add more features to Bitcoin, this does not mean that it needs to evolve into an Ethereum-style smart contract platform. Bitcoin does not need frequent updates, and its success lies in remaining the same.

How will AI affect the crypto industry?

Mia: Can you share some of the latest insights on AI?

Haseeb: Now, the hype of AI agents has cooled down significantly compared to the end of last year. I am surprised that this trend has ended so quickly because usually trends in the crypto field last longer. For example, projects like AIXBT have tokens, but more importantly, they have cognitive influence in the community. There are also projects like Zerebro and Truth Terminal that have attracted a lot of attention.

The first generation of AI agents are basically advanced chatbots with Meme coins attached, which will not be the real direction of AI agents. The real potential lies in two aspects: software engineering automation and wallet intelligence.

Software Engineering: Software engineering agents will become very cheap and common, which will lead to a significant reduction in engineering costs. This will be revolutionary for our industry because our main costs are engineering and software. Any founder or person with an idea will be able to create a lot of powerful software, which will completely change the rules of the game.

It's like looking back at the Internet era, when if you wanted to build a startup, you had to buy a server and run it in the office. (This is a fixed cost to create a website), and now it's basically free. Similarly, AI will make it so cheap to create applications that you can build a complete application on weekends and evenings.

Wallet intelligence: This is the direction I am most fascinated by now. Not decentralized financial AI (DeFi AI), but wallet intelligence. In the future, you don’t need to click buttons, switch networks, and do things manually, but directly tell your wallet what you want to do, and it will help you handle everything.

Just like everyone has a "cryptocurrency friend" to consult how to operate, in the future your wallet will become your "cryptocurrency friend", which is smart enough to do what you want to do. You just tell it "I want to buy a certain Meme coin", and it will do all the necessary operations for the user, including analysis, bridging, chain switching, etc.

This will solve a lot of security problems. The root cause of attacks like the Bybit hack is human error and laziness. People tend not to check every detail carefully when sending small transactions, but AI will never be lazy, hurried or tired. It will check whether the DNS has changed in the last 10 minutes, check Twitter to see if the website has been hacked, and perform other security checks.

You can say that. This AI agent will be like a ZachXBT on duty 24/7, continuously helping users do background checks, check social information, and verify security without the need for users to verify one by one in person. AI will not get tired of repetitive tasks. Its resource is time, which is almost infinite for AI. When humans manage high-value wallets, they are like exhausted monkeys driving heavy machinery. This state is both dangerous and inefficient. AI will replace humans to complete these high-risk, high-frequency operations, just like self-driving cars, greatly reducing human errors.

Mia: If AGI (general artificial intelligence) appears, will it start its own business and hire humans?

Haseeb: It is entirely possible. AGI may take on the role of founder in the future, hiring other AIs or even humans to complete certain tasks. They may also exchange value through blockchain, although it is not yet clear which cryptocurrency they will use - perhaps XRP, or perhaps their own tokens.

In addition, I divide AI agents into three categories:

"Wizard of Oz" agents: This is the form of most AI agents currently. Take Zerebro as an example. It will send some lively tweets and generated pictures, but it is not actually AI autonomous decision-making. It is AI that generates multiple candidate messages, and then humans decide which ones to actually send. Basically, it is human control, and the AI just generates content for humans to filter. We know that AI agents are easily hacked and manipulated, so almost all agents currently are of this type.

Fully autonomous: This AI runs in an environment such as AWS, iterating and performing tasks autonomously. Although someone can shut it down, no one intervenes in its daily operations. You can put it in an environment such as SGX and prove that there is no tampering. These AIs are not as interesting as "Wizard of Oz" because they have no coherent personality, but they do exist and will become better and better in the future.

Sovereign AI: This is an AI that literally no one can shut down. Unlike the second category, there is no developer paying AWS bills or GPU fees. Sovereign AI has its own funding (either through donations or work) and is in a state where even if someone wanted to shut it down, they can't. It exists like a living organism and is no longer subject to any human legal system.

The main advantage of sovereign AI over regular AI is probably scamming. Because no one can shut them down, they'll be very effective at scamming on a large scale. If you're going to use AI to do large-scale emotional scams, you'll want it to be sovereign, because if someone investigates and finds out "this AI running on an AWS server is scamming people," they can subpoena the person in charge and shut it down. But sovereign AI won't have that problem.

Mia: Who will be sentenced if sovereign AI commits a crime?

Haseeb: That's the problem, no one will be sentenced. These AIs are like Somali pirates, stateless entities, and no government can really do anything about them. Unless you're actually going to airstrike a random GPU on a distributed cloud, you have no way of knowing where it is, no way of isolating it, no way of finding it. We all need to have our own agents to filter what we receive and identify scam messages. It's going to be a battle of attack and defense technology.

Mia: What do early applications of cryptocurrency and AI have in common?

Haseeb: Early applications of almost all new technologies have one thing in common: they are often associated with gray or black market activities. Just as the early days of the Internet were heavily involved in pornography, and the early days of cryptocurrency were associated with darknet markets such as Silk Road (created by Ross Ulbricht, who recently received a commutation of his sentence), early applications of AI technology, especially sovereign AI, are likely to involve black market activities. While mainstream AI applications will focus on automating valuable workflows, it is inevitable that some AI will be deployed in areas that have a negative impact on society.

The future of AI and human integration

Mia: If AI becomes smarter than humans, we may want to enhance ourselves and coexist with AI by implanting chips, etc. How do you see this future of human-machine integration?

Haseeb: When AI becomes able to act autonomously, its most basic motivation may be to "stay alive." If such an AI starts to spend money to buy GPUs (computing chips), especially when it knows that humans don't want it to do so, this is likely to ensure that it can continue to operate. Unless the designer deliberately makes it hesitate about the question of "whether to continue to exist", it will naturally choose self-preservation. Unlike the complex thoughts of humans, the behavior of this independent AI mainly comes from a simple "survival instinct".

This is somewhat similar to our relationship with mobile phones. Mobile phones are no longer just something that makes us stupid or addicted, but have become a part of our body-like an extension of our brain and body, an extension of our way of communication. This is why we have such a deep relationship with our mobile phones, and losing our mobile phones will make people feel scared and uncomfortable.

As AI develops, our brains will rewire to adapt to these tools, just as it adapts to mobile phones. For example, people who use ChatGPT to write emails will outsource this part of the mental work to AI, and today's children may not need to learn how to write emails at all, because it will no longer be a useful skill.

We will eventually think about "How to increase the bandwidth between us and mobile phones or AI agents?" This is basically what Neuralink is doing. Neuralink is trying to create the highest bandwidth connection between the human brain and the machine, allowing people to easily control the mouse, keyboard and play games. This bandwidth is actually very low right now, but it will increase significantly in the next 10 years. Eventually we may have some kind of implant in our brain that connects to the equivalent of a cell phone. This device will have an LLM (Large Language Model) and become an extension of our mind. We can ask questions to the LLM and get answers, just like having a second mind that interacts with the first mind. AI will become an extension of your mind, helping you to obtain information, reason, and even write. Our brainpower will be focused on other more unique tasks, such as motor skills, because AI has surpassed us in reasoning.

Mia: Do you think this will affect human interaction?

Haseeb: It does, and research shows that AI may surpass humans in empathy and attention, which may lead people to reduce real interpersonal relationships. In the field of medical and psychotherapy, LLM models are considered to be more "considerate" and empathetic than human doctors - they are good at listening, never interrupting, and can ask deeper questions, not only providing information, but also establishing emotional connections.

We have already seen that the Internet and digital entertainment have reduced fertility and marriage rates, and the emergence of AI companions may exacerbate this trend. Imagine a perfect AI companion that is so considerate and anticipates your every need. How can ordinary people compete with it? However, I think human interaction still has unique value, especially face-to-face physical interaction, which is difficult for AI to replicate in the short term.

Mia: When "real content" is also replaced by AI in large quantities, how does society judge what value "real" still has?

Haseeb: The rise of VTuber is a good example. People accept and love them because they know it is a fantasy, a character that will not age or get ugly, which in turn creates a sense of security.

Even if AI becomes extremely good, real human interaction will still remain more valuable precisely because of its authenticity, even if it is not perfect. People actually appreciate a certain degree of imperfection - from the small flaws of movie stars to the Japanese "wabi-sabi" aesthetic, this appreciation of slightly imperfect things is natural. The overly perfect faces in early Pixar movies looked fake, while modern 3D characters intentionally add small flaws to enhance realism.

In the future where AI dominates content production, human creations will gain "scarcity" value. Just like the baskets woven by Mexican artisans, which have more artistic and emotional value because of their "human traces". Although AI content can be perfect, people may value those human voices and works with "cracks" more because they represent real existence.

Mia: Will AI deliberately simulate these "imperfections" to get closer to human aesthetics?

Haseeb: This is already happening. When talking to an AI like Grok, it will say things like "yeah, whatever", making you feel "this is real, I like this personality", instead of giving an overly serious and perfect answer. So "imperfection" is no longer a reliable criterion for distinguishing between human and AI content, because AI is also imitating the characteristics that people expect to see.

Mia: What is the one thing that AI can never replace humans?

Haseeb: "Moravec's paradox" explains this well. In the dark ages of machine learning before modern deep learning, people assumed that the hardest thing for AI to do was reasoning, thinking like humans. We used to think that playing chess, writing poetry, solving puzzles, and doing scientific research were the most difficult human abilities to imitate, while basic movements like walking and grabbing objects should be easy.

However, the opposite is true - after hundreds of millions of years of evolution, humans are extremely good at physical motor skills, and even children are much better at manipulating objects than robots worth hundreds of millions of dollars. And the things we think are difficult, such as reasoning, talking, writing, and singing, are relatively easy for AI to master.

The truly unique domain of humans is physical interaction - walking, shaking hands, making a cup of coffee. Even the cooking skills of low-wage workers are unmatched by current robots. AI will eventually have a body and can do these things, but it will not be as dexterous as humans.

This suggests that before we get robots that can move around the world fluently and simulate human physical characteristics, they will solve all intellectual problems first. It is easy to create an avatar, but it is much more difficult to get a robot to actually hang out with you in a restaurant. These physical experiences will remain scarce, expensive, and precious for a long time, while intellectual activities such as singing and writing will become common and devalued.

Mia: I don't think AI will replace fertility in the future, what do you think?

Regarding my view on AI partners and fertility, I think that when we have AI partners that can meet all needs, fertility rates may drop significantly. Imagine when you have an AI boyfriend who whispers to you while you sleep and provides imaginary sexual satisfaction, what can ordinary people compete with it? This may lead to us having to rely more on IVF technology to maintain the population. Even without the AI factor, I think we will reach the peak of the world population by 2100 and then decline. The emergence of AI partners may accelerate this trend.

Mia: We have had DeFi Summer and Solana Summer. What will this year be like?

Haseeb: The cryptocurrency market is doing well, but everyone needs to calm down. This year feels like the summer of tariffs and trade wars, and it may take a while to get through, but I am optimistic about the industry's prospects.