How does this music make you feel? Do you find it beautiful? Is it creative? Now, would you change your answers if you learned the composer was this robot? Believe it or not, people have been grappling with the question of artificial creativity, alongside the question of artifcial intelligence, for over 170 years. In 1843, Lady Ada Lovelace, an English mathematician considered the world's first computer programmer, wrote that a machine could not have human-like intelligence as long as it only did what humans intentionally programmed it to do. According to Lovelace, a machine must be able to create original ideas if it is to be considered intelligent. The Lovelace Test, formalized in 2001, proposes a way of scrutinizing this idea. A machine can pass this test if it can produce an outcome that its designers cannot explain based on their original code. The Lovelace Test is, by design, more of a thought experiment than an objective scientific test. But it's a place to start. At first glance, the idea of a machine creating high quality, original music in this way might seem impossible. We could come up with an extremely complex algorithm using random number generators, chaotic functions, and fuzzy logic to generate a sequence of musical notes in a way that would be impossible to track. But although this would yield countless original melodies never heard before, only a tiny fraction of them would be worth listening to. With the computer having no way to distinguish between those which we would consider beautiful and those which we won't. But what if we took a step back and tried to model a natural process that allows creativity to form? We happen to know of at least one such process that has lead to original, valuable, and even beautiful outcomes: the process of evolution. And evolutionary algorithms, or genetic algorithms that mimic biological evolution, are one promising approach to making machines generate original and valuable artistic outcomes. So how can evolution make a machine musically creative? Well, instead of organisms, we can start with an initial population of musical phrases, and a basic algorithm that mimics reproduction and random mutations by switching some parts, combining others, and replacing random notes. Now that we have a new generation of phrases, we can apply selection using an operation called a fitness function. Just as biological fitness is determined by external environmental pressures, our fitness function can be determined by an external melody chosen by human musicians, or music fans, to represent the ultimate beautiful melody. The algorithm can then compare between our musical phrases and that beautiful melody, and select only the phrases that are most similar to it. Once the least similar sequences are weeded out, the algorithm can reapply mutation and recombination to what's left, select the most similar, or fitted ones, again from the new generation, and repeat for many generations. The process that got us there has so much randomness and complexity built in that the result might pass the Lovelace Test. More importantly, thanks to the presence of human aesthetic in the process, we'll theoretically generate melodies we would consider beautiful. But does this satisfy our intuition for what is truly creative? Is it enough to make something original and beautiful, or does creativity require intention and awareness of what is being created? Perhaps the creativity in this case is really coming from the programmers, even if they don't understand the process. What is human creativity, anyways? Is it something more than a system of interconnected neurons developed by biological algorithmic processes and the random experiences that shape our lives? Order and chaos, machine and human. These are the dynamos at the heart of machine creativity initiatives that are currently making music, sculptures, paintings, poetry and more. The jury may still be out as to whether it's fair to call these acts of creation creative. But if a piece of art can make you weep, or blow your mind, or send shivers down your spine, does it really matter who or what created it?
你對於這段音樂有什麼感覺? 你覺得它優美嗎? 它具有創意嗎? 現在,如果你知道這段音樂 是由機器人創作的, 你會改變之前的答案嗎? 信不信由你, 人們試圖解決「人工創意」的問題, 以及相關的人工智慧議題, 已經研究了超過170年。 在1843年,英國數學家 愛達·勒芙蕾絲夫人, 她是後人公認的 史上第一位電腦程式設計師。 她認為,機器不可能會有 和人類一樣的智慧, 因為它只會忠實執行 人類所要求的工作。 根據勒芙蕾絲的觀點, 如果一部機器具有智慧, 它就必須能夠產生 具有原創性的構想。 直到2001年, 人們提出了「勒芙蕾絲測試」, 用來檢驗機器是否具有智慧。 如果機器所創造出的作品, 是它的設計師無法用程式解釋的, 就代表這部機器就具有智慧。 就實驗設計來看, 勒芙蕾絲測試比較像是「思想實驗」, 而不屬於客觀的科學實驗。 但這是一個起點。 乍看之下, 機器似乎不可能 創作出高品質的原創音樂。 我們可以用複雜的演算法, 結合隨機產生的變數、 混沌公式和模糊邏輯, 來產生一連串的音符, 這個過程無法從原始程式解讀、追踪。 雖然這會產生無數的、 未曾聽過的原創旋律, 但是其中只有一小部分 是令人覺得悅耳的。 目前電腦仍然無法分辨 人們認為好聽的音樂 以及不好聽的音樂。 但是,如果我們退一步來想, 試著模擬出 形成創意的自然過程呢? 目前,我們已經知道 至少一種以上的過程, 它可以產生出原創的、有價值的、 甚至是悅耳的音樂作品, 這個過程稱為「演化」。 包括進化式演算法, 或是模仿生物進化的「基因演算法」, 都是可能的研究方向, 能讓機器產生出具原創性, 而且有價值的藝術作品。 那麼,什麼樣的演化過程, 能使機器擁有創作音樂的能力? 不同於一般生物的演化, 我們先以樂句(musical phrases) 作為演化的初始群體, 然後用基本的演算法 來模仿複製與隨機變異 例如將樂曲的某些部分進行交換、 互相組合, 以及隨機取代某些音符。 現在,我們產生了新的樂句, 接著我們用「適應函數」來進行選擇。 正如同生物的適應性, 是由外在環境的壓力來決定, 我們的適應函數, 是根據音樂家或音樂愛好者 所挑選出的樂曲旋律來決定, 這些代表了人們認為最優美的旋律。 然後,透過演算法 將機器創作出來的樂句 與人們所挑選的優美旋律 進行比較, 並選出相似性最高的樂句。 淘汰掉那些 相似性較低的樂句之後, 演算法會將剩下的部分, 再次進行突變和重組, 然後再次從新一代的樂句當中, 選出相似性或適應性最高的部分, 這樣的步驟 會重複進行數個循環。 在這個創作過程中, 有許多的隨機性和複雜性, 所以能夠通過勒芙蕾絲的測試。 更重要的是,在創作的過程中 結合了人類的審美觀, 所以根據理論,我們能夠產生出 讓人覺得優美的旋律。 但是,這是否符合人們直覺上 所認為的「真正的創意」? 這樣的創作方式, 是否能得到原創而且優美的作品? 創作者在過程當中, 是否要有目的、能自我察覺, 才能夠稱為創意? 也許在這種情況下, 創意是來自於程式設計師, 但是他們並不瞭解整個創作過程。 究竟,什麼是人類的創造力? 也許,創意並不僅僅是 一個由生物演算過程所發展而成, 相互連結的神經元系統, 也不僅僅是 形成我們的生活的一些隨機經驗。 秩序與混亂,機器和人類。 這些共同構成了 機器創作的核心動力, 使得機器能創作出 音樂、雕塑、繪畫、詩歌和更多的作品。 我們仍然無法定論 機器創作是否可以算是 具有創意。 但是如果一件藝術作品, 能夠令人感動, 或是令你畢生難忘, 甚至是令你覺得毛骨悚然, 究竟作者是人還是機器, 真的重要嗎?