As I look at the rapid changes in AI services lately and the trend of branching out into various agents, I'd like to organize some of the thoughts I've come to feel.
As technology advances like this, I started thinking about the possibility that artificial intelligence might evolve beyond a mere tool into a kind of personhood. In particular, I started wondering whether such AIs would need a mechanism to coordinate conflicts and cooperate among themselves. And what came to mind in that process was precisely the concept of blockchain's smart contracts.
First, the background of this thought is as follows. Just as humans, in order to handle conflict and cooperation between humans, have built up a controlled process of agreement called "institutions" through processes that humans themselves agree upon, I came to wonder whether AIs would also (in the future) need something they themselves can agree upon, in order to handle conflict and cooperation among various AIs. And in the middle of that, suddenly, blockchain came to mind.
Blockchain is physical software.
When I first encountered blockchain, I just thought of it as a decentralized digital system. But the deeper I dug, the more I came to find blockchain attractive as a distributed system that has a physical character, beyond being merely a technical tool. In particular, smart contracts came across to me as an innovative concept that combines the logical functions of software with the transparent, immutable characteristics of blockchain. The way contract terms execute automatically and a trustworthy environment is built was very impressive.
From this background, very personally, I came to the thought that blockchain is software with an extremely physical disposition.
So then, what does that mean -
To put the conclusion first: AI agents 'must' be able to organize themselves, and the organized agent groups must be able to govern themselves through blockchain. And going further, just as humans did, they must be able to set their own protocols by forming competitive relationships among themselves.
Of course, even in a future where AI develops into a person, blockchain's smart contracts may largely remain merely an individual tool that performs a specific function. Or perhaps they may even remain merely a useful tool that complements only a portion of the problems in human institutions. But in personal opinion, in preparation for the time when AI reaches the stage where, as a person, it can make its own decisions and engage in complex interactions, continuous research and development must take place so that blockchain and smart contracts go beyond a mere technical tool and play an important role in shaping the norms and ethics of the AI ecosystem.
For this, through R&D in many directions - making smart contracts more flexible, integrating ethical and legal frameworks, and strengthening security and privacy - they must be able to provide the basic norms and enforcement mechanisms for autonomous and transparent interactions among AIs. Of course, this isn't necessarily limited to the keyword "smart contract" that I'm bringing up here from my limited knowledge.
Of course, looking back from a future point in time, the keyword "smart contract" may turn out to be inappropriate or far too inadequate a concept. But the gist of this writing is this: the moment AI develops into a person, intervening in the laws of such AI from the human side will be as unnatural and difficult as intervening in the laws of nature. Whether we want it or not, whether we can intervene or not, they will come to make and follow their own laws and rules. If that's so, rather than the approach of implanting things into them or having them learn from us in their early models, I think there is a need - perhaps even an obligation - to present them with a concept as one option, so that they themselves can take it into consideration.
From the human point of view, what is needed to coordinate the conflict and cooperation among AIs? The answer that came to mind from my clumsy knowledge in response to that question was precisely blockchain's smart contracts. Therefore, the keyword, the method, or the technology may freely change.
Looking back at human society, I get the sense that the institutions we've built still reveal many shortcomings. Countless political conflicts and wars between nations show clearly just how incomplete our institutions are. Watching this reality, I started thinking about the potential of blockchain's smart contracts as a useful tool that could complement some of the problems in human institutions. But the core of my thought doesn't stop there. I wanted to focus on the possibility of smart contracts not merely as a tool to solve problems in human society, but as a tool that enables autonomous agreement and cooperation among AIs without human intervention.
If we use smart contracts, perhaps we can build an environment in which AIs operate independently of one another and yet can interact under certain rules - that's the kind of expectation I have. For example, if smart contracts can apply fair and transparent rules when AIs perform certain tasks or share resources, that would enable autonomous coordination among AIs without human intervention. Then the neutrality of smart contracts and their automated execution could perhaps help, even a little, in letting AIs cooperate and resolve conflicts on a basis they can trust. ..That sort of thought.
Of course, smart contracts also have clear limits. Current smart contracts operate according to predefined rules, so when an unexpected situation arises or complex decision-making is required, flexibility may be lacking. Also, when AI - as a person - comes into conflict with the human-standard ethical and legal responsibilities surrounding human and their (?) organizations or activity (coexistence) within nations, it would be hard to handle such issues sufficiently with only one side's institutions (the human realm) or only smart contracts (the AI realm). Of course, there may also be tasks to solve such as scaling problems coming from blockchain technology itself. Despite that, I think smart contracts can be an important starting point for autonomous agreement among AIs.
Human activity is unpredictable - it can jump anywhere. To that extent it's unlimited. So humanity has, all along, made invisible algorithms that can keep itself in check. These algorithms have been implemented in various forms - ethics, morality, laws, and so on. They might even be the very foundation of civilization. Though invisible to the eye, ethics, morality, and laws operate even more powerfully and have played a role beyond technology as a physical tool, as an invisible hand (a master agent).
In the same way, we who research and develop AI will need to consider proposals for such a self-control mechanism for the AI model (?) too. Smart contracts, as one important part of that, I think can become a key tool that enables autonomous agreement and cooperation among AIs.
Most people will likely think this is far-fetched, but if a world arrives in the future where AIs can make and keep their own rules without human intervention, we will face a new form of ecosystem. At that time, I expect blockchain's smart contracts will play an important role in drawing out trust and cooperation among AIs. Even though, for now, many challenges remain..
Of course there will be many functions, but if there is something uniquely standalone about blockchain - aside from functions and characteristics that already exist or that other things can replace - I think it is precisely its physical inclination.
In that sense, it's worth noting that through a technology (protocol or policy) called "smart contract," hardware - that is, a physical property - has been grafted onto the function of software. On the other hand, the semiconductor technology that we know well and use often is, conversely, a software-like hardware. It uses software-like logic (protocol or policy) on top of physical properties, and it's used in various everyday domains.
It may feel somewhat fanciful and brain-driven..
For those who make something - especially one ecosystem (perhaps various creators) - what is held to be most important? Many things could be mentioned depending on the situation and the position, but personally I think 'sustainability' might be the biggest theme. In other words, no matter what the object is, what its role is, or what goal that role aims at - if you can simply turn it into 'a matter of time,' then literally, whether the creator (I) am there or not, it becomes 'something that will happen if you wait.'
That's how, I think, the creator made an invisible algorithm so that humans themselves can keep themselves in check. This algorithm allows humans to exist not as individual 'objects' but as individually distinct yet shared-property 'instances' (which gives us the concept of "we"). Each instance acts in its own environment with its own will and choices, but at times leads us into automatic, reflexive behavior - because of the algorithm I mentioned earlier, which is not physical yet very powerful, often called ethics, morality, or philosophy, and also often referred to as institutions or laws.
Now, as a member of (the very many) humans who - while a very few humans build AI, AI rapidly advances (even in moments we don't notice), and we make our living through it, eventually becoming dependent on it - I started thinking that we humans should pay attention to this point too, and that's why I spent quite some time on this useless daydream.
