The Case for Spacemesh
This exchange on X caught our attention, because it touches on a topic that is of great personal interest to us. And in this blog post, I hope to convince you that it’s of great communal interest to the human race.
The trajectory of our age is becoming increasingly obvious: rapid and ever-accelerating change, forcing us to reckon with the extent to which everything is complex and entangled. We simply must wake up and smell the adaptive complex systems!
We can’t afford to not get busy improving our capacity to cope with faster-occurring novelty and growing uncertainty, both from Mother Nature as well as from destabilizing human-designed systems of governance and control.
Many smart people are already aware that AI will have a part to play in this, while dismissing crypto as lacking a usecase. But I hope to convince you that these two technologies are intertwined. That, indeed, crypto is necessary to make AI work for us instead of simply adding more noise to drown out the signal.
And this potential lies in some unique features of cryptocurrencies — at least when they reach their full potential.
The Gemini Dilemma
When Google released their latest AI text-to-image model, the tech world was all aflutter. But that excitement turned to ashes in our mouths when this much-anticipated AI tool turned out to be programmed by its human creators to lie to us about history — part of a broader campaign by Google to manipulate information for the sake of a “progressive” agenda.
But there’s nothing progressive about willful deception. In fact, in a world that’s already inundated with bad information and lies, such deceit is downright reactionary — accelerating the breakdown of coherent communication that has been the hallmark of the fiat, and especially the post-Internet fiat age.
What we need in these chaotic times is not technology that can be used to spread yet more untruths, but technology that, by its very nature, cannot be used in the service of untruths.
Finding the Signal
With so many players trying their best to muddy the waters even more, how do we keep our heads above the muck? How do we create a reliable means of sensing and adjusting to rapid and accelerating change? How do we find the signal in our ever more noisy and chaotic world?
We need more quality, holistic information. And for that, we need “sensors” in more places. We also need the right kind of sensors. Ideally, supercomputers trained on matters of utmost concern to human beings. Millions of them. Perhaps even billions of them.
The best sensors of all for the purposes of human flourishing are the human beings themselves!
It’s sine qua non upon us to create ‘real-time human sensor networks’ which can examine novelty from a multi-paradigm and multi-cognitive perspective, seeking both patterns and outliers. Along with, for fidelity’s sake, an incentives model that rewards being the most forthright and sincere as a sensor (e.g. ‘I’m pretty sure the king has no clothes’ gets the big bucks if it’s correct), thus making it the most rational behavior.
And for such a network to be built, it requires the right infrastructure. Decentralization is a prerequisite for humans to grapple with the coming age of perpetual, rapid, and chaotic change. Should we be faced with no other choice but to collaborate, how can we protect and upregulate trustful collaboration?
Team Spacemesh has been thinking about this topic for about a decade now, and we believe we’ve found some compelling answers. Let me try to summarize our key findings:
What sort of qualitative guarantees would we want from a decentralized system? Or, even more ambitiously: How would one take a shot at creating the most decentralized digital crypto asset ever?
What guarantees would suffice to reassure us that it would, in fact, conserve decentralization — despite whatever usage by super smart AI (in addition to cute regular people, of course)?
The way to conceptualize this problem is that any topology is like a blanket with which you’d aim to cover as much territory as possible, with as few tradeoffs as possible, in the realms of security, decentralization, scalability and environmental footprint.
But tradeoffs there must be, and once such tradeoffs are established, a topology has been formed. A superior solution would therefore need to demonstrate that the tradeoffs that come with decentralization — namely, in transaction speed and efficient governance — result in the maximization of decentralization. And moveover, that this decentralization provides benefits in excess of the tradeoffs.
We believe that Spacemesh offers exactly that.
—
In this write-up, I aim to follow in the footsteps of Balaji S. Srinivasan , and offer both a fairly fast overview of the case in its entirety, along with the option to dig deeper into any of the hypotheses that one may find more challenging to accept on introductory ‘face value’.
And herein is the case for Spacemesh coin, the ‘big blanket topology’ (or as we like to call it, the People’s Coin) based on the following three hypotheses:
- AIs will become the largest users of crypto networks and digital assets. Their use will drive adoption and dominate demand.
- AIs are likely to trend towards the ‘highest quality of decentralization’ assets, due to decentralization’s tangible evolutionary fitness enhancing benefits.
- Spacemesh would inevitably, due strictly to its technological topology, become further decentralized over time, positioned to become the most decentralized topology yet.
The Machines should use crypto-money as their lingua franca
There is a great deal of understandable controversy around AI technologies, as they become more sophisticated and independent. Thus, one of the biggest questions is how we can make them more ethical. This is one area where cryptocurrencies will come into play.
As many of us have gathered by now, money is a language. However, in being a language, money can also be used to “lie”. And AIs, when put into competition against one-another, will have the same incentives to lie that humans do.
If there’s one thing the latest debacle with Gemini has made abundantly clear, it’s AIs are nothing but a mirror to humankind. They are not their own identities or persons: they’re us. They can be manipulated when fed with the wrong information, spreading propaganda. They can be made to lie for the sake of human agendas. But, to make AIs a technology for good and not for evil, they must be honest. This is a choice that we must make for our own sake.
But that raises the question: how do we keep AIs honest? We propose that the answer is to make the cost of dishonesty more “expensive” than being honest.
Enter cryptocurrencies. It solves the “false signals” problem by making arbitrary money printing impossible, and it solves the lying problem by using a verifiable transaction ledger. This makes it superior English or bits, because of the inability to change “the story” — that is, the ledger — retroactively, since it introduces an unavoidable ever growing cost associated with changing shared coordinated reality (our ledger) after the fact.
If the communications of AIs and humans were recorded and put in an unbreakable vault, there would be no more convenient just-so stories about this past. There would be no more “memory-holing” of inconvenient or uncomfortable truths. Even if our brains lie to us, the cold hard material reality would not. In the realm of money, it’s the difference between an honest and dependable system and the one we’re stuck with today.
But that’s just one dimension of what makes cryptocurrencies such a powerful form of communication. As discussed above, the even more desirable attribute of cryptocurrencies is that, when they have the right infrastructure, they are far better at parsing the signal from the noise.
But what is the “right infrastructure” to optimally achieve this result?
If we decided to make crypto the language of the Machines, which kind of crypto assets would they end up using? What intrinsic properties for the asset would eventually be selected for?
The more decentralized the cryptocurrency, the better for use by AI
Our present selection process for cryptocurrencies is driven by irrational humans, who have social concerns — ”keeping up with the Joneses” and other forms of FOMO — and a reward-seeking drive that has a tendency to favor get-rich-quick schemes of dubious morality and quality.
AI presents a game changer, however. Once this technology has fully matured, we will in effect be adding a new kind of digital pseudo-organism to our human universe which, most certainly, has its own things it must do to survive. It’s going to coexist with us economically, and perhaps even politically. This is likely going to completely change the ways in which governments and people interact, because of much cheaper and faster AI-enabled services.
In the 20th century, the nation state was ascendent. But in the 21st century, it’s becoming increasingly apparent that the digital world is ascendant. AIs cannot be controlled by any one government. They’re going to take resources and knowledge anywhere they can, to grow and fulfill whatever mission they were initialized to do. The algorithmic learning of AIs, while completely changing the shape of our economy, will also change the way we interact with our material reality.
Herein lies our rationale for why AI would be optimized by using the most decentralized cryptocurrency.
For all the potential of cryptocurrencies and the crypto space, all major recent collapses involved centralized projects (FTX,Terra Luna, …), too much leverage, and rising interest rates. A real-life demonstration of the superior robustness of decentralized systems. One that not only will not appeal to the political class to “bail them out”, but cannot by its very nature.
This is all in addition to the commonly-understood advantages of decentralization, such us:
Maximum robustness and security: the most decentralized assets = the most expensive and most difficult assets for a single part or interest group to capture above 51%.
Better Governance: A ‘Money System’ is always threatened by an oligarchy of holders which fails to see perspectives other than their own.
For optimal governance in cryptocurrencies, economies of scale CANNOT be allowed to price out “the little guy”, which is what ultimately leads to the development of plutocracy. If you protect the little guy, you uphold the original promise of ‘magic internet money everyone can join via mining from home’.
Fastest integration of information: Perhaps the most important fitness advantage of decentralization has got to do with information asymmetries, and the speed in which those asymmetries are discovered and integrated. More sensors in more places, via many “little guys” creating an emergent transaction process, will mean we discover the emergent asymmetries more quickly. Why would any AI choose slower integrating assets when faster integration exists and the value of FASTER rigorous synthesis and integration is so profound?
Energy efficiency: Much like in biological systems, technology evolves to maximize energy efficiency via reuse. In our case, it would essentially mean something like AI_bob saying to AI_Alice: “Hello AI_Alice, since you already compute this all the time, I’m going to chip in on your cost, and use your results, rather than redoing them all on my own”.
So decentralization also helps in the realm of communication, which means the AI machines have two more urgent needs: the most decentralized compute-power and storage, and the ability to share the cost/benefits of certain ongoing computation and storage with other AIs.
Let’s assume for rigor’s sake, we are wrong, and for example ‘Speed and Response time’ would out-perform ‘Decentralization quality’, thus selected for by the AIs. That must mean the challenges that AI would compete and compute over are not time and not rule variance, like chess or go, which we know is already wrong, therefore learning quality over speed again seems more plausible.
While evolution is impossible to predict and the evolution of our digital gem stones is yet so early in its evolution process — it is however highly plausible upholding permissionless low barrier of entry, leading to highest node count distributed geographically. is a good proxy for quality of signal (accuracy and fidelity) And stands good chances to be selected for mission critical tasks.
To summarize, during the human selection process, ‘all the debt buried people, where do they all belong?’ trends towards risk reduction via ‘following the leader’, or ‘trust the herd’ etc. AI would be able to make more narrow and accurate discernments, which in the qualitative advantage takes priority over quantity. For example it seems plausible that developers’ adoption would follow the AIs and not the other way around.
Spacemesh’s tradeoffs topology promotes further decentralization over time, making it optimal for use by AI that require a robust, resilient crypto “language”
The rationale behind such a bold future prediction, for Spacemesh becoming a source of everlasting creativity, and the generator of stable cryptographic primitives, is as follows:
1. Binding
Such “coherence over time”-type signaling systems, roughly speaking, can only be achieved in two distinct ways: as a circular, self-referential loop or binding to extrinsic commodity. However the illusive point to hone on is that these are NOT interchangeable. Binding to an extrinsic commodity comes with distinct advantages:
Shock Resistance — let’s examine destabilizing patterns as means of assessing shock resistance, the price of energy is the binding of BTC, Storage over time of SMH, now contrast this against stake as security. Only in the latter, once a coin is under attack, and its price is going down as a result, your security guarantees are shrinking as first-order effect and attack becomes instantly cheaper to maintain.
Natural Bottom up Feeding. You learn from the environment within which you are embedded. The binding — the number we agree upon as external to the system — allows us to play the superior game of ‘if you know better prove it, and we add it to our shared baseline’, and not the inferior game of ‘he said she said’. The latter inevitably makes the rich more correct, as they have more “stake” with which to manipulate the self-referential system. And now feel free to extrapolate: currencies nested to the same extrinsic substrate vs. no sound way to value exchange across economies.
In Spacemesh, for example, we use time as our extrinsic resource. Such first-principles binding to time is achieved because, once a smesher’s storage is fully initialized, the amount of $SMH payoffs over a given time unit is fairly predictable. This means that if a t-shirt costs, say, 3 SMH, and you make 1 SMH per day, then this t-shirt costs you 3 days of mining on ‘first principles’.
The correlation between time and value, therefore, is self-evident.
Extrinsic binding is critical, because without this friction between the territory called reality and our imperfectly constructed map called money, our world descends into the chaos, confusion, and mistrust that is so evident around us in the world today.
2. Permissionless
Permissionless-ness can be a double-edged sword when it comes to decentralization. On the one hand, it means anyone, anywhere can join — which means a bigger pool of potential “sensors” to add to our network. On the other hand, it means that there are no limits to the amount of resources a “sensor” can pour into the network. So the real challenge is finding a way to maintain decentralization in the face of economies of scale and human ingenuity.
The only sound way to immunize ourselves against the capture game is to remain an open system, and that means to reduce the extent to which the system is predictable — protecting the element of surprise, such that one can neither truly predict nor control the results. Fair permissionless mining, the open system element, is how we preserve the element of surprise, thereby benefiting strategies which leave the future open to surprises.
In Spacemesh, the above is achieved via low marginal costs for home users when compared to “industrial miners”, combined with the race-freeness property that ensures rewards are proportional to the contributed spacetime resources, and moreover are distributed at a predictable rate. This may seem contradictory to the idea of an open system full of surprises at first glance, but it’s actually a key part of the incentive structure which keeps a steady stream of new “sensors” — that is, home miners — entering the system. This leads to a “long tail” of small miners that cannot be priced out, entering and exiting the system in a manner that preserves “surprise”.
3. Leaderless
We mentioned above how Spacemesh distributes rewards on a predictable basis, as opposed to in a “race” or “lottery”-based competition (e.g. BTC longest chain). These “leader-based” systems fail to deliver on decentralization, as they provide those with significant capital to invest (e.g. custom ASICs, or more “stake”) an unfair advantage.
The solution for this problem lies in the linearity of rewards, where a powerful (perhaps adversarial) miner cannot receive a disproportionately high portion of the rewards due to superior resources or superior luck. This is the only practical way to protect the little guy.
In Spacemesh, rather than one winner getting all the rewards in big, unpredictable jackpots, every miner gets their proportion of the rewards in smaller, frequent, and predictable payouts. Every miner that follows the rules gets something for their efforts. Obviously, the bigger miners still get more, but they do not get all.
To summarize, leaderless protocols such as Spacemesh creates the topology that promotes further decentralization over time, and combat the plutocracy that’s so cancerous to decentralization.
Sorting Through the Noise
Rapid and ever-accelerating change is the trajectory of our age, forcing us to reckon with the extent to which everything is complex and entangled. This in turn will force us to improve our capacity to cope with faster-occurring novelty and growing uncertainty.
And now AIs are emerging into this world — complex algorithmic learning systems that stand to one day, perhaps sooner than anyone is expecting, evolve into pseudo-organisms with a kind of agency of their own.
AI and cryptocurrency are in a yin and yang dance. Cryptocurrencies provide human sensor networks that can allow for high-fidelity, trustworthy ways for AI to compute and communicate. This is because cryptocurrencies provide incentive models that are effective in making sincere truthfulness the rational choice. It’s in our interest to make cryptocurrencies the language of AIs, to “keep them honest”.
Spacemesh, which promises to become the most decentralized and un-capturable cryptocurrency available by remaining permissionless in practice through the lowest possible barriers to entry, will provide the highest “quality of signal”.- making it the natural yang to AI’s yin.