Robintes wrote: ↑13 Apr 2023
Not to blow my own trumpet of prescience, but I already asked this AI question in General cat. here but not so forcefully. I just suggested that Music Producers were living on borrowed time. With all the new music-by-numbers goodies available now in DAWs (and R12 has so many ways to get into mischief) I predict that AI Soundscapes will be in your face by 2024 and you will be hard pressed to tell the product from a human composed version. AI can examine the sound market and guess what its looking for (and pays best) then it can.
hook into a DAW and create melodies and patches to match - cant you all just see it
coming. Fake avatars abound already. AI B movie plots, screenplays, Pulp fiction novels planned and ghosted. Even fake picassos. My OH must have an AI implant - a good faker
We already know what formulaic music looks like. That's cheesy pop. And while it dominated the charts in the 90s, it failed to innovate and bring us the songs that stand the test of time, break new ground, and relate to people in a new way.
Creating music isn't just a factor of intelligence and analysis.
There's also a highly subjective aspect to its creation that defies logic, but by sheer coincidence does not contradict it.
That's not to suggest it's 100% impossible to model something equivalent to subjective human experience (that may be easier than AI), it's that the subjective human experience more often than not has shit intuition.
And that's what genius really is.
Genius is the sheer luck of having an intuition that defies logic but does not contradict it. And so they can make huge intuitive leaps into the right spaces that would have taken decades to reach through mere exploration alone.
It's why neural networks were key in beating human beings at Go, and why they're so key in chess engines. Chess engine strength used to be largely determined by how many moves it could evaluate. With neural networks, they are able to search far fewer moves by utilizing the "intuition" of the neural net.