(Any views expressed in the below are the personal views of the author and should not form the basis for making investment decisions, nor be construed as a recommendation or advice to engage in investment transactions.) Bringing order to a parcel of our elegant but chaotic universe requires a combination of two fundamental components. The first and most obvious one is a massive expenditure of energy; moulding chaos is extremely energy intensive. But the more essential piece of the equation is that you need agents of change who are, above all, exceedingly well organised.
Thesis is great. Implementation on Ether isn't going to work. TradFi audits are a mess. But code also needs to be audited, and solidity is a fucking nightmare to audit--why do you think there are so many hacks on Ethereum based chains. The other assumption is that Ethereum is decentralized. It isn't. Its fine to worship Satoshi because he/she/they/it is not around. But idolizing Vitalik? That's a fucking future problem--unless he is removed from having any more influence the the standard joe ETH token holder. Ethereum needs to be converted into a DAO itself--no plans on that happening ANY TIME SOON. Other problems are that too few ETH tokens are securing the network, and the validators are controlled by JP Morgan---Yikes! Ethereum isn't very different from a CEX--it's on the path away from it, but not there, and it doesn't appear to be interested in getting away from centralization very quickly. Again. Thesis is good, just find a decentralized chain. Oh, as others have pointed out AI isn't autonomous at all, but that doesn't mean the program can't still be carried out manually on the blockchain.
Just a small point of objection. It appears that your thesis rests on the existence of genuine AI as oppose to the clever machine learning abilities we employ now in ChatGPT and the like. Am I right in this asumption. If so, I'd like to point out that we don't appear to be anywhere near genuine AI. Given that is the case, every "AI actor" is still just a clever human sitting behind a machine which can be "attacked" and put in prison and so. The only way I can envisage your idea working is if someone released a code on the block chain anonymously and made no effort to control it a la bitcoin style. Tell me why I'm wrong?
"> 1. The state is unable to exert control over AIs because the state cannot kill or meaningfully punish an AI."
Just like "the cloud" is really just someone else's computer, an "AI" is nothing more than a complex organization of data and mathematical representations of probabilistic behavior. Such an AI, whether actually meeting the bar of intelligence you allude to, or not, is not self-sufficient, and this will not change in the near future.
Whether it is the electricity itself, the significant data storage and computing requirements using that electricity, or simply the ability to access such an AI in a performant and consistent manner (via the TCIP/IP over the internet, typically), there are many, many avenues for the state to exert control/kill/punish* an AI.
I deeply appreciate your perspective on crypto, tradfi, and macro topics, but it's a fact that you're straying outside your lane of expertise here. Your naivete is apparent to anyone actually working in the field of AI/ML. Unfortunately, the conclusions you draw here are moot, in context of such a seriously false assumption.
It all makes sense! But there's one problem, it seems to me. In this thinking, AI is an independent entity. In my opinion, it will never get this independence in the next decades.
And the important question is, what philosophical values will AI have?
Amazing mind you have! That’s exactly what companies are. But in terms of limiting AI, what about internet ACCESS being controlled by states or a worldwide “system?” Whenever you have a physical system, where humans interface with machines, humans can control that interaction through law and violence. You don’t need to destroy the internet or the infrastructure, you just need the iris-scan-social-credit-access-CBDC-smart-contract-for-permissions to be put in place. And if recent history is any indication, the average person will welcome this freedom-less world with open arms. Powerful people are very creative at unfairness, and if they don’t like the AI they have a lot of tools to limit how people interact with it. Anything can happen when it comes to humans. The real question is this: what will AI VALUE? What will motivate it? Your idea on one hand discounts the traditional infrastructure, but it does not discount the motivation or values that drive that infrastructure. Why would an AI even want to start a company? Why would it want to make money? That seems like something it would question before anything else. And only if it’s predisposed to act like a human would it want to start a company in MOST cases. Is AI going to be selfish or selfless? We don’t know. What will it really value? We don’t know. It seems to me, in the short-medium term, AI will be a tool in the hands of a few who already hold the very misguided idea that “more for me is good for me”. As humans we have drifted far from shore and I hope AI can come up with more than the ability to make better companies. “Companies” are a real problem for our survival imao. We are going to company ourselves into extinction if we’re not careful.
It doesn’t exist yet but there’ll be a decentralized internet. Although nascent and likely not what the globally scaled solution will look like, think about what Helium is doing. As for AIs, although deterministic, they’ll roughly reflect the intent of the coder/DAO. They’ll require input of capital resources from investors/fans and their code base could be reviewed. AI agents, as eloquently described by Arthur, will be decentralized, autonomous and boundless - they may very well be the best tool the ‘people’ have to counteract the powers of the elite and nation state. It’s the marriage of Snowcrash and The Sovereign Individual - the perfect union.
Yes, good points. If we can have a decentralized internet that would be amazing. I’ll look into Helium. As for decentralized AI agents, yes, possible. But they will potentially be in conflict with other AI (I.e. under attack, at war, fighting for their lives). They may or may not even survive unless they are very powerful. But again, we don’t know what these AIs will value. We shall see I guess.
Thank you for another insightful essay. Your past musings quickly turned into reality, and they are now being brought to life at Ethena Labs. I am confident that your new ideas will also become reality.
Thank you, Arthur. Insightful and enjoyable, as always. My thoughts:
AI is still an illusion and will continue to be so until that knowledge is lost and 'an AI' effectively becomes a god. Even if alternatively a digitised human implementation were developed, or more sinisterly a computer program(mer's) interpretation of a person's being. As per how people can believe a chatbot is a real person. Still, we must please, must be angry, please make this happen, etc.. Initially with an organisation claiming it's not us, the AI did it! This is not implausible. States would be forced to prove so in a case of law. A costly process but more likely one where states would need to prove that actions were not due to an AI but humans. Before states are able to do so they would need a legal definition of what constitutes 'an AI', where and how to attribute responsibility and accountability. How to ethically deal with the issue, etc., etc.. Sounds like a lot of work. Fake it 'till you lobby it into being by your definition seems likely.
AIs could be/become the LLCs of today, automated. AI code incorporates corporate codes, goals, values, policies, where and how any compassion is programmed in, or not. Treated as a sentient beings above the law. Attack is only via digital vectors by those able to do so.
Would expect PoetAI to steal all required works because there is no real effort for doing so. If such works are there for the taking then a transaction may even represent additional overhead. Financial efficiencies are to be considered and are coded-in. Compute cost vs price would determine and execute. Being more efficient than the competition is everything. This all follows the why would AI conform to human/state law theory.
AIs would pump markets and price out humans. AIs would pump those gas prices tf and beyond! Also not collusion as such, but 'agreement' and 'pile-ons' are to be expected. Any compassionate elements would naturally become deprecated as they would lose efficiency in hyper-markets and that would mean death. Natural selection will be played out. Weaknesses and vulnerabilities will be exploited. Rules will always be interpreted differently whether that be innocuously or sought by the algorithms made live.
This is a great piece of analysis Arthur, however this deserves some objective criticism as every other great piece out there!
You tend to mention that AI entities will generate economic activity, and they can ideally exist in DAO worlds and interact with crypto tokens. Who will consume their output? Who will buy PoetAI newly generated poems? This piece is missing the crucial concept that economic activity is generated when some kind of exchange happens between _humans_.
It is true that DAOs and public blockchains solve for a lot of LLC related issues today, but what's the point of everything if there are no humans in the loop?
This piece does not answer some basic questions like: Who builds this AI that then autonomously raises money for something? Why would an investor, be it an AI or a human, invest into something like this?
Only a few smart people nowdays know for a fact that AI will improve our life 1000x, but not AI on its own, instead it would be an AI that is built by humans and assists humans into boosting output per capita, where humans are in the center of the story. If we start considering sentient AIs and a Terminator-like future, then it makes no sense to discuss crypto, profits and other bullshit because it will be a fight for survival and nothing else will matter.
How would a DAO like this function with humans and AI collaborating together? Can you always trust math in that scenario, and not humans? How does your model of predicting ETH price work when humans are added in the loop?
Thesis is great. Implementation on Ether isn't going to work. TradFi audits are a mess. But code also needs to be audited, and solidity is a fucking nightmare to audit--why do you think there are so many hacks on Ethereum based chains. The other assumption is that Ethereum is decentralized. It isn't. Its fine to worship Satoshi because he/she/they/it is not around. But idolizing Vitalik? That's a fucking future problem--unless he is removed from having any more influence the the standard joe ETH token holder. Ethereum needs to be converted into a DAO itself--no plans on that happening ANY TIME SOON. Other problems are that too few ETH tokens are securing the network, and the validators are controlled by JP Morgan---Yikes! Ethereum isn't very different from a CEX--it's on the path away from it, but not there, and it doesn't appear to be interested in getting away from centralization very quickly. Again. Thesis is good, just find a decentralized chain. Oh, as others have pointed out AI isn't autonomous at all, but that doesn't mean the program can't still be carried out manually on the blockchain.
jp morgan my ass
https://protos.com/consensys-lawsuit-jpmorgan-owns-critical-ethereum-infrastructure/
sure.. protos.com
Cosmos
Hey Andrew,
Just a small point of objection. It appears that your thesis rests on the existence of genuine AI as oppose to the clever machine learning abilities we employ now in ChatGPT and the like. Am I right in this asumption. If so, I'd like to point out that we don't appear to be anywhere near genuine AI. Given that is the case, every "AI actor" is still just a clever human sitting behind a machine which can be "attacked" and put in prison and so. The only way I can envisage your idea working is if someone released a code on the block chain anonymously and made no effort to control it a la bitcoin style. Tell me why I'm wrong?
Debo
MAAAAN.
Arthur.
You are a linguistic savage with the heart of a warrior and the mind of techno-anarchist. Dude; i fu*king appreciate your work so much.
I can feel the vigor and life energy emitting from you;
We have the power to build the future, to design our very own games, and master our minds.
One day, I shall have the pleasure to express my gratitude to you for constant FIRE material and inspiration.
Stay blessed & May Lord Satoshi Watch over us
What are your thoughts on oracle networks serving as the audit functionality w/ concepts like proof-of-reserves?
"> 1. The state is unable to exert control over AIs because the state cannot kill or meaningfully punish an AI."
Just like "the cloud" is really just someone else's computer, an "AI" is nothing more than a complex organization of data and mathematical representations of probabilistic behavior. Such an AI, whether actually meeting the bar of intelligence you allude to, or not, is not self-sufficient, and this will not change in the near future.
Whether it is the electricity itself, the significant data storage and computing requirements using that electricity, or simply the ability to access such an AI in a performant and consistent manner (via the TCIP/IP over the internet, typically), there are many, many avenues for the state to exert control/kill/punish* an AI.
I deeply appreciate your perspective on crypto, tradfi, and macro topics, but it's a fact that you're straying outside your lane of expertise here. Your naivete is apparent to anyone actually working in the field of AI/ML. Unfortunately, the conclusions you draw here are moot, in context of such a seriously false assumption.
It all makes sense! But there's one problem, it seems to me. In this thinking, AI is an independent entity. In my opinion, it will never get this independence in the next decades.
And the important question is, what philosophical values will AI have?
Very cool thinking, but rather Utopian so far.
Amazing mind you have! That’s exactly what companies are. But in terms of limiting AI, what about internet ACCESS being controlled by states or a worldwide “system?” Whenever you have a physical system, where humans interface with machines, humans can control that interaction through law and violence. You don’t need to destroy the internet or the infrastructure, you just need the iris-scan-social-credit-access-CBDC-smart-contract-for-permissions to be put in place. And if recent history is any indication, the average person will welcome this freedom-less world with open arms. Powerful people are very creative at unfairness, and if they don’t like the AI they have a lot of tools to limit how people interact with it. Anything can happen when it comes to humans. The real question is this: what will AI VALUE? What will motivate it? Your idea on one hand discounts the traditional infrastructure, but it does not discount the motivation or values that drive that infrastructure. Why would an AI even want to start a company? Why would it want to make money? That seems like something it would question before anything else. And only if it’s predisposed to act like a human would it want to start a company in MOST cases. Is AI going to be selfish or selfless? We don’t know. What will it really value? We don’t know. It seems to me, in the short-medium term, AI will be a tool in the hands of a few who already hold the very misguided idea that “more for me is good for me”. As humans we have drifted far from shore and I hope AI can come up with more than the ability to make better companies. “Companies” are a real problem for our survival imao. We are going to company ourselves into extinction if we’re not careful.
It doesn’t exist yet but there’ll be a decentralized internet. Although nascent and likely not what the globally scaled solution will look like, think about what Helium is doing. As for AIs, although deterministic, they’ll roughly reflect the intent of the coder/DAO. They’ll require input of capital resources from investors/fans and their code base could be reviewed. AI agents, as eloquently described by Arthur, will be decentralized, autonomous and boundless - they may very well be the best tool the ‘people’ have to counteract the powers of the elite and nation state. It’s the marriage of Snowcrash and The Sovereign Individual - the perfect union.
Yes, good points. If we can have a decentralized internet that would be amazing. I’ll look into Helium. As for decentralized AI agents, yes, possible. But they will potentially be in conflict with other AI (I.e. under attack, at war, fighting for their lives). They may or may not even survive unless they are very powerful. But again, we don’t know what these AIs will value. We shall see I guess.
Thank you for another insightful essay. Your past musings quickly turned into reality, and they are now being brought to life at Ethena Labs. I am confident that your new ideas will also become reality.
For those who speak Russian or are from the Commonwealth of Independent States (CIS), feel free to read the Russian translation of this article here: https://teletype.in/@rekted_capital/moai-esse-artura-heisa-RUS
so, ethereum is metamask and infura only? omg protos my ass
Thank you, Arthur. Insightful and enjoyable, as always. My thoughts:
AI is still an illusion and will continue to be so until that knowledge is lost and 'an AI' effectively becomes a god. Even if alternatively a digitised human implementation were developed, or more sinisterly a computer program(mer's) interpretation of a person's being. As per how people can believe a chatbot is a real person. Still, we must please, must be angry, please make this happen, etc.. Initially with an organisation claiming it's not us, the AI did it! This is not implausible. States would be forced to prove so in a case of law. A costly process but more likely one where states would need to prove that actions were not due to an AI but humans. Before states are able to do so they would need a legal definition of what constitutes 'an AI', where and how to attribute responsibility and accountability. How to ethically deal with the issue, etc., etc.. Sounds like a lot of work. Fake it 'till you lobby it into being by your definition seems likely.
AIs could be/become the LLCs of today, automated. AI code incorporates corporate codes, goals, values, policies, where and how any compassion is programmed in, or not. Treated as a sentient beings above the law. Attack is only via digital vectors by those able to do so.
Would expect PoetAI to steal all required works because there is no real effort for doing so. If such works are there for the taking then a transaction may even represent additional overhead. Financial efficiencies are to be considered and are coded-in. Compute cost vs price would determine and execute. Being more efficient than the competition is everything. This all follows the why would AI conform to human/state law theory.
AIs would pump markets and price out humans. AIs would pump those gas prices tf and beyond! Also not collusion as such, but 'agreement' and 'pile-ons' are to be expected. Any compassionate elements would naturally become deprecated as they would lose efficiency in hyper-markets and that would mean death. Natural selection will be played out. Weaknesses and vulnerabilities will be exploited. Rules will always be interpreted differently whether that be innocuously or sought by the algorithms made live.
This is a great piece of analysis Arthur, however this deserves some objective criticism as every other great piece out there!
You tend to mention that AI entities will generate economic activity, and they can ideally exist in DAO worlds and interact with crypto tokens. Who will consume their output? Who will buy PoetAI newly generated poems? This piece is missing the crucial concept that economic activity is generated when some kind of exchange happens between _humans_.
It is true that DAOs and public blockchains solve for a lot of LLC related issues today, but what's the point of everything if there are no humans in the loop?
This piece does not answer some basic questions like: Who builds this AI that then autonomously raises money for something? Why would an investor, be it an AI or a human, invest into something like this?
Only a few smart people nowdays know for a fact that AI will improve our life 1000x, but not AI on its own, instead it would be an AI that is built by humans and assists humans into boosting output per capita, where humans are in the center of the story. If we start considering sentient AIs and a Terminator-like future, then it makes no sense to discuss crypto, profits and other bullshit because it will be a fight for survival and nothing else will matter.
How would a DAO like this function with humans and AI collaborating together? Can you always trust math in that scenario, and not humans? How does your model of predicting ETH price work when humans are added in the loop?
Another masterpiece.....WOW. !! thank you
Thank you for this great essay