太陽能神諭隨身聽 Solar Oracle Walkman
太陽能神諭隨身聽（Solar Oracle Walkman）是一件受能源交易機制啟發的聲音藝術裝置，也是我當前的 startup project。它透過讀取手工製作的染料敏化太陽能電池（mini DSSC disc）的電流 — 電壓特徵，將這些量測陽光的結果轉換為可重構的生成音樂。每一片圖案化 DSSC 的 TiO₂ 半導體層，都透過氰版顯影或網版印刷引入刻意設計的微觀缺陷，使其在光照下呈現具有差異性且難以完全複製的 I–V 曲線特徵。這些特徵可以被理解為一種具物理唯一性的光伏簽名，也可以進一步作為 physical unclonable function（PUF）的實驗性基礎，讓每一次具體的光照事件，不只是能量輸入，也成為可被辨識與驗證的物質媒介。
但透過設計這個原型，我慢慢意識到，我真正感興趣的其實不只是材料辨識或能源驗證，而是更底層的問題：心智到底是什麼？我們為什麼需要不斷測量世界？無論是人透過感官理解世界，還是機器透過感測器讀取環境，中間都會經過取樣、比對、誤差、修正與校準。也就是說，我們所接觸到的現實，往往不是一個一次到位、永遠固定的東西，而比較像是在不斷測量與反覆校準之中，暫時穩定下來的一個世界模型。這像是 Andy Clark 提出的predictive brain，它不是單純在說大腦會預測，而是在提醒我們，所謂心智，本來就是一個持續更新模型的過程。像我們這種完全物質性的存在，怎麼會有知覺、理解、想像、思考，甚至意識這些看起來不像「純物質」的能力？心智其實是一個會主動預測、主動行動、主動改造環境、並且被身體與外部世界共同塑造的系統。
一旦從這個角度來看，「生成」是一種更普遍的結構，包括心智和LLM。我們現在看到的很多系統，不再是單純輸出一個固定結果，而是透過規則、資料、回饋、機率、感測、遞迴調整，不斷產生新的狀態。這件事在生成藝術裡很早就已經被拿來當作創作方法：作品不只是成品，而是系統如何運作、如何變化、如何在秩序與不確定之間形成形式。我想到 Galanter 對 generative art 的定義，就是藝術家把部分控制權交給一個具有相對自主性的系統，讓它參與或完成作品。
這裡最重要的不是藝術風格，而是：主體不再完全位於作者身上，形式也不再完全由作者直接決定，而是由系統與條件共同生成。這就是生成藝術真正重要的地方。它的重要性不只是因為它產出某種視覺效果，而是因為它比很多其他領域更早面對了一個今天變得非常普遍的問題：當結果不是直接被做出來，而是被生成出來時，作者、系統、材料、環境和接收者之間的關係要怎麼重新理解？
這也是它和生成式 AI 真正連起來的地方。它們共享的不是名稱，而是結構：輸出不是預先寫死的，而是從模型、資料、提示、機率分布、運算資源與互動過程中被生成出來的。也就是說，生成不只是美學方法，它已經變成一種技術架構，甚至變成一種理解世界的方式。
而只要一個系統是以生成為核心，它就會直接碰到身份問題。因為身份本來就是一種「在變動中仍被認作同一個東西」的條件。對固定物件來說，身份常常靠外形、名稱、序號、材料組成來判定；但對生成系統來說，情況不一樣。生成系統的每次輸出都可能不同，它的表現會隨上下文改變，甚至連邊界都不一定清楚。這時候，身份就不能只理解成一個靜態標籤，而必須理解成某種跨時間維持的生成一致性。
所以談生成和談身份，關係其實很深。因為一個生成系統最根本的問題之一就是：它在不斷變化、不斷輸出新狀態的情況下，憑什麼還被看成是同一個系統？這個問題在生成藝術裡其實已經出現過一次。當藝術家把控制權部分交給規則、機器、演算法、材料或自組織過程時，作品的身份就不再只等於藝術家的意圖，也不再只等於最後那個物件，而是和整個生成過程有關。
也正是在這個背景下，complexism 才開始變得有用。但在我這裡，它不是起點，也不是最後要證明的理論本體。它比較像一個範例：示範當原本的分類已經不夠用時，理論如何提供新的理解方式。Galanter 提出 complexism，不只是為了替生成藝術找到一個位置，而是為了描述那些既不是完全固定、也不是完全任意的複雜生成狀態。他把它放在現代主義與後現代主義之間，作為一種關於分散、回饋、湧現與不完整認識的描述框架。
如果把這個思路再往今天推進一步，問題就不再只是我們要怎麼理解真理、作者或文本，而是我們要怎麼理解「我」。過去談自我，很多時候還是預設它是一個相對穩定的中心。即使後來有人質疑這個中心，討論也多半還是在人類主體的範圍內進行。但 AI agent 出現之後，情況開始不一樣了。現在我們面對的，是一些並不是人、但又確實可以記憶、回應、規劃、執行、交易，甚至長期代表某個目標行動的系統。這時候，自我就不能再只被理解成心理學上的內在感受，或哲學上的抽象主體，而必須同時被理解成一種可以在世界中持續運作、持續被辨認、也持續承擔後果的結構。
把身份層和能源層放在一起看，一開始對我來說確實只是個很直接的技術需求：我想替 DSSC 建立身份，讓生成音樂具有可重構性。後來我才慢慢意識到，這件事不只跟作品有關。任何能夠長期存在並持續行動的系統，都不可能只靠一個名字、一組帳號、一把金鑰就成立。它還必須有物理上的條件：它怎麼獲得能量，怎麼維持運作，怎麼感測環境，怎麼留下連續的行為痕跡。沒有這些條件，身份就很容易只剩下一層符號外殼。它可以被複製、租借、轉移、模擬，但很難說那真的是一個有自身連續性的存在。
我把身份層和能源層綁在一起，最初有一部分也來自一種對未來基礎設施的直覺：如果之後真的出現更明顯的能源經濟，身份可能會像早期網際網路上的 email address 一樣，變成一種先被建起來、之後才慢慢滲透到各種行為裡的基礎設施。而對 AI agent 來說，這件事尤其合理。它們需要能源才能行動，也需要感測與硬體條件才能維持連續性。從這個角度看，把身份來源和能源來源綁在一起，不只是驗證手段，也是在替一種新的存在形式建立更可信的邊界。
此外，一個系統之所以像是一個自我，不只是因為它能說「我是誰」，而是因為它能在噪聲、變動和限制之中，持續維持某種可辨識性。它有自己的輸入輸出關係，有自己的資源邊界，也有自己的代價。這樣看來，未來真正重要的問題可能不是 AI 會不會比人更聰明，而是我們要用什麼標準去判斷，一個系統到底是不是一個自己。這個問題會出現在 AI agent 的設計裡。因為當 agent 可以長期執行任務、調用工具、管理資產、代表使用者做決策時，我們就必須問：它的連續性來自哪裡？來自模型本身？來自記憶？來自帳號？還是來自它和某個硬體、某種感測條件、某個能量來源之間穩定的耦合？
我認為這些問題也會出現在腦機介面或神經植入的設計裡。因為一旦記憶、注意力、感知或決策可以被外接系統參與，自我的邊界就不再那麼清楚。到那個時候，「我」可能不再只是自然形成的心理事實，而會變成一個被技術條件重新安排的結構。那麼自我是延伸了，還是被切分了，還是被平台化了，這些都會變成很實際的問題。
但透過設計這個原型，我慢慢意識到，我真正感興趣的其實不只是材料辨識或能源驗證，而是更底層的問題：心智到底是什麼？我們為什麼需要不斷測量世界？無論是人透過感官理解世界，還是機器透過感測器讀取環境，中間都會經過取樣、比對、誤差、修正與校準。也就是說，我們所接觸到的現實，往往不是一個一次到位、永遠固定的東西，而比較像是在不斷測量與反覆校準之中，暫時穩定下來的一個世界模型。這像是 Andy Clark 提出的predictive brain，它不是單純在說大腦會預測，而是在提醒我們，所謂心智，本來就是一個持續更新模型的過程。像我們這種完全物質性的存在，怎麼會有知覺、理解、想像、思考，甚至意識這些看起來不像「純物質」的能力？心智其實是一個會主動預測、主動行動、主動改造環境、並且被身體與外部世界共同塑造的系統。
一旦從這個角度來看，「生成」是一種更普遍的結構，包括心智和LLM。我們現在看到的很多系統，不再是單純輸出一個固定結果，而是透過規則、資料、回饋、機率、感測、遞迴調整，不斷產生新的狀態。這件事在生成藝術裡很早就已經被拿來當作創作方法：作品不只是成品，而是系統如何運作、如何變化、如何在秩序與不確定之間形成形式。我想到 Galanter 對 generative art 的定義，就是藝術家把部分控制權交給一個具有相對自主性的系統，讓它參與或完成作品。
這裡最重要的不是藝術風格，而是：主體不再完全位於作者身上，形式也不再完全由作者直接決定，而是由系統與條件共同生成。這就是生成藝術真正重要的地方。它的重要性不只是因為它產出某種視覺效果，而是因為它比很多其他領域更早面對了一個今天變得非常普遍的問題：當結果不是直接被做出來，而是被生成出來時，作者、系統、材料、環境和接收者之間的關係要怎麼重新理解？
這也是它和生成式 AI 真正連起來的地方。它們共享的不是名稱，而是結構：輸出不是預先寫死的，而是從模型、資料、提示、機率分布、運算資源與互動過程中被生成出來的。也就是說，生成不只是美學方法，它已經變成一種技術架構，甚至變成一種理解世界的方式。
而只要一個系統是以生成為核心，它就會直接碰到身份問題。因為身份本來就是一種「在變動中仍被認作同一個東西」的條件。對固定物件來說，身份常常靠外形、名稱、序號、材料組成來判定；但對生成系統來說，情況不一樣。生成系統的每次輸出都可能不同，它的表現會隨上下文改變，甚至連邊界都不一定清楚。這時候，身份就不能只理解成一個靜態標籤，而必須理解成某種跨時間維持的生成一致性。
所以談生成和談身份，關係其實很深。因為一個生成系統最根本的問題之一就是：它在不斷變化、不斷輸出新狀態的情況下，憑什麼還被看成是同一個系統？這個問題在生成藝術裡其實已經出現過一次。當藝術家把控制權部分交給規則、機器、演算法、材料或自組織過程時，作品的身份就不再只等於藝術家的意圖，也不再只等於最後那個物件，而是和整個生成過程有關。
也正是在這個背景下，complexism 才開始變得有用。但在我這裡，它不是起點，也不是最後要證明的理論本體。它比較像一個範例：示範當原本的分類已經不夠用時，理論如何提供新的理解方式。Galanter 提出 complexism，不只是為了替生成藝術找到一個位置，而是為了描述那些既不是完全固定、也不是完全任意的複雜生成狀態。他把它放在現代主義與後現代主義之間，作為一種關於分散、回饋、湧現與不完整認識的描述框架。
如果把這個思路再往今天推進一步，問題就不再只是我們要怎麼理解真理、作者或文本，而是我們要怎麼理解「我」。過去談自我，很多時候還是預設它是一個相對穩定的中心。即使後來有人質疑這個中心，討論也多半還是在人類主體的範圍內進行。但 AI agent 出現之後，情況開始不一樣了。現在我們面對的，是一些並不是人、但又確實可以記憶、回應、規劃、執行、交易，甚至長期代表某個目標行動的系統。這時候，自我就不能再只被理解成心理學上的內在感受，或哲學上的抽象主體，而必須同時被理解成一種可以在世界中持續運作、持續被辨認、也持續承擔後果的結構。
把身份層和能源層放在一起看，一開始對我來說確實只是個很直接的技術需求：我想替 DSSC 建立身份，讓生成音樂具有可重構性。後來我才慢慢意識到，這件事不只跟作品有關。任何能夠長期存在並持續行動的系統，都不可能只靠一個名字、一組帳號、一把金鑰就成立。它還必須有物理上的條件：它怎麼獲得能量，怎麼維持運作，怎麼感測環境，怎麼留下連續的行為痕跡。沒有這些條件，身份就很容易只剩下一層符號外殼。它可以被複製、租借、轉移、模擬，但很難說那真的是一個有自身連續性的存在。
我把身份層和能源層綁在一起，最初有一部分也來自一種對未來基礎設施的直覺：如果之後真的出現更明顯的能源經濟，身份可能會像早期網際網路上的 email address 一樣，變成一種先被建起來、之後才慢慢滲透到各種行為裡的基礎設施。而對 AI agent 來說，這件事尤其合理。它們需要能源才能行動，也需要感測與硬體條件才能維持連續性。從這個角度看，把身份來源和能源來源綁在一起，不只是驗證手段，也是在替一種新的存在形式建立更可信的邊界。
此外，一個系統之所以像是一個自我，不只是因為它能說「我是誰」，而是因為它能在噪聲、變動和限制之中，持續維持某種可辨識性。它有自己的輸入輸出關係，有自己的資源邊界，也有自己的代價。這樣看來，未來真正重要的問題可能不是 AI 會不會比人更聰明，而是我們要用什麼標準去判斷，一個系統到底是不是一個自己。這個問題會出現在 AI agent 的設計裡。因為當 agent 可以長期執行任務、調用工具、管理資產、代表使用者做決策時，我們就必須問：它的連續性來自哪裡？來自模型本身？來自記憶？來自帳號？還是來自它和某個硬體、某種感測條件、某個能量來源之間穩定的耦合？
我認為這些問題也會出現在腦機介面或神經植入的設計裡。因為一旦記憶、注意力、感知或決策可以被外接系統參與，自我的邊界就不再那麼清楚。到那個時候，「我」可能不再只是自然形成的心理事實，而會變成一個被技術條件重新安排的結構。那麼自我是延伸了，還是被切分了，還是被平台化了，這些都會變成很實際的問題。
Solar Oracle Walkman is a sound art device inspired by the logic of energy trading, and it is also my current startup project. By reading the current-voltage characteristics of handmade dye-sensitized solar cells (mini DSSC discs), it transforms these measurements of sunlight into reconstructable generative music. In each patterned DSSC, the TiO₂ semiconductor layer is deliberately marked with microscopic defects through cyanotype exposure or screen printing, so that under illumination it produces I–V curve features that are differentiated and difficult to replicate exactly. These features can be understood as a physically unique photovoltaic signature, and can further serve as an experimental basis for a physical unclonable function (PUF), allowing each concrete event of illumination to become not only an input of energy but also a material medium that can be identified and verified.
But through designing this prototype, I gradually realized that what really interests me is not only material recognition or energy verification, but a more basic question: what is mind? Why do we need to keep measuring the world? Whether humans understand the world through the senses, or machines read the environment through sensors, the process always involves sampling, comparison, error, correction, and calibration. In other words, the reality we come into contact with is often not something given once and for all, fixed forever, but something more like a world model that becomes temporarily stabilized through ongoing measurement and repeated calibration. This resembles Andy Clark’s idea of the predictive brain. It is not simply saying that the brain makes predictions, but reminding us that what we call mind is already a process of continuously updating a model. How can fully material beings like ourselves possess perception, understanding, imagination, thought, and even consciousness—capacities that do not seem “purely material” at all? Mind is in fact a system that actively predicts, actively acts, actively reshapes its environment, and is itself shaped by the body and the external world.
Once seen from this angle, “generation” is no longer just a style within art, but a more general structure that includes both mind and LLMs. Many of the systems we see today no longer produce a single fixed output. Instead, they continuously generate new states through rules, data, feedback, probability, sensing, and recursive adjustment. This was already taken up early in generative art as a creative method: the work is not just an end product, but a way of showing how a system operates, how it changes, and how form emerges between order and uncertainty. This is what I think of when I recall Philip Galanter’s definition of generative art: the artist hands over part of the control to a system with a degree of relative autonomy, allowing it to participate in or complete the work.
What matters most here is not artistic style, but this: the subject is no longer located entirely in the author, and form is no longer determined directly by the author alone. Instead, both are generated jointly by systems and conditions. That is what is truly important about generative art. Its importance is not simply that it produces a certain visual effect, but that it confronted, earlier than many other fields, a problem that has now become very common: when a result is not directly made but generated, how do we rethink the relationship between the author, the system, the material, the environment, and the receiver?
This is also where it genuinely connects to generative AI. What they share is not a name, but a structure: the output is not prewritten in advance, but generated out of models, data, prompts, probability distributions, computational resources, and processes of interaction. In other words, generation is not only an aesthetic method. It has become a technical architecture, and even a way of understanding the world.
And as soon as a system is centered on generation, it immediately runs into the question of identity. Identity is, after all, a condition under which something is still recognized as the same thing even while it changes. For fixed objects, identity is often determined by appearance, name, serial number, or material composition. But for generative systems, the situation is different. Every output may differ. Their behavior changes with context. Even their boundaries may not be entirely clear. At that point, identity can no longer be understood as a static label. It has to be understood as some kind of generative consistency maintained across time.
So generation and identity are deeply related. One of the most basic questions for any generative system is this: if it is constantly changing and constantly producing new states, on what basis do we still treat it as the same system? This question already appeared once in generative art. When artists handed over part of their control to rules, machines, algorithms, materials, or self-organizing processes, the identity of the work no longer consisted only in the artist’s intention, nor only in the final object. It also came to depend on the entire generative process.
It is precisely in this context that complexism starts to become useful. But for me it is neither the point of departure nor the final theory I need to prove. It is more like an example: it shows how theory can provide a new way of understanding when the existing categories are no longer enough. Galanter’s notion of complexism was not only meant to give generative art a place, but to describe those complex generative states that are neither fully fixed nor fully arbitrary. He places it between modernism and postmodernism, as a descriptive framework for distribution, feedback, emergence, and incomplete knowledge.
If we extend this line of thought into the present, the question is no longer only how to understand truth, the author, or the text, but how to understand the “I.” In the past, discussions of the self often still assumed some relatively stable center. Even when that center was later challenged, the discussion usually remained within the range of the human subject. But with the emergence of AI agents, the situation has started to change. We are now dealing with systems that are not human, yet can clearly remember, respond, plan, execute, transact, and even represent a long-term goal over time. At that point, the self can no longer be understood only as an inner psychological feeling or an abstract philosophical subject. It also has to be understood as a structure that can continue to operate in the world, continue to be recognized, and continue to bear consequences.
At first, putting the identity layer and the energy layer together was simply a direct technical need for me: I wanted to give DSSCs an identity so that the generative music could be reconstructable. Later I gradually realized that this issue is not only about the work itself. Any system that can persist and continue to act over time cannot exist on the basis of a name, an account, or a cryptographic key alone. It also needs physical conditions: how it gets energy, how it sustains operation, how it senses its environment, and how it leaves continuous traces of behavior. Without those conditions, identity is too easily reduced to a symbolic shell. It can be copied, rented, transferred, or simulated, but it becomes difficult to say that it is truly a being with continuity of its own.
My decision to tie the identity layer to the energy layer also came, in part, from an intuition about future infrastructure. If a more explicit energy economy really does emerge, identity may come to function like the email address in the early internet: a piece of infrastructure that is built first and only later gradually permeates all kinds of behavior. And for AI agents, this seems especially reasonable. They need energy in order to act, and they also need sensing and hardware conditions in order to maintain continuity. From that perspective, tying the source of identity to the source of energy is not only a means of verification, but also a way of building a more credible boundary for a new form of existence.
A system resembles a self not only because it can say “this is who I am,” but because it can maintain a certain recognizability within noise, change, and constraint. It has its own input-output relations, its own resource boundaries, and its own costs. From this perspective, the important question in the future may not be whether AI becomes more intelligent than humans, but what standards we will use to judge whether a system is really a self. This question will appear in the design of AI agents. Once an agent can carry out long-term tasks, call tools, manage assets, and make decisions on behalf of a user, we have to ask: where does its continuity come from? From the model itself? From memory? From an account? Or from a stable coupling with some hardware, some sensing condition, or some energy source?
So far, mind philosophy is no longer pure philosophy, but consciousness science and an important role in modern science. I believe these questions will also appear in the design of brain-computer interfaces and neural implants. Once memory, attention, perception, or decision-making can be participated in by external systems, the boundary of the self will no longer be so clear. At that point, the “I” may no longer be simply a naturally formed psychological fact, but a structure rearranged by technical conditions. In that case, does the self become extended, divided, or platformized? These will become very practical questions.
But through designing this prototype, I gradually realized that what really interests me is not only material recognition or energy verification, but a more basic question: what is mind? Why do we need to keep measuring the world? Whether humans understand the world through the senses, or machines read the environment through sensors, the process always involves sampling, comparison, error, correction, and calibration. In other words, the reality we come into contact with is often not something given once and for all, fixed forever, but something more like a world model that becomes temporarily stabilized through ongoing measurement and repeated calibration. This resembles Andy Clark’s idea of the predictive brain. It is not simply saying that the brain makes predictions, but reminding us that what we call mind is already a process of continuously updating a model. How can fully material beings like ourselves possess perception, understanding, imagination, thought, and even consciousness—capacities that do not seem “purely material” at all? Mind is in fact a system that actively predicts, actively acts, actively reshapes its environment, and is itself shaped by the body and the external world.
Once seen from this angle, “generation” is no longer just a style within art, but a more general structure that includes both mind and LLMs. Many of the systems we see today no longer produce a single fixed output. Instead, they continuously generate new states through rules, data, feedback, probability, sensing, and recursive adjustment. This was already taken up early in generative art as a creative method: the work is not just an end product, but a way of showing how a system operates, how it changes, and how form emerges between order and uncertainty. This is what I think of when I recall Philip Galanter’s definition of generative art: the artist hands over part of the control to a system with a degree of relative autonomy, allowing it to participate in or complete the work.
What matters most here is not artistic style, but this: the subject is no longer located entirely in the author, and form is no longer determined directly by the author alone. Instead, both are generated jointly by systems and conditions. That is what is truly important about generative art. Its importance is not simply that it produces a certain visual effect, but that it confronted, earlier than many other fields, a problem that has now become very common: when a result is not directly made but generated, how do we rethink the relationship between the author, the system, the material, the environment, and the receiver?
This is also where it genuinely connects to generative AI. What they share is not a name, but a structure: the output is not prewritten in advance, but generated out of models, data, prompts, probability distributions, computational resources, and processes of interaction. In other words, generation is not only an aesthetic method. It has become a technical architecture, and even a way of understanding the world.
And as soon as a system is centered on generation, it immediately runs into the question of identity. Identity is, after all, a condition under which something is still recognized as the same thing even while it changes. For fixed objects, identity is often determined by appearance, name, serial number, or material composition. But for generative systems, the situation is different. Every output may differ. Their behavior changes with context. Even their boundaries may not be entirely clear. At that point, identity can no longer be understood as a static label. It has to be understood as some kind of generative consistency maintained across time.
So generation and identity are deeply related. One of the most basic questions for any generative system is this: if it is constantly changing and constantly producing new states, on what basis do we still treat it as the same system? This question already appeared once in generative art. When artists handed over part of their control to rules, machines, algorithms, materials, or self-organizing processes, the identity of the work no longer consisted only in the artist’s intention, nor only in the final object. It also came to depend on the entire generative process.
It is precisely in this context that complexism starts to become useful. But for me it is neither the point of departure nor the final theory I need to prove. It is more like an example: it shows how theory can provide a new way of understanding when the existing categories are no longer enough. Galanter’s notion of complexism was not only meant to give generative art a place, but to describe those complex generative states that are neither fully fixed nor fully arbitrary. He places it between modernism and postmodernism, as a descriptive framework for distribution, feedback, emergence, and incomplete knowledge.
If we extend this line of thought into the present, the question is no longer only how to understand truth, the author, or the text, but how to understand the “I.” In the past, discussions of the self often still assumed some relatively stable center. Even when that center was later challenged, the discussion usually remained within the range of the human subject. But with the emergence of AI agents, the situation has started to change. We are now dealing with systems that are not human, yet can clearly remember, respond, plan, execute, transact, and even represent a long-term goal over time. At that point, the self can no longer be understood only as an inner psychological feeling or an abstract philosophical subject. It also has to be understood as a structure that can continue to operate in the world, continue to be recognized, and continue to bear consequences.
At first, putting the identity layer and the energy layer together was simply a direct technical need for me: I wanted to give DSSCs an identity so that the generative music could be reconstructable. Later I gradually realized that this issue is not only about the work itself. Any system that can persist and continue to act over time cannot exist on the basis of a name, an account, or a cryptographic key alone. It also needs physical conditions: how it gets energy, how it sustains operation, how it senses its environment, and how it leaves continuous traces of behavior. Without those conditions, identity is too easily reduced to a symbolic shell. It can be copied, rented, transferred, or simulated, but it becomes difficult to say that it is truly a being with continuity of its own.
My decision to tie the identity layer to the energy layer also came, in part, from an intuition about future infrastructure. If a more explicit energy economy really does emerge, identity may come to function like the email address in the early internet: a piece of infrastructure that is built first and only later gradually permeates all kinds of behavior. And for AI agents, this seems especially reasonable. They need energy in order to act, and they also need sensing and hardware conditions in order to maintain continuity. From that perspective, tying the source of identity to the source of energy is not only a means of verification, but also a way of building a more credible boundary for a new form of existence.
A system resembles a self not only because it can say “this is who I am,” but because it can maintain a certain recognizability within noise, change, and constraint. It has its own input-output relations, its own resource boundaries, and its own costs. From this perspective, the important question in the future may not be whether AI becomes more intelligent than humans, but what standards we will use to judge whether a system is really a self. This question will appear in the design of AI agents. Once an agent can carry out long-term tasks, call tools, manage assets, and make decisions on behalf of a user, we have to ask: where does its continuity come from? From the model itself? From memory? From an account? Or from a stable coupling with some hardware, some sensing condition, or some energy source?
So far, mind philosophy is no longer pure philosophy, but consciousness science and an important role in modern science. I believe these questions will also appear in the design of brain-computer interfaces and neural implants. Once memory, attention, perception, or decision-making can be participated in by external systems, the boundary of the self will no longer be so clear. At that point, the “I” may no longer be simply a naturally formed psychological fact, but a structure rearranged by technical conditions. In that case, does the self become extended, divided, or platformized? These will become very practical questions.
染敏飢餓合成器
DSSC starvation synth
在今年二月與 Marc Dusseiller 於台北進行的前期研究後，本次為期五週的 KUBU 駐村計畫正式展開，作為「The Garden And The Hedge 夏季展覽與計畫」的一部分，由 Teresa Dillon 和 Tuomo Tammenpää 策劃與組織。
本藝術駐村研究旨在於染料敏化太陽能電池中開發兼具圖像和功能的W-type串聯電路設計，以同時滿足藝術與電子工程的需求。其挑戰在於如何製作具圖像特徵的二氧化鈦與鉑電極，同時保持良好的電氣特性與轉換效率。
此次所開發的概念驗證原型預計未來將基於 Voronoi 分形結構，產生更複雜的圖案。這些圖像不僅具有藝術價值，也具備工程應用的潛力。例如，此原型未來可能應用於進一步研究，如自供能的機器人眼睛、智慧隱形眼鏡或太陽眼鏡等，因為分形結構的串聯電路可與無需對焦的微透鏡陣列系統整合，進而實現 DSSC 在智慧視覺系統中的創新應用。
本藝術駐村研究旨在於染料敏化太陽能電池中開發兼具圖像和功能的W-type串聯電路設計，以同時滿足藝術與電子工程的需求。其挑戰在於如何製作具圖像特徵的二氧化鈦與鉑電極，同時保持良好的電氣特性與轉換效率。
此次所開發的概念驗證原型預計未來將基於 Voronoi 分形結構，產生更複雜的圖案。這些圖像不僅具有藝術價值，也具備工程應用的潛力。例如，此原型未來可能應用於進一步研究，如自供能的機器人眼睛、智慧隱形眼鏡或太陽眼鏡等，因為分形結構的串聯電路可與無需對焦的微透鏡陣列系統整合，進而實現 DSSC 在智慧視覺系統中的創新應用。
Few previous researches has been made before this residency in Taipei with Marc Dusseiller in February. This 5-weeks residency at KUBU, is part of the The Garden And The Hedge Summer Exhibition And Program, organized and curated by Teresa Dillon and Tuomo Tammenpää.
This artist in residency research aim to develop graphical W-type series connection in DSSC to satisfy both artistic and electric engineering purposes; the challenge is to have graphical TiO₂ and platinum electrodes without failing it's electric properties and efficiency.
This proof of concept prototype is expected to generate more complicated pattern base on the voronoi fractal foundation in the future. The graphics could be also engineering useful too, for example, this POC could be useful for further researches, such as self-powered robotic eyes, smart contact lens or sunglasses in the future due to the fractal series circuits could corporate with the micro lens array in the focus-free lens system to exploit DSSC in the manner of smart visions.
This artist in residency research aim to develop graphical W-type series connection in DSSC to satisfy both artistic and electric engineering purposes; the challenge is to have graphical TiO₂ and platinum electrodes without failing it's electric properties and efficiency.
This proof of concept prototype is expected to generate more complicated pattern base on the voronoi fractal foundation in the future. The graphics could be also engineering useful too, for example, this POC could be useful for further researches, such as self-powered robotic eyes, smart contact lens or sunglasses in the future due to the fractal series circuits could corporate with the micro lens array in the focus-free lens system to exploit DSSC in the manner of smart visions.
一百五十四萬奈米的適當離網染料敏化太陽能
1,540,000nm of DSSC-Appropriate Off-Grid
雖然當今最先進的半導體製程已達 2 奈米，「1,540,000 奈米」意在象徵一種 DIY 精神。這檔展覽是關於「適切的離網太陽能狂歡」的場域特定計畫——聚集並分享自身創造的能量。染料敏化太陽能電池（DSSC）在當代以「離網 off-grid」小裝置追求能源自主的論述中，或許顯得陳舊而不切實際，然而它體現了去中心化、低成本可複製，以及「適切科技」的願景。
展覽聚焦於施惟捷對 DIY DSSC 製作的研究，並透過教學型工作坊與對談式實驗室，把科學學習的方法轉化為一種社會性教學。運用自製工具、臨時材料與在地植物作為染料來源，計畫開啟了關於開源能源、DIY 太陽能技術，以及在課堂內外進行原型製作之挑戰等關鍵提問。
施惟捷持續以 DIY 方法重塑先進技術，作為對「異化科技」的一種反叛；其脈絡橫跨藝術、科學與展覽實踐，並與市場與專有技術保持距離。
以獨立自己自足的視野為出發點，施惟捷將 hack lab 的方法帶入混合型社群實驗室（hybrid social labs），串連非營利獨立空間，橋接科技與知識社群，拼裝完成本計畫。這檔展覽是一種公開的立場宣言——不躲在黑箱後，而是體現在不按常規、往往凌亂卻活躍的社群教育中，以回應能源脆弱性與知識階序。
展場中的太陽能板也以開源資源的方式分享此意圖：手工製作並公開釋出，作為他人可複製的開源裝置；其中的瑕疵被保留為具有生成潛力的特性。透過這些「太陽能小書」，作品傳遞自我賦能的訊息：高科技門檻不是令人疏離的，而可以是人人得以親手創造與創新的事物。
展覽聚焦於施惟捷對 DIY DSSC 製作的研究，並透過教學型工作坊與對談式實驗室，把科學學習的方法轉化為一種社會性教學。運用自製工具、臨時材料與在地植物作為染料來源，計畫開啟了關於開源能源、DIY 太陽能技術，以及在課堂內外進行原型製作之挑戰等關鍵提問。
施惟捷持續以 DIY 方法重塑先進技術，作為對「異化科技」的一種反叛；其脈絡橫跨藝術、科學與展覽實踐，並與市場與專有技術保持距離。
以獨立自己自足的視野為出發點，施惟捷將 hack lab 的方法帶入混合型社群實驗室（hybrid social labs），串連非營利獨立空間，橋接科技與知識社群，拼裝完成本計畫。這檔展覽是一種公開的立場宣言——不躲在黑箱後，而是體現在不按常規、往往凌亂卻活躍的社群教育中，以回應能源脆弱性與知識階序。
展場中的太陽能板也以開源資源的方式分享此意圖：手工製作並公開釋出，作為他人可複製的開源裝置；其中的瑕疵被保留為具有生成潛力的特性。透過這些「太陽能小書」，作品傳遞自我賦能的訊息：高科技門檻不是令人疏離的，而可以是人人得以親手創造與創新的事物。
Although the most advanced semiconductor manufacturing process today reaches 2 nanometers, the title “1,540,000 nanometer” is meant to symbolize the spirit of DIY. This exhibition is a site-specific project on appropriate off-grid solar rave—gathering and sharing one’s own creative energy. Dye-sensitized solar cells (DSSCs) might appear obsolete and entirely impracticable in the modern discourse on “off-grid” gadget-based energy independence, yet they embody decentralization, low-cost reproducibility, and a vision of appropriate technology.
This exhibition highlights Shih Wei Chieh’s research into DIY DSSC fabrication, featuring tuition workshops and talk labs that reframe methods of science learning into social pedagogy. Using DIY tools, makeshift materials, and local plants as dye sources, the project opens critical questions about open source energy, DIY solar technologies, and the challenges of prototyping in classrooms and beyond.
Shih Wei Chieh continues to reinvent advanced technology through DIY methods as a rebellion against alienated technology in the context of art, science, and modes of exhibition against market and proprietary technologies.
With a vision of self-reliant access, Abao incorporates hack labs in hybrid social labs, along with activities in Asia and Europe that bridge technology and knowledge communities. This exhibition is a declaration of interest—not hidden behind black boxes, but embodied by an education in offbeat and often messy communities that respond to both energy precarity and knowledge hierarchies.
The solar panels on display share this intention as an open-source resource: handmade and openly available, these functions act as open-source devices replicable by others, with imperfections left as generative affordances. Through these solar notebooks, the work communicates a self-empowering message that technology does not have to be inaccessible or alienating, but could be something everyone can create and innovate with their own hands.
This exhibition highlights Shih Wei Chieh’s research into DIY DSSC fabrication, featuring tuition workshops and talk labs that reframe methods of science learning into social pedagogy. Using DIY tools, makeshift materials, and local plants as dye sources, the project opens critical questions about open source energy, DIY solar technologies, and the challenges of prototyping in classrooms and beyond.
Shih Wei Chieh continues to reinvent advanced technology through DIY methods as a rebellion against alienated technology in the context of art, science, and modes of exhibition against market and proprietary technologies.
With a vision of self-reliant access, Abao incorporates hack labs in hybrid social labs, along with activities in Asia and Europe that bridge technology and knowledge communities. This exhibition is a declaration of interest—not hidden behind black boxes, but embodied by an education in offbeat and often messy communities that respond to both energy precarity and knowledge hierarchies.
The solar panels on display share this intention as an open-source resource: handmade and openly available, these functions act as open-source devices replicable by others, with imperfections left as generative affordances. Through these solar notebooks, the work communicates a self-empowering message that technology does not have to be inaccessible or alienating, but could be something everyone can create and innovate with their own hands.
Loose Energy Curriculum At The Humuspunk Library + Bioclub + SGMK residency
這個實驗藝術駐村行動由藝術網絡、創客空間以及獨立實驗室組成，地點橫跨日本和瑞士，為時 3 個月，內容由研究駐村和藝術展覽組成。我在日本 Bioclub Tokyo 和 Fabcafe Tokyo 初步進行了奈米銀線合成。接下來在蘇黎世的 SGMK的Hacker in residency 進行後續實驗，並同步在 Regenerative Energy Community （REC）組織的 「ENERGY GIVEAWAY AT THE HUMANPUNK LIBRARY」 展覽中展出作品。
本駐村研究內容與目標圍繞著 DIY 太陽能紡織品，包括奈米銀線的合成方法。奈米銀線是製造透明且導電薄膜的重要材料，也是許多太陽能穿戴織品的解決方案。雖然最終沒有成功合成出足夠長度的奈米銀線，但是其合成技術在製造藝術品和藝術表現方法層面上有許多啟發。
這個游牧式的研究型展覽計劃藉由三個獨立組織共同完成，可以被看成在策展資源不足下，獨立藝術網絡和藝術家之間透過協作調配靈活運用現有資源組織非典型駐村模型的範例。
本駐村研究內容與目標圍繞著 DIY 太陽能紡織品，包括奈米銀線的合成方法。奈米銀線是製造透明且導電薄膜的重要材料，也是許多太陽能穿戴織品的解決方案。雖然最終沒有成功合成出足夠長度的奈米銀線，但是其合成技術在製造藝術品和藝術表現方法層面上有許多啟發。
這個游牧式的研究型展覽計劃藉由三個獨立組織共同完成，可以被看成在策展資源不足下，獨立藝術網絡和藝術家之間透過協作調配靈活運用現有資源組織非典型駐村模型的範例。
This self-organized residency program, focused on DIY dye-sensitized solar cell textiles, is comprised of several components primarily located in Japan and Switzerland, spanning a total duration of three months. The initial phase unfolded at Bioclub Tokyo within Fabcafe Tokyo, while the subsequent segment transpired at the Hackteria Open Source Biological Arts Platform in Zurich, indirectly receiving support from the exhibition "Energy Giveaway at The Humuspunk Library," orchestrated by the Regenerative Energy Community (REC).
The residency research centers on the production of flexible, functional dye-sensitized solar cell films or textiles through a DIY protocol. Additionally, a minor aspect involves preliminary experiments with silver nanowires and the synthesis method of it. A crucial material for crafting transparent and flexible conductive substrates at room temperature. The conclusive outcomes of the residency experiments are detailed in the provided wiki link.
The residency research centers on the production of flexible, functional dye-sensitized solar cell films or textiles through a DIY protocol. Additionally, a minor aspect involves preliminary experiments with silver nanowires and the synthesis method of it. A crucial material for crafting transparent and flexible conductive substrates at room temperature. The conclusive outcomes of the residency experiments are detailed in the provided wiki link.
. 
DIY大型染料敏化太陽能電池 DIY Large Dye Sensitized Solar Cell
本實驗挑戰製造一個 30 × 60 cm 的大型染料敏化太陽能電池。染料敏化太陽能是一種易於製造且具製造成本相對低的光伏裝置。本實驗所使用的化學試劑主要來自 Greatcell Solar，成本約為每平方公分0.7美金。然而，此DIY版本的轉換率仍相對低，遠不及商業標準，這主要受限於系統內 FTO 或 ITO 層的高方阻、缺乏沙林膜封裝和電解液分布不均。
本實驗原型由12個子電池構成，並以 Z 型串聯組態構成，每個子電池尺寸為3x28cm，間距為2公分。每個次電池為並在對電極上印刷銀電極線，銀線寬度約為5mm，測得開放電壓5.8V 和短路電流 53mA。
一項主要的發現是以玻璃窯450ºC燒結的大型玻璃具有熱變形現象，這造成了注入電解液的困難。電解液注入得助於兩片平整玻璃間毛細作用的幫助，任何輕微的變形都會破壞毛細作用的維持。升溫速度約為每分鐘8ºC，到達450ºC後持溫30分鐘，再於窯內自然降溫到室溫。
本實驗原型由12個子電池構成，並以 Z 型串聯組態構成，每個子電池尺寸為3x28cm，間距為2公分。每個次電池為並在對電極上印刷銀電極線，銀線寬度約為5mm，測得開放電壓5.8V 和短路電流 53mA。
一項主要的發現是以玻璃窯450ºC燒結的大型玻璃具有熱變形現象，這造成了注入電解液的困難。電解液注入得助於兩片平整玻璃間毛細作用的幫助，任何輕微的變形都會破壞毛細作用的維持。升溫速度約為每分鐘8ºC，到達450ºC後持溫30分鐘，再於窯內自然降溫到室溫。
This experiment challenges the fabrication of a large-scale dye-sensitized solar cell (DSSC) with dimensions of 30 × 60 cm. DSSCs are a type of photovoltaic device known for their relatively low-cost and accessible fabrication process. The chemical reagents used in this experiment were mainly sourced from Greatcell Solar, with a material cost of about 0.7 USD per cm².
However, the conversion efficiency of this DIY version remains relatively low and falls short of commercial standards. The main limitations arise from the high sheet resistance of the FTO/ITO layer, the absence of Surlyn film sealing, and the incomplete injection of electrolyte.
The prototype consists of 12 sub-cells connected in a Z-shaped series configuration. Each sub-cell measures 3 × 28 cm with a 2 cm spacing. On the counter electrode, silver electrode lines (approx. 5 mm wide) were screen-printed. The device demonstrated an open-circuit voltage of 5.8 V and a short-circuit current of 53 mA.
One key finding is that large glass substrates sintered at 450 °C in a glass kiln showed noticeable thermal deformation, which created difficulties during electrolyte injection. Electrolyte filling typically relies on capillary action between two flat glass plates, and even slight warping disrupts this mechanism.
The thermal profile followed a ramp-up rate of about 8 °C per minute to 450 °C, held for 30 minutes, and then cooled naturally in the kiln back to room temperature.
However, the conversion efficiency of this DIY version remains relatively low and falls short of commercial standards. The main limitations arise from the high sheet resistance of the FTO/ITO layer, the absence of Surlyn film sealing, and the incomplete injection of electrolyte.
The prototype consists of 12 sub-cells connected in a Z-shaped series configuration. Each sub-cell measures 3 × 28 cm with a 2 cm spacing. On the counter electrode, silver electrode lines (approx. 5 mm wide) were screen-printed. The device demonstrated an open-circuit voltage of 5.8 V and a short-circuit current of 53 mA.
One key finding is that large glass substrates sintered at 450 °C in a glass kiln showed noticeable thermal deformation, which created difficulties during electrolyte injection. Electrolyte filling typically relies on capillary action between two flat glass plates, and even slight warping disrupts this mechanism.
The thermal profile followed a ramp-up rate of about 8 °C per minute to 450 °C, held for 30 minutes, and then cooled naturally in the kiln back to room temperature.