You Can’t Fool All The People All The Time: Why Computer Music Instruments Don’t Make Music Like Acoustic Music Instruments
Robin Daniel Z. Rivera
Socio 297: Technology and Society
Fr. Tagura
30 August 2002
The debate about artificial intelligence has brought out fundamental questions about what actually constitutes human intelligence, and if this can be replicated by computers. A similar debate can be made about the difference between acoustic and computer music instruments. This paper will attempt to apply the binary nature of the argument to the problem of why virtual computer music instruments are theoretically opposed to natural acoustic musical instruments.
Herbert Dreyfuss is a long time critic of the philosophical basis for artificial intelligence (AI). Much of his criticisms are aimed at what is now called Good Old Fashioned Artificial Intelligence or GOFAI. Ron Barnett describes Dreyfuss’ critique:
“In a tone of almost "I told you so---but why won't it go away?!," Dreyfus tries to lay to final rest the AI approach dubbed by John Haugeland (1985) 'Good Old Fashioned Artificial Intelligence,' or GOFAI. Inspired by a Turing model of intelligent behavior as essentially computational, the thesis of GOFAI is that the processes underlying intelligence are symbolic in nature. More specifically: GOFAI models human intelligence as von Neumann computational architectures that (1) perform computations on abstract symbolic representations, (2) by computations governed by a stored program that contains an explicit list of instructions or rules which transform these symbolic representations into new symbolic states, and (3) in terms of which these computations are performed in serial fashion by a CPU that has information stored in the computers' permanent memory. As such, GOFAI depicts mentality within the context of what philosophers know as the Representational Theory of Mind, according to which the mind is an entity which performs calculations over mental representations, or inner tokens or symbols which refer to features of the outer world. In short, the mind is thus viewed as a symbolic information-processing device, operating in a serial fashion, and governed by rules which are, at base, the language of thought.”(1)
A simple example of a GOFAI isa linear process in which 1) knowledge is reduced to matter in the form of information, 2) information is inputted and stored into to the computer, 3) information is invoked in response to a query for appropriate action, 4) the computer will scan the database of information for an appropriate action based on existing information, 5) the selected action will be carried out by the computer to a connected controlled device, or relayed to the user. More recent types of AI such as Fuzzy Logic (FL) attempts to to incorporate vagueness into the process by deliberately obscuring the precision of both inputted information, the resulting response to queries using this information, and the resulting action. But the model of FL remains serial in nature, and still cannot simulate the essentially non-linear nature of human thought. Dreyfus critiques even more recent AI models such as Parallel Distributed Processing (PDP), or Connectionist, architectures and programming because it still lacks many human variables such as context and relevance. But Dreyfus’ nevertheless argues that intelligence is not just a function of the brain, but of a host of other components that make up the human body, the wealth of an individual’s human experience, and an infinite number of relationships that occur in the interaction of these components and experiences. Dreyfus leans toward a phenomenological take on intelligence, bordering on the metaphysical, that not only basically defies quantification, and suggests that until computers can never be capable of true intelligence as long as rationality remains tied down to positivist empiricism.
The same argument can be applied to the difference between acoustic and computer musical instruments. The basic philosophies in these two types of devices emerge from opposite sides of the spectrum, and despite efforts to forge a compromise, this remains elusive.
One root concept in virtual computer music instruments is the representation of sound as waves of air pressure. This representation was applied not just to CMI’s, but to electro-mechanical, and analog electronic musical instruments as well. But CMI’s take this one step further because the characteristics of the system that creates the sound can be stored in digital memory and called up at any given time. This kind of predictability is one of the virtues of computer control. A single “natural” sound can be dissected and a CMI can simulate or replicate the waveform. The strategy is similar to that of artificial intelligence in that a finite number of sound objects, and all their empirical characteristics are collected to form a database that will determine the range of possible outcomes. There are three popular methods for creating sounds in CMIs, these are Subtractive Synthesis, Additive Synthesis, and Sampling.
In subtractive synthesis, a digital oscillator produces a complex but periodic waveform. The waveform is then subjected to various forms of filtering to until it is deemed usable. The filtering comes in many forms. For example, Timbre (or harmonic content) envelope (or time-related characteristics), or modulation (vibrato) can be manipulated.
Additive synthesis, on the other hand, combines simple (but still periodic) waveforms to obtain complex waveforms. While these waveforms can be further subjected to the filtering techniques similar to subtractive synthesis, the process remains opposed because the sound is a product of the interaction between simple waveforms.
Sampling differs from the two because it begins with a digital recording, or “sample” of a natural sound and not from an oscillator giving off periodic waveforms. Like the other two methods, filtering can also be applied to modify its content. But unlike them, the waveform and be literally “redrawn” into a different waveform.
Whatever method of generation or filtering is used, the basic waveform is always the same. A square wave from a subtractive synthesis digital oscillator will always be the same every time is is invoked. A complex wave from an additive synthesis algorhythm will always be the same when invoked. And a sample, when invoked, will always be the same because it is a recording of a single sound event. Each digitally generated sound is a result of a finite database of waveform information. While such waveforms can be manipulated before, during, and even after performance (in the case of recorded musical performances), the possible variations are inherently limited by the nature of the “basic” waveform. In addition to this, CMI’s can be activated by computer sequencers, which store waveform and performance information in computer memory, and can invoke a stored performance identically, as often as needed. This gives CMI and computer sequencer users a great deal of certainty whenever a musical performance is invoked.
Acoustic instruments and sounds come to us from an entirely different lifeworld. It is safe to say that in nature, each individual sound is unique and cannot be repeated, regardless of the source of the sound. This can also be true for the sounds from natural, physical musical instruments. This is because acoustic musical instruments are subject to a completely different set of variables from CMIs. For example, one particular note in an instrument like a acoustic piano, played in apparently the same manner, can never duplicate the exact same sound twice. This is because, among other, the natural environment is always changing, thus affecting the sound as well as the characteristics of the material of the instrument. This is unlike a virtual environment that can controlled and or replicated. Second, the sound of acoustic instruments are usually activated by a human user, and humans are not theoretically known for precise repetition. Human musicians usually approach a level of repetitive precision by regular and intense practice. But no matter how consistent a musician may be, a performance always has an air of uncertainty. CMI’s, on the other hand, can be called to precisely replicate the activation of a waveform ad nauseum. The list of variables goes on and on. So every sound in a natural environment is based on a unique phenomenon with an infinite number of physical variables, the inherent uniqueness of every human action in the activation of the instrument, and an overall sense of uncertainty that preceeds the production of a sound.
Each and every natural sound from an acoustic musical instrument is unique, despite human attempts to make them consistent. While musical virtuosos are valued for their consistency, they are also treasured for their ability to come up with new sounds that defy sameness. This shows the ironic nature of acoustic instruments, sound, and performances. In turn, each sound in a CMI is born from a single instance of a waveform, in spite of massive efforts to create variations. Digital instruments remove at least one important variable in this ability because the root sound is just a repetition of a previously synthesized or sampled sound, and the effort needed to vary a sound is bound by a finite database of variables. While acoustic instruments thrive on uncertainty, virtual CMIs sounds boasts of unhuman consistency and predictability.
Just as the Turing Test attempted to fool some respondents into thinking that they were conversing with a human instead of a computer, CMI’s have managed to fool some listeners into thinking that they were listening to an acoustic instrument. But as the saying goes, you can’t fool all the people all of the time. I will not be as bold as Dreyfus by saying that CMI’s will never possess all the characteristics of acoustic instruments. But until a new computer algorhythm is devised to produce virtual sound in the same model as acoustic instruments, the binary nature of this argument will remain.
Footnote
(1)Ron Barnette.”A Critical Review of What Computers Still Can't Do, by Hubert Dreyfus”. <http://www.valdosta.edu/~rbarnett/phi/dreyfus.html> (20 August 2002, 22:45 PST).
Sources
Barnette, Ron. “A Critical Review of What Computers Still Can't Do, by Hubert Dreyfus ”. (20 August 2002, 22:45 PST).
Dreyfus, Herbert; Dreyfus, Stuart. “From Socrates to Expert Systems:The Limits and Dangers of Calculative Rationality”. (20 August 21:30 PST).
Truax, Barry. Acoustic Communication, 2nd edition. Westport, Connecticut: Ablex Publishing, 2001.
1Ron Barnette.”A Critical Review of What Computers Still Can't Do, by Hubert Dreyfus”. <http://www.valdosta.edu/~rbarnett/phi/dreyfus.html> (20 August 2002, 22:45 PST).
Robin Daniel Z. Rivera
Socio 297: Technology and Society
Fr. Tagura
30 August 2002
The debate about artificial intelligence has brought out fundamental questions about what actually constitutes human intelligence, and if this can be replicated by computers. A similar debate can be made about the difference between acoustic and computer music instruments. This paper will attempt to apply the binary nature of the argument to the problem of why virtual computer music instruments are theoretically opposed to natural acoustic musical instruments.
Herbert Dreyfuss is a long time critic of the philosophical basis for artificial intelligence (AI). Much of his criticisms are aimed at what is now called Good Old Fashioned Artificial Intelligence or GOFAI. Ron Barnett describes Dreyfuss’ critique:
“In a tone of almost "I told you so---but why won't it go away?!," Dreyfus tries to lay to final rest the AI approach dubbed by John Haugeland (1985) 'Good Old Fashioned Artificial Intelligence,' or GOFAI. Inspired by a Turing model of intelligent behavior as essentially computational, the thesis of GOFAI is that the processes underlying intelligence are symbolic in nature. More specifically: GOFAI models human intelligence as von Neumann computational architectures that (1) perform computations on abstract symbolic representations, (2) by computations governed by a stored program that contains an explicit list of instructions or rules which transform these symbolic representations into new symbolic states, and (3) in terms of which these computations are performed in serial fashion by a CPU that has information stored in the computers' permanent memory. As such, GOFAI depicts mentality within the context of what philosophers know as the Representational Theory of Mind, according to which the mind is an entity which performs calculations over mental representations, or inner tokens or symbols which refer to features of the outer world. In short, the mind is thus viewed as a symbolic information-processing device, operating in a serial fashion, and governed by rules which are, at base, the language of thought.”(1)
A simple example of a GOFAI isa linear process in which 1) knowledge is reduced to matter in the form of information, 2) information is inputted and stored into to the computer, 3) information is invoked in response to a query for appropriate action, 4) the computer will scan the database of information for an appropriate action based on existing information, 5) the selected action will be carried out by the computer to a connected controlled device, or relayed to the user. More recent types of AI such as Fuzzy Logic (FL) attempts to to incorporate vagueness into the process by deliberately obscuring the precision of both inputted information, the resulting response to queries using this information, and the resulting action. But the model of FL remains serial in nature, and still cannot simulate the essentially non-linear nature of human thought. Dreyfus critiques even more recent AI models such as Parallel Distributed Processing (PDP), or Connectionist, architectures and programming because it still lacks many human variables such as context and relevance. But Dreyfus’ nevertheless argues that intelligence is not just a function of the brain, but of a host of other components that make up the human body, the wealth of an individual’s human experience, and an infinite number of relationships that occur in the interaction of these components and experiences. Dreyfus leans toward a phenomenological take on intelligence, bordering on the metaphysical, that not only basically defies quantification, and suggests that until computers can never be capable of true intelligence as long as rationality remains tied down to positivist empiricism.
The same argument can be applied to the difference between acoustic and computer musical instruments. The basic philosophies in these two types of devices emerge from opposite sides of the spectrum, and despite efforts to forge a compromise, this remains elusive.
One root concept in virtual computer music instruments is the representation of sound as waves of air pressure. This representation was applied not just to CMI’s, but to electro-mechanical, and analog electronic musical instruments as well. But CMI’s take this one step further because the characteristics of the system that creates the sound can be stored in digital memory and called up at any given time. This kind of predictability is one of the virtues of computer control. A single “natural” sound can be dissected and a CMI can simulate or replicate the waveform. The strategy is similar to that of artificial intelligence in that a finite number of sound objects, and all their empirical characteristics are collected to form a database that will determine the range of possible outcomes. There are three popular methods for creating sounds in CMIs, these are Subtractive Synthesis, Additive Synthesis, and Sampling.
In subtractive synthesis, a digital oscillator produces a complex but periodic waveform. The waveform is then subjected to various forms of filtering to until it is deemed usable. The filtering comes in many forms. For example, Timbre (or harmonic content) envelope (or time-related characteristics), or modulation (vibrato) can be manipulated.
Additive synthesis, on the other hand, combines simple (but still periodic) waveforms to obtain complex waveforms. While these waveforms can be further subjected to the filtering techniques similar to subtractive synthesis, the process remains opposed because the sound is a product of the interaction between simple waveforms.
Sampling differs from the two because it begins with a digital recording, or “sample” of a natural sound and not from an oscillator giving off periodic waveforms. Like the other two methods, filtering can also be applied to modify its content. But unlike them, the waveform and be literally “redrawn” into a different waveform.
Whatever method of generation or filtering is used, the basic waveform is always the same. A square wave from a subtractive synthesis digital oscillator will always be the same every time is is invoked. A complex wave from an additive synthesis algorhythm will always be the same when invoked. And a sample, when invoked, will always be the same because it is a recording of a single sound event. Each digitally generated sound is a result of a finite database of waveform information. While such waveforms can be manipulated before, during, and even after performance (in the case of recorded musical performances), the possible variations are inherently limited by the nature of the “basic” waveform. In addition to this, CMI’s can be activated by computer sequencers, which store waveform and performance information in computer memory, and can invoke a stored performance identically, as often as needed. This gives CMI and computer sequencer users a great deal of certainty whenever a musical performance is invoked.
Acoustic instruments and sounds come to us from an entirely different lifeworld. It is safe to say that in nature, each individual sound is unique and cannot be repeated, regardless of the source of the sound. This can also be true for the sounds from natural, physical musical instruments. This is because acoustic musical instruments are subject to a completely different set of variables from CMIs. For example, one particular note in an instrument like a acoustic piano, played in apparently the same manner, can never duplicate the exact same sound twice. This is because, among other, the natural environment is always changing, thus affecting the sound as well as the characteristics of the material of the instrument. This is unlike a virtual environment that can controlled and or replicated. Second, the sound of acoustic instruments are usually activated by a human user, and humans are not theoretically known for precise repetition. Human musicians usually approach a level of repetitive precision by regular and intense practice. But no matter how consistent a musician may be, a performance always has an air of uncertainty. CMI’s, on the other hand, can be called to precisely replicate the activation of a waveform ad nauseum. The list of variables goes on and on. So every sound in a natural environment is based on a unique phenomenon with an infinite number of physical variables, the inherent uniqueness of every human action in the activation of the instrument, and an overall sense of uncertainty that preceeds the production of a sound.
Each and every natural sound from an acoustic musical instrument is unique, despite human attempts to make them consistent. While musical virtuosos are valued for their consistency, they are also treasured for their ability to come up with new sounds that defy sameness. This shows the ironic nature of acoustic instruments, sound, and performances. In turn, each sound in a CMI is born from a single instance of a waveform, in spite of massive efforts to create variations. Digital instruments remove at least one important variable in this ability because the root sound is just a repetition of a previously synthesized or sampled sound, and the effort needed to vary a sound is bound by a finite database of variables. While acoustic instruments thrive on uncertainty, virtual CMIs sounds boasts of unhuman consistency and predictability.
Just as the Turing Test attempted to fool some respondents into thinking that they were conversing with a human instead of a computer, CMI’s have managed to fool some listeners into thinking that they were listening to an acoustic instrument. But as the saying goes, you can’t fool all the people all of the time. I will not be as bold as Dreyfus by saying that CMI’s will never possess all the characteristics of acoustic instruments. But until a new computer algorhythm is devised to produce virtual sound in the same model as acoustic instruments, the binary nature of this argument will remain.
Footnote
(1)Ron Barnette.”A Critical Review of What Computers Still Can't Do, by Hubert Dreyfus”. <http://www.valdosta.edu/~rbarnett/phi/dreyfus.html> (20 August 2002, 22:45 PST).
Sources
Barnette, Ron. “A Critical Review of What Computers Still Can't Do, by Hubert Dreyfus ”. (20 August 2002, 22:45 PST).
Dreyfus, Herbert; Dreyfus, Stuart. “From Socrates to Expert Systems:The Limits and Dangers of Calculative Rationality”. (20 August 21:30 PST).
Truax, Barry. Acoustic Communication, 2nd edition. Westport, Connecticut: Ablex Publishing, 2001.
1Ron Barnette.”A Critical Review of What Computers Still Can't Do, by Hubert Dreyfus”. <http://www.valdosta.edu/~rbarnett/phi/dreyfus.html> (20 August 2002, 22:45 PST).
