Cole Ingraham: Build generative AI with deterministic components, not pure randomness
Cole Ingraham · AI-Powered Music // Drew Silverstein and Cole Ingraham, Amper Music (FirstMark's Data Driven) · Apr 6, 2017 · at 9:57
Dr. Cole Ingraham, Head of Composition at Amper Music, explains software architecture principles for creating controllable and debuggable generative music systems.
0:00 / 0:23exact quote · 23.8s
▶ Watch the full episode on YouTube → 720p mp4 · rendered on demand · StarZero watermark
“I find it's much better and easier to focus on writing writing systems which are deterministic in their individual components and putting them together in a way that it's complex and seems non-deterministic, but that way you can actually reason about what it does. Versus relying on some sort of totally pseudo-random process that you can't actually control with certainty of what it will do.”
quote is from the automated transcript, cleaned for reading: filler sounds and stutters are removed, nothing is rephrased. names can be misheard (the analysis reads context, assessments check outside sources). how →
More from Cole Ingraham
Cole Ingraham: Generative AI relying solely on datasets produces homogeneous outputs
“The result of that that you're gonna get is usually things are somewhat homogenous. They will look like your data set. You're not going to get things that are not in the data set because you got everything from the data set.”