Digital recreation of percussive instruments, or even classical instruments en masse is pretty convincing these days, but you still can't get a convincing syntesized guitar solo, or a convincing synthesized solo violin, or any stringed instrument, or synthetized woodwind, brass, etc..
There are two sides to this: the complexity of the tone to be synthesized (or modelled) and its capabilities for modulating tone, pitch, timbre, etc., on the one hand, and the complexity of what can be injected into those capabilities to utilize them in creative combinations, by humans, as against AIs at the current state of development.
The sound side is pretty close, actually, in many ways. Percussion (drums, including rock drums, dance beats, etc.) is a done deal. But of course that's because percussion is the easy case, because it's a one hit deal, it doesn't have to be continuously controlled and modulated like a violin.
I fancy with the developments like being able to separate sounds out of a mix, this will eventually improve, and synthesized sounds, be capable of being continuously controlled and modulated.
But the AI is the big one. I think basically because what's being expressed in music does actually happen to be something that's tied intimately to our biology (incredibly fine and refined expressions of emotional nuance), and the AI would have to model that too - i.e., AI that's capable of making music will have to wait until AI that's capable of emotion generally, comes online.
Needs AI with juices (something to model the hormonal bath and how it affects the total state of the organism qua signaller-of-internal-states).
Slashdot Mirror
Isn't a machine's "mistake" as much a deterministic output as its "functionality"?
Digital recreation of percussive instruments, or even classical instruments en masse is pretty convincing these days, but you still can't get a convincing syntesized guitar solo, or a convincing synthesized solo violin, or any stringed instrument, or synthetized woodwind, brass, etc.. There are two sides to this: the complexity of the tone to be synthesized (or modelled) and its capabilities for modulating tone, pitch, timbre, etc., on the one hand, and the complexity of what can be injected into those capabilities to utilize them in creative combinations, by humans, as against AIs at the current state of development. The sound side is pretty close, actually, in many ways. Percussion (drums, including rock drums, dance beats, etc.) is a done deal. But of course that's because percussion is the easy case, because it's a one hit deal, it doesn't have to be continuously controlled and modulated like a violin. I fancy with the developments like being able to separate sounds out of a mix, this will eventually improve, and synthesized sounds, be capable of being continuously controlled and modulated. But the AI is the big one. I think basically because what's being expressed in music does actually happen to be something that's tied intimately to our biology (incredibly fine and refined expressions of emotional nuance), and the AI would have to model that too - i.e., AI that's capable of making music will have to wait until AI that's capable of emotion generally, comes online. Needs AI with juices (something to model the hormonal bath and how it affects the total state of the organism qua signaller-of-internal-states).