AI music models must learn inductively without hardcoded music theory
Mikey Shulman · No Priors Ep. 64 | With Suno CEO and Co-Founder Mikey Shulman · May 16, 2024 · at 9:35
Mikey Shulman, CEO of Suno, explains the modeling philosophy behind Suno's music generation architecture.
0:00 / 0:19exact quote · 19.7s
▶ Watch the full episode on YouTube → 720p mp4 · rendered on demand · StarZero watermark
“The model shouldn't know about music theory. You don't tell GPT, this is a noun and this is a verb. GPT figures it out. If I tell my model, there are only 12 tones. My model will only know how to output 12 tones. If I tell my model there's 50 different instruments, I will never get that unique sound. And so we've really tried very hard not to do anything like that.”
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 Mikey Shulman
Suno aims to be a mass consumer product, not a DAW plugin
“We are trying to change how the entire globe interacts with music and to open new experiences for people. And so what that means is that this is a consumer product. This is not sprinkling AI into Ableton or Logic or Pro Tools. This isn't for the person already…”
Generative AI SaaS pricing models are likely outdated vestiges
“Everybody's doing kind of something that looks like SAS pricing, and it's kind of done very crudely, and we are certainly no exception to that. But I don't know if this is right in the long term, and it strikes me as probably just a vestige of, it is the same …”
Generative AI will blur the boundary between music creation and consumption
“And I think these set of technologies have the ability to skew that much, much farther because the creation process is so enjoyable. But I actually think if we do this right in the future, these are not going to be the Terms that we use to describe what we're …”
Generative AI will dramatically increase time and money spent on music
“If we are correct that there are just modes of experience around music that people don't have access to, that we can get a billion people much more engaged with music than they are now, that just in terms of the number of dollars or the amount of time people a…”
Suno builds on transformers and focuses innovation on audio tokenization
“We don't make it a secret that these are just transformers. This is somewhat our backgrounds doing text before, but also transformers scale nicely. A lot of work ends up being done for you by the open source text community, which is always really nice. We can …”
AI evaluation benchmarks fail for audio, necessitating human aesthetic judgment
“I think in, in all branches of AI, we become slaves to our metrics, and you say, I did this accuracy on this benchmark, and this accuracy on this benchmark, and in the real world, sometimes it doesn't necessarily matter, and these benchmarks are extra terrible…”