Vitalik answers everything: AI disaster, Zuzalu experiment, Bitcoin culture, what does he like and dislike the most?
Original: Vitalik Twitter Reply
Compiled by: GaryMa, Wu Says Blockchain
Q: What are your favorite book\game\movie?
A: 1. The foreign Harry Potter fan fiction Harry Potter and the Methods of Rationality;
- Unsong, a fantasy science fiction novel by Scott Alexander, set in a world based on the mysterious Kabbalah.
Q: Have you ever had a supernatural experience or encountered something you couldn't explain?
A: Seeing many of our worst enemies self-destruct, or at least make themselves look very foolish.
Q: Are you concerned that Ethereum culture might follow a path too similar to Bitcoin culture, where the loudest voices are non-technical?
A: I'm not sure if non-technical itself is the problem; more importantly, it's the interplay of (i) non-technical, (ii) overconfidence, and (iii) aggressiveness. I think the latter two traits are more significant. In any case, we should be cautious!
Q: What excites you about the current developments in regenerative/public goods funding?
A: Better provenance tracking in retrospective funding, allowing for the identification and rewarding of the entire path leading to a selective work. Some relevant work has already begun to emerge in this area, and it would be great to see more related developments.
Q: Please answer the following questions, ignoring those that are not so quick, simple, or interesting!
ps: Zuzalu is a pop-up city experiment initiated by Vitalik in Montenegro from March to May 2023.
● What aspects of Zuzalu surprised you compared to expectations?
● Who are the leaders you admire now or in history?
● What daily habits have you developed in the past few years that you particularly enjoy?
● Who do you guess is the writer you read the most in 2023?
● If you had to choose one country to live in for the rest of your life, which country would you choose?
● What do you estimate is the probability of you living to 123 years or longer?
● Is there something you measure or track that most people don't pay attention to?
A:
I think Zuzalu has been a success as an experiment! Many core assumptions have been validated:
● It is possible to organize the whole thing
● People are willing to stay
● Successful idea exchanges
● Integration of zero-knowledge technology works
● Successfully established a positive and healthy culture
The biggest unresolved issue I see is how to balance quality and inclusivity at scale. A multi-layered "community network" currently seems to be the approach I tend to discuss with others (including myself). How to implement it in practice depends on the specific details…
There is a significant divergence on the "jurisdiction" issue: what are your expectations of the government you are in?
Possible answers include:
● Actually nothing, just wanting to maintain unity
● Visa-friendly
● Providing more meaningful freedoms for specific industries (biotech? drones? other industries?)
Zuzalu itself is just a temporary experiment, but the community includes people who care about these three aspects.
However, the latter is inherently more difficult than the former; thus, these three aspects seem to trend towards different types of countries.
Q: (Continuing from the above question) If my understanding is correct, then you make sense. You delegate the responsibility of filtering quality to each community to make scaling more feasible while maintaining a certain level of quality control through careful selection of the communities themselves.
A: Maintaining a certain level of quality control is achieved through careful selection of communities.
I feel that even such a setup is not quite accurate! Rather, anyone can create a community, and there is a concept of relationship between communities that can become closer if they trust each other.
Q: This is a question about Zuzalu: What was the biggest achievement? What improvements will there be next time? When is the next time?
A: This is just one of over 10 ideas, but:
● I hope to see a more coordinated adoption of free/open/decentralized/privacy-focused technologies that goes deeper than just having Zupass and Zupoll.
● Using ETH for payments on L2, ditching Telegram, and using @ethstatus or @skiffprivacy or other alternatives.
The same goes for health. Making it easier for people to access their metrics and personalized advice, providing them with suitable diets and exercise, etc. Making the community a comprehensive health resource.
Q: What are your thoughts on the ultimate state of L2 governance? Will we forever remain in mechanisms like token voting, "security councils," multi-signatures, etc.? This seems far from the ideal of cryptocurrency governance (for me, the ideal refers to: "voting based on the node software I run," plus immutable smart contracts).
The same question can also be considered from another angle: Is it worth (eventually) fixing a single L2 (Rollup) as part of the core client code, subjecting it to the governance mechanisms on the Ethereum chain rather than the governance of individual L2 projects?
Because if we cannot incorporate scaling solutions under the same governance mechanism, people like me who are cautious about mechanisms like token voting may be drawn to monolithic L1 projects.
For core functionality, governance is the only thing that should not be delegated from the underlying base chain.
A:
Implementing ZK-EVM verification internally in Ethereum is definitely an idea worth exploring!
I tend to achieve our goals with minimal L1 fixation. You can enshrine the entire L2, but if you fix ZK-EVM verification and allow people to use it in any way, you might achieve 90% of the benefits with three times the flexibility.
In particular, I expect the non-ZK-EVM part of L2 to be small enough to be formally verified.
I hope we can enter the second phase in the coming years, where security councils will only adjudicate in cases of divergence between two different proof systems.
ps: In response to concerns about the centralization of security councils, Vitalik believes that security councils with cross-institutional participation and (importantly) a 75% threshold, trusting the code before reaching that threshold, represent a fairly high standard of decentralization.
Q: What is your longest running record? What is your longest walking record?
A: Running: 44 km, taking 4 hours, pace 5:26 min/km.
Walking: 113 km, taking 23 hr 12 min.
Q: Do you listen to music/podcasts while walking?
A: Dan Carlin's Hardcore History series, especially the first four episodes about World War I.
Q: Will you continue to donate to Ukraine?
A: Yes, I have donated in several places this year.
Q: What are the top five to ten catastrophic scenarios you think are most likely to occur for humanity in the next century?
I guess this question relates to whether humanity will face further climate and planetary resource-related issues. Are you optimistic about global coordination?
A:
● Artificial Intelligence (AI)
● Engineered super pandemics
When it comes to extinction-level disasters, nuclear war is one of them. Unstoppable AI-enhanced authoritarianism is also concerning.
Climate issues might rank fifth or close to fifth. It's very important to push for advancements in solar energy, nuclear energy, etc.
As for "resources will run out" type of issues, I'm not too worried; as resources approach depletion, items will become more expensive, and we will find alternatives.
Q: What are your thoughts on the current state of Bitcoin Cash?
I remember you recently mentioned in a public comment that Bitcoin Cash largely failed due to being overly focused on schism rebellion. However, has the stability over the past 2.5 years and the current unity of the community made you reconsider?
A: I do see cultural improvements! Good luck to you all.
Q: What do you think are the latest research directions in P2P networks related to Ethereum that are worth emerging researchers/research engineers' time?
A: Making the existing Ethereum P2P layer more robust (including protections against intentional censorship, firewalls, etc.) may not be very exciting, but it's very important.
Q: What are your thoughts on the idea of DAI fully moving off-chain (more support for RWA physical assets) without insisting on decentralization?
A: I think there is some reasonableness in this design space, but it needs to be engineered carefully to ensure broad distribution of supported assets and stable governance.
Q: In your opinion, which country is most likely to become the most crypto-friendly and long-term livable cryptocurrency hub?
A: I am more optimistic about small countries than large ones.
Q: Besides cryptocurrency and longevity, what other areas are you most interested in?
A: Wandering?
Q: It is concerning that the AI Act excludes the military from regulation across various fields of AI. It is naive for the drafters of the bill to believe that the military will act according to its own knowledge and beliefs. If there is an opportunity to exploit a technology that could provide a significant advantage, someone will eventually misuse it for personal gain. Even more troubling is the exclusion of intelligence agencies under the guise of regulation, seemingly granting them complete freedom. This is not only morally questionable but also extremely dangerous.
A: Yes, I absolutely oppose AI regulatory plans that assume the military is more trustworthy than others.
Q: Will quantum computing destroy blockchain or make it more secure?
A: Blockchain will have to adapt to QC, but we will be fine.
Q: Does public key encryption exist unconditionally?
A: What you mean by "unconditional" refers to the security of public key encryption not relying on any specific assumptions or conditions. The security of public key encryption relies on the intractability of certain complex mathematical problems. These problems include large number factorization, discrete logarithm calculations, etc., which are considered computationally difficult to solve effectively. However, some mathematical structures, such as curves and lattices, are believed to provide the conditions necessary for public key encryption security, while hash functions do not meet this requirement.
Q: What are the pros and cons of MPC-based EOA wallets versus smart contract wallets?
A: MPC-based external accounts (EOAs) fundamentally have flaws because they cannot revoke keys (resharing keys doesn't count, as the original holder can still recover the key), and I point out that smart contract wallets are the only option because they can solve this issue.
Q: Defects of smart wallets:
1. Only ETH/EVM, networks like Bitcoin are not supported
2. L1 setup and recovery costs are expensive (recovery costs $40)
3. Smart contracts can also be revoked (not always the case)
A:
Yes, Bitcoin needs to upgrade its technology. We have known about this issue for over 5 years.
This is because the overall cost of L1 is high. That's why L2 and smart contract wallets need more protocol-level support.
I'm not quite sure what you mean here?
Q: What is your favorite dinosaur?
A: I like ducks.
Q: What aspects of socialism do you find interesting?
A: I think the focus on the harms caused by private monopolies and resource inequality is important, and I appreciate the willingness to address these issues in a principled way. Additionally, the spirit of internationalism, especially in early socialism, is very good and valuable.
Q: What is the timeline for AGI and p(doom)?
ps: AGI refers to Artificial General Intelligence, also known as superintelligence, which refers to AI systems that can perform at or above human intelligence across multiple domains.
p(doom) refers to the probability of doom, which is an estimate of the likelihood of severe disasters or catastrophic events occurring. In this context, it refers to the probability of severe disasters related to artificial intelligence.
A: For the timeline, the range is very broad. I have 95% confidence that the timeline for ASI is between 2029 and 2200 (I think AGI may remain roughly at human levels for a while).
p(doom) is about 0.1. This certainly makes it necessary for us to take AI risks seriously.
Q: What do you think will be the warning signals? Or what do you think might raise your p(doom) to >10% or <10 years timeline?
A: People have asked this question, and I find it hard to express accurately what important things AI still cannot do. It can be roughly understood as "interacting with other agents in complex environments, formulating and executing plans involving novel behaviors." If AI starts to do this, then I would say we are very close to superintelligence.
I think one important reason my p(doom) is not very high is that I am not entirely convinced by the arguments for rapid takeoff and "no AGI fire alarms" (I know rapid takeoff is not a prerequisite for disaster, but it has played an important role in pushing in that direction!).
If I were convinced by these points, my p(doom) would rise.
Q: If we assume AGI is achieved at some point, what do you think the path to human flourishing looks like afterward?
A: Personally, I hope to see a path that preserves human agency. Of course, this necessarily means that there needs to be higher bandwidth and more frequent communication between technology-supported computation and human will…
Q: The term "bandwidth" just attracts me :)
In your model, does "human agency" mean biological-based?
If it is biological-based, the non-biological side will dominate, and we may face the risks of AI.
If it is non-biological, this outcome is not so pleasant for most people.
A dilemma. Do you see other options?
A: The ultimate form of non-biological (unless synthetic biologists can compete with metal computers, that would be cool). I expect that over time, as this technology becomes more normalized and lowers risks, it will become more accepted, just like all radical technologies.
Q: What do you dislike most about the crypto world, and what drains your energy and motivation the most?
A: As always, what I dislike most are those who think the main innovation of crypto is bringing tokens into various things. Secondly, I think the following points are:
● Massive capital misallocation
● Those who easily stray from principles (ahem, fake libertarians…)
● On the other hand, completely unrealistic purist beliefs ("12-word mnemonics are good enough for anyone!")
● Unnecessary conflicts
Q: What are your thoughts on the latest developments in proof systems (especially proof aggregation, recursion, and composition)?
A: Techniques like aggregation and recursion are crucial for quickly proving very complex statements. For example, we ultimately need to prove the validity of an Ethereum block in less than 4 seconds; you just need to parallelize and combine in some way to achieve this.
Q: Do you think governance systems with non-trivial power can indefinitely remain unseized? What are the parameters of the "timetocapture" function?
A: @SamoBurja would say this is an unsolved problem. I tend to agree. You can get close to a solution, but you have to accept some trade-offs. Having very specific and limited goals (like maintaining price oracles) can help, although even in such cases, problems will arise once extreme situations occur.
Q: What is your imagination of the scale and complexity of a human city on Mars? If by 2050, there are 1 million people settled on Mars, a ticket to Mars costs only $50,000, and Ethereum's transaction volume reaches 60 million per second, would you permanently move to Mars?
A: It depends on the quality of life! (And the final legal jurisdiction, culture, etc., of the Martian city.)
Q: Do you think it is worthwhile to promote a close relationship between the Ethereum community and the Bitcoin community, and if so, how can we achieve this goal?
A: I think a potential area for collaboration is to work harder to jointly support non-blockchain freedom and privacy tools. End-to-end encrypted messaging (without phone numbers), internet anonymity, end-to-end collaboration tools, secure operating systems (like Graphene, Qubes?), open hardware, open virtual reality, etc.
Q: What would be the downsides of introducing EIP1559b, which allocates a portion of burned tokens to household independent validators? For those running a few clients, allocations could increase based on a sliding scale.
A: The protocol cannot distinguish who is a household individual independent validator.
Q: What is the safest cold wallet?
A: Safe Wallet.
