Interview
Mikey Shulman, CEO @Suno: The Future of Music, What is Gonna Happen? | E1244
- A final distinct model version (4 or 5) is expected before updates shift exclusively to product releases, coinciding with a market transition where consumers stop paying premiums for newer models within approximately five years as capabilities integrate into standard offerings.
- Music generation will likely rely on many relatively small models rather than a single massive entity due to the subjective nature of the art form, with scaling laws unable to resolve all technical challenges and GPU burn expenses projected to increase significantly despite sublinear returns.
- The industry aims to surpass the current video game sector's value through interactive, engaging experiences similar to gaming, while the overall market size is predicted to grow rather than remain a fixed pie.
- New capabilities will be aligned to human preferences using "taste" techniques adapted from chat interfaces, with the goal of rendering specific model versions irrelevant to users who will focus instead on emotional output and interaction metrics.
- User acquisition is expected to grow alongside revenue without significant trade-offs, driven by the removal of "prompt" terminology within six to twelve months in favor of more intuitive interfaces, though a superior web app interface could have accelerated adoption beyond the current 400,000 Discord users.
- Discovery will likely remain algorithmic with TikTok dominating short-form and Spotify leading list-based discovery, while Spotify's video initiatives aim to transform background listening into more engaging consumption habits.
- Future business models are expected to reward remixing and fan engagement more than traditional stream-share structures, potentially enabling individual artists to create personalized models provided they possess the necessary rights.
- Timbaland is expected to continue advising and legitimizing the technology, while the company anticipates the RIAA lawsuit will not force a cessation of operations but would suffer negative consequences from a trial loss.
- The company does not plan to offer APIs to generative video companies to avoid positioning music as background content, though the music industry may eventually resemble a social blend of peer-to-peer interaction and professional content creation.
- Significant risks include the potential for rampant unauthorized artist impersonation and the creation of hyper-personalized, antisocial listening experiences that remove human connectivity.
- Operational challenges involve the difficulty of managing extensive remote work as the company scales, the extreme difficulty of maintaining focus by selecting only three out of 30 impactful initiatives, and the prediction that large model providers will likely prioritize application layers like customer support over the music domain.
- Future growth relies on more people creating music for more hours with economics flowing downstream, though the speaker expresses uncertainty regarding the potential for a broader "AI winter" and warns against misjudging effect sizes by focusing on second-order rather than first-order benefits.