WS CONNECT Q&A with Jesse Engel

TL;DR

Research scientist Jesse Angle discusses the role of machine learning in music and art, highlighting the importance of having a problem that is personally interesting to ground learning.

Transcript

hi everyone I'm Laura with the World Science Festival thanks so much for joining our live Q&A we're here with Jesse angle hi Jesse right Jesse thanks so much for being with us today it's my pleasure can you tell our viewers a little bit about what you do and where you work what your research is sure so I am a research scientist at Google brain so I... Read More

Key Insights

• ❤️‍🩹 Machine learning and creativity intersect in projects like magenta, where algorithms are developed to augment human abilities for open-ended problems in art and music.
• 💄 Personal interest and finding problems that are personally meaningful help in making the learning process more engaging and intuitive.
• 🎼 AI can potentially generate commercial music, but human biases, context, and social interactions will always be essential in the music industry.
• 😫 Teaching algorithms involves setting up a search space of functions and using data to find the best possible function through optimization.
• ❓ AI becoming conscious is a distant concern compared to addressing more immediate ethical issues in human interactions.
• 😑 AI has the potential to expand the audio spectrum and create new artistic expressions beyond human comprehensibility.
• 💭 Interpretable analysis of data and structuring information can help in better understanding the thoughts and opinions of individuals and society as a whole.

Questions & Answers

Q: What is the goal of research in creative AI?

The goal is to create algorithms that enhance human abilities in expressing themselves and exploring new solutions to open-ended problems in art, music, and other creative fields. It involves finding new ways to think about AI and machine learning beyond traditional problem-solving approaches.

Q: Can AI create top-selling albums and replace human musicians?

While it's possible for AI algorithms to generate sounds that people would pay for, there will always be human biases, context, and the social interaction that musicians bring to their work. AI-generated music may have commercial applications, but it's unlikely to completely replace human musicians and the unique experiences they offer.

Q: How do you teach an algorithm?

Teaching an algorithm is metaphorical, as there is no actual teaching involved. It involves setting up a space of possible functions and using data to find the best function that can map inputs to desired outputs. Machine learning algorithms like TensorFlow help in refining and searching for the best possible function through optimization techniques.

Q: Do you have concerns about AI and robots becoming conscious?

While ethical concerns about AI and its impact on society are important, the complexity of consciousness is far from being achieved in current AI models. There are more pressing ethical concerns in human interactions that need to be addressed before such issues become relevant.

Summary & Key Takeaways

• Jesse Angle is a research scientist at Google Brain and works on the magenta project, which focuses on the intersection of machine learning and creativity in music.

• Machine learning is a powerful tool to find optimal solutions, but creativity and open-ended problems require new algorithms to augment human abilities in expressing themselves and finding new solutions.

• Jesse emphasizes the importance of finding problems in machine learning that are personally interesting, as it makes the learning process more intuitive and engaging.