SNDOUT (sound out) is a generative AI music discovery tool that gives you the power to wander the globe and explore music history.

The intersection of music, discovery and storytelling has been a personal obsession of mine since the early 90s. Multimedia tools like Hypercard and Director unlocked the idea of storytelling as an interactive journey — the ultimate choose your own adventure. Over the years, I've tried to build experiences, imagine new paradigms, and advise companies that would capture the joy of music discovery. Limited content and resources have always been a barrier.

The rapid advancements in generative AI, powered by transformer architectures and large language models (LLMs), have been impressive, yet I approach these developments with some caution. While the leap in capabilities is undeniable, it remains to be seen if AI can transcend mere imitation of human intellect. Nevertheless, I’m enamored by the profound capabilities these innovations unlock.

What excites me the most is large language models as a reflection of the human experience — an encapsulation of global collective knowledge and experience. There are many flaws and challenges with AI, but one thing it’s great at is connecting dots. When interrogating the AI service to find paths that connect, and then asking it to articulate and visualize those connections, we witness the true beauty of LLMs emerging. And since these connections simply result from probability, the AI service can surface weak ties, sometimes as hallucinations, as probable stories and insights from history. This is the closest I’ve felt to AI as being a creative contributor.

Incorporating AI into products and startups presents significant challenges. The cost of AI services, such as generating images with DALL-E 3 or processing complex GPT-4 prompts, can quickly accumulate, making the simple act of sharing an idea a potential financial burden. To manage these expenses for SNDOUT, I’ve devised a token-based system aligned with AI service costs. Concerns around content rights, bias, and the diminishing benefits for musicians, among other issues, pose significant challenges. Addressing these fundamental concerns is crucial for the sustainable future of AI services.

This project is a prototype — it’s the best I could do with my limited time. Additionally, Spotify has not sanctioned it, so I suspect they will shut it down at some point. The core intent for the experience is to connect context between two points (time, place, artist, genre). Features like the “random” button, which generate unexpected musical discoveries, and the ability to explore music based on your current location, are surprising highlights of the experience. I had a blast building this tool. Enjoy!