Basically one neural network, called the imagitron, is bathed in simulated heat. It generates new ideas. The other, called the perceptron, monitors the first. It has opinions and governs the amount of heat stimulating the imagitron.... I would smile or frown as the machine generated sound and the perceptron could see my reactions. If I liked something, it would remember that and map a pattern of how I would rate a song, freezing in memories of likable tunes.Thaler plans to release an album of machine-created music on iTunes in the near future.
This isn't Thaler's first foray into composition. In 1994, he claimed U.S. Copyright no. PAu-1-920-845 on 11,000 "musical hooks" generated by the Creativity Machine. From the company's history:
October, 1995 [sic], Most Prolific Musician of All Time. A Creativity Machine, trained on top-ten melodies over the preceding 30 years proceeds to invent 11,000 new musical hooks that are promptly copyrighted. Interestingly enough human musical artists disdain the copyright, protesting that "only human minds can conceive music!" Thaler holds back on a million song database generated via Creativity Machine in view of the spirited response from human artists. ...On the flip side, computational musicians don't get it either, feeling that they can do the same thing, spending months or years concocting new works of art via computer. What they don't realize was that only a few hours had been spent by Thaler in translating sheet music to a representation more conducive to neural nets. Beyond that, Creativity Machine function was spontaneous and voluminous.Just in case you weren't sure that Thaler was prone to confuse quality with quantity, he's also copyrighted (no. TX-5-725-954) a million new machine-generated English words.
Let's leave aside the absolute inanity of the Copyright office on this. (Every time I write a new phrase, I'm supposed to check and make sure I'm not infringing on 11,000 randomly-generated variations? Please.) If you get the sense that Thaler has a tendency to indiscriminately anthropomorphize the output of these neural networks, you're right. From the article:
Q. And music from your computers is different?No database of prior music—that is, except the examples that Thaler's fed into the machine as raw materials. It's not filled with feeling, it's filled with processes gradually filtered out by a feedback loop. The filtering itself, in its speed and efficiency, may be impressive, but all it's doing is throwing sonic spaghetti against the refrigerator door of Thaler's ear and noting which strands stick.
A. It's filled with feeling. It's generated much the same as music from the mind of a human composer. It is spontaneously and autonomously generated by machines, using only raw emotional response from a human being. There are no explicit rules, no databases of prior music and no templates of any kind.
No doubt Thaler would claim that's all human composers do, and we're just arrogantly attributing it inspiration or whatever because we don't consider the process. The machine, though, doesn't imagine the intended emotional state and then try to realize that state in sound. It's doing what Thaler has programmed it to do, methodically trying permutations until the desired feedback goal—Thaler smiling—is consistently attained. It's the ultimate pop music dispenser: it gives you what you like, and only what you like, all the time. Does Thaler really think this is what composers try to do? (In his patent for the machine, Thaler theorizes that ideas are noting more than "degraded memories," which makes me wonder just what it was, say, Harry Partch was trying to recall.) And note that the computer is beholden to whatever styles are used as examples. As Thaler says: "We put in a wide spectrum—classical, rock, hip-hop." The prospect of an experimental aesthetic only arises if the listener providing feedback has formulated that aesthetic already.
Don't get me wrong—the technology certainly sounds cool. It's just that Thaler's reading into it something that's just not there. In between programming machines to design warheads or toothbrushes, Thaler has started In Its Image, Inc., a non-profit dedicated to "Raising awareness, through public seminars, of the artificial intelligence technology that promises to be the computational vehicle for immortality." The idea is that we'll all upload our meat-based, decrepit neural networks onto robust Creativity Machines and thus maintain our own unique consciousness indefinitely.
This sounds suspiciously like a humble cousin to a scenario laid out by physicist Frank Tipler in his book The Physics of Immortality, which proposes the universe evolving into an all-powerful quantum computer that will simultaneously recreate every possible state of consciousness in a perpetual virtual reality. If that sounds fishy to you, well, you're not alone. Thaler doesn't go nearly that far, but reducing human experience to mere neural interaction implicitly devalues the powerful artistic spur of human fallibility, decay, imperfection—the magical, cascadingly fruitful effort of pushing up against the boundaries of time, understanding, and physical ability. I admit, this might just be my own post-Romantic bias against crude 17th-century body vs. soul dualism—an analysis of this blog's archive would no doubt reveal a standing prejudice against the immutably unbreakable. The cultural expressions I treasure are the products of imaginations inseperably bound up with the frustration, desperation, and exhilaration of corporeal existence.
Besides, the proof is in the pudding, right? Well, if this brief, Mannheim-Steamroller-worthy MP3 sample of Thaler's handiwork is really the music of an immortal future, I have to admit, a mortal opt-out is looking better and better.