Podcast Summary
From Music to Physics to AI: Mikey Schulman's Unconventional Career Path: Mikey Schulman's journey from a music lover to a physics PhD, machine learning opportunities, and eventually to AI and music generation demonstrates that unconventional paths can lead to exciting and innovative careers.
Mikey Schulman, the co-founder and CEO of Suno, has a unique background that led him from a love of music to a career in physics, and eventually to the field of AI and music generation. Mikey's passion for music started at a young age, but he soon realized that he wasn't good enough to make a career out of it. Instead, he pursued a PhD in physics, where he discovered a love for quantum computing. However, he soon found himself at a local company called Kentro, where he stumbled into machine learning opportunities. With a background in physics and a newfound passion for machine learning, Mikey built a team and created fun products. Kentro was acquired by S&P Global in 2018, and Mikey continued to pursue interesting projects. He eventually found his way into AI and music generation with Suno, which allows users to create songs with just a text prompt. Despite an unconventional path, Mikey's love for music, physics, and AI have come together in an exciting and innovative way. Suno is a young company, but it's already making waves in the AI music industry. Mikey's story is a reminder that sometimes the most unexpected paths can lead to the most rewarding careers.
From speech transcription to music generation: The Bark team started with speech transcription but found unique challenges in music generation, focusing on tokenization as a primary area of innovation, and acknowledged limitations while looking forward to scaling transformer models.
The founders of Bark, an open-source music generation model, initially started with an open-source project focused on transcribing earnings calls for S&P Global after their acquisition. Although they were musicians, they didn't initially plan to focus on music generation. However, they discovered a lack of creativity in speech transcription and became captivated by the potential in music. The team, which includes individuals with backgrounds in text and transformer models, found that music generation posed unique challenges due to the continuous nature of audio data. They focused on tokenization, converting the audio signal into manageable units, as a primary area of innovation. The team's approach to measuring model quality is not explicitly stated but can be inferred as a combination of human evaluation and potentially other metrics. Additionally, the team acknowledged the limitations of current models and their reliance on open-source text community advancements. They also mentioned the potential for scaling transformer models in music generation, although the specifics of how this will be achieved were not discussed. Overall, the team's journey from text processing to music generation highlights the potential for innovation in the field of audio AI, particularly in the area of music generation, and the importance of understanding the unique challenges posed by the continuous nature of audio data.
Balancing metrics and human evaluation in AI music development: Importance of using human ears to evaluate AI music models, recognizing limitations of benchmarks in audio domain, and balancing quantitative metrics with qualitative human evaluation
While metrics are important in AI development, including in the field of music generation, aesthetics and human evaluation cannot be overlooked. The speaker emphasized the importance of using human ears to evaluate models, acknowledging that benchmarks can be less effective in the audio domain. They also shared that the music background of the team has influenced the development of Suno, particularly in the early stages. The team has tried to avoid implicit bias in their model, but they may need to reconsider this approach as they explore the unique challenges of AI music. The speaker admitted that they have not given much thought to the specific difficulties the model faces in music generation, focusing more on easily measurable aspects like stereo and bit rate. Overall, the team recognizes the importance of balancing quantitative metrics with qualitative human evaluation in the development of AI music.
Exploring music creation through AI: The company aims to make music creation accessible to all, experimenting with various business models to understand user motivation and sustainability, and ultimately changing how the world interacts with music.
The creation and experience of music through AI is a deeply emotional and subjective process that is not yet fully understood, and it caters to a wide range of human emotions, cultural backgrounds, and age groups. The company discussed in the conversation aims to make music creation accessible to everyone, not just professionals or hobbyists, and is currently exploring various business models to understand what motivates users to pay for the product. The history of digital business models shows that what works in the short term may not necessarily be the most sustainable solution in the long term. The company is trying to pioneer new behaviors around music creation and therefore, it's important to experiment with different pricing structures to understand what resonates with the user base. The ultimate goal is to change how the world interacts with music and open up new experiences for people.
Revolutionizing Music Creation and Collaboration: Suno's technology enables unexpected collaborations and creativity in music, bringing people together and providing joy in the music creation process.
Suno's technology is revolutionizing the way music is created and shared by enabling collaboration and creativity in unexpected ways. It's not just about the final product but also about the journey and the joy of creating music together. This is reminiscent of the way people have always resonated with music and the desire to make it with others. The technology has opened up a magical experience for users, allowing them to collaborate with AI or each other in various ways, such as co-writing lyrics or trading off verses and choruses. It's fulfilling for Suno to see how this technology brings people together and brings them joy, even if it may not be curing cancer, but it's a significant step forward in the world of music creation.
Suno's potential to change the creator-listener ratio in music: Suno empowers creators to share niche music, learn new genres, and feel ownership, potentially skewing the creation-to-consumption ratio
Suno, a creation platform, has the potential to significantly change the skewed ratio of creators to listeners in the music industry. The platform opens up opportunities for smaller, niche micro-sharing, allowing for the creation of songs that resonate with a specific group of people. This dynamic is currently absent in music. Additionally, Suno provides a ground-up learning experience, allowing users to discover new genres and even create hybrid genres. The platform's simple features, such as editing song titles, have led to unexpected user behavior, further demonstrating the desire for creators to feel ownership and pride in their work. The enjoyment of the creation process on Suno could potentially skew the creation-to-consumption ratio even further, making it a unique and exciting space in the music industry.
The future of music: blending creation and consumption: The future of music will see a blurred line between creation and consumption, leading to increased engagement and participation from a larger population, resulting in a faster-paced music culture with a strong emotional connection between fans and artists.
The future of music consumption and creation will blur together, leading to increased engagement and participation from a larger population. This shift, driven by accessible technology, will result in a faster-paced music culture where new styles and trends emerge more frequently. Despite this, the emotional connection between fans and their favorite artists is expected to remain strong. The advent of digital tools like DAWs (Digital Audio Workstations) in the past has already revolutionized music production, allowing more people to create music and contributing to a more diverse musical landscape. This trend is likely to continue, making music more accessible and interesting for both creators and listeners alike.
Revolutionizing Music Creation with AI Technology: Suno uses AI technology to generate unique sounds, unlock new song structures, and create melodically new music, making music creation more accessible and easier for all.
Suno is revolutionizing music creation by making it more accessible and easier through AI technology. This technology not only generates unique sounds but also unlocks new song structures, chord changes, and the ability to mix different styles. It has the potential to create melodically new music that could keep listeners engaged for longer periods of time. Suno is growing and hiring new team members who share a passion for technology and music. The company's synthetic songs, which include machine-created vocals and music, showcase the incredible capabilities of this technology. The machine doesn't even recognize the concept of voice, yet it produces sounds that resonate with humans. Suno's mission is to bring more music to the world, and they are always looking for talented individuals to join their team. If you're interested, check out their website for career opportunities.