模擬.感知

在馬修·斯圖爾特(Matthew Stewart)在《艾利評論》(Avery Review)中題為“官僚機構的體系結構:WeWork和算法設計”的文章中,斯圖爾特指出,WeWork 的把工作包裝成一種具崇高想像力的生活方式,而並不僅是一種平凡的勞動。同時,WeWork 的內部設計部門使用電腦自動化的程式生產最優化的室內設計,把每個空間的桌子數量最大化,讓WeWork達到最佳成本效益。 WeWork 對工作的包裝和他們的設計實踐造成了諷刺的反差感。在一個勞動密集型工作越來越被自動化取代的時代,WeWork聲稱,傳統上一些精神密集型的工作(例如設計)也是無法避免被電腦自動化替代。他們認為,若電腦自動化可以解決某些重複性的設計任務,這就能有效地將設計任務系統化,讓人類設計師有更多時間專注於考量其他更重要、更具人性化的設計,從而節省成本,提高可靠性和生產率。為此,WeWork的內部算法是根據參數輸入,然後生成房間和家具的佈局,最終獲得最優化的結果。在這種情況下,即使是工作室內的辦公桌數量龐大,但經參數算法而成的工作空間仍然會為用家帶來令人滿意的設計。

因此,相比之下,人類似乎在優化和速度的領域中處於劣勢。 AI是經驗主義者,僅依靠外部信息進行操作。另一方面,人是理性主義者,我們對世界的了解是與生俱來的,並且我們懷有感知力。這二元性的闡述延伸到以下幾個問題:

1. AI可以模擬情感並反映人類獨有的特質嗎?

2.語言可以作為一個輸入參數,從而輸出一個有感知的結果嗎?



Simulating.Sentience

In Matthew Stewart’s piece in The Avery Review titled “The Architecture of Banal Bureaucracy: WeWork and Algorithmic Design”, Stewart pointed out the eery juxtaposition of the promotion of a cultish imagination of work not as mundane labour but as a way of life, with the banal automation WeWork uses to generate space.  In an era where physically intensive jobs are increasingly on their way to be replaced by automation, WeWork claimed that traditionally mentally intensive jobs such as design would not be safe from the replacement by automation either.  They argued that automation would increase cost savings, reliability, and productivity by systemising repetitive tasks, freeing up the human designer to use their creativity in other ways.  To accomplish this, WeWork’s in house algorithm allows room and furniture layouts to be generated upon the input of parameters, ultimately to have the most optimised outcome.  In this case, the maximum number of desks while still providing spatially satisfactory experience.  

Therefore, in comparison, humans seem to be subpar in the realm of optimisation and functional accelerationism.  The AI is an empiricist, relying solely on external information to operate.  On the other hand, the human is a rationalist, and our knowledge of the world is innate and we operate with sentience.  The elaboration of this duality extends to a few queries:

1. Can AI simulate sentience and respond to subjective human qualities?

2. Can language be an input parameter to generate sentient algorithmic outputs?