[INTRO] Unfair use? The ethics of generative AI for music
Introducing an 8-part series of posts that dive deep into the PEOPLE & PRODUCTS aspects of the ethics of generative AI for 'creating' music
TL;DR Just because we can generate melodies now with generative AI, that doesn’t mean we should turn these capabilities loose in the world (yet).
We need to get the data and AI ethics right. As with planting a tree, the best time was 20 years ago, and the second best time is now.
INTRODUCTION
This 8-part article series tackles one of the key questions about the use of generative AI for music: Is it ethical, for musicians (writers and performers) and for music users?
Prerequisites: No prior knowledge is required for how generative AI techniques use training data or how AI models work under the hood.
Audience: The expected audience for this article series includes executives, technical and product leaders, musicians, music publishers and copyright owners, music users and consumers, responsible AI teams, and legal advisors.
This article series is not a substitute for legal advice and is meant for general information only.
Here’s the planned “table of contents” for the article series. For convenience, links will be added here after each PART is published.
BACKGROUND
HOW & WHO (the deep dives on 5 bigs)
Part 4: Ethics of Top 5 - Training & Reference Data
Part 5: Ethics of Top 5 - Traceability & Fair Use
Part 6: Ethics of Top 5 - Bias Minimization and Diversity
Part 7: Ethics of Top 5 - Cost-Benefit Analysis
CONCLUSION
Part 8: Current State of Ethics
References (page)
In PARTS 1-3, we identify the PEOPLE affected by genAI music initiatives, identify key aspects of ETHICS of generative AI for music, and survey the landscape of recent PRODUCT announcements and events in use of generative AI for producing and customizing (editing/controlling) musical melodies.
PARTS 4-7 are deep dives which include close looks at where the major players (OpenAI, Google, Meta, Microsoft, and Adobe) and their PRODUCTS and PEOPLE stand today on:
Respecting IP and legal considerations on generating and customizing melodies by:
fair use of sets of human-created songs for training models and enabling controls
fair use of user-provided ‘reference’ music and melody styles
Minimizing potential for bias in use of Text To Audio (TTA) techniques for generating music
Supporting traceability of human-created sources to song outputs from the tools
Evaluating whether the benefits of these tools warrant the societal costs of using AI (including environmental impacts)
In PART 8, we will share our conclusions and some recommendations.
Look for PART 1 soon! (follow or subscribe to be notified when each of the articles in this series is published)
Part 2 of our series on ethics of generative AI for music is now also out: https://sixpeas.substack.com/p/part-2-unfair-use-genai-music-ethics It covers how the 3 stakeholder groups in Part 1 are affected by ethical concerns (including model biases), the “pro-choice” view of ownership rights, and the importance of "informed consent".
Part 3 is up next, on the who, what & when of the current genAI music tools. While we’re working on finalizing Part 3 and beyond, we’d love to hear from musicians, music lovers & consumers, technical folks, and people who are two or more of those! What do you think? Did we miss any important concerns that matter to you, or do you have a different view on some points?
Happy to announce that Part 1 is now published in our series of articles on ethics of generative AI for music! My collaborators and I are looking forward to hearing your thoughts :) https://open.substack.com/pub/sixpeas/p/part-1-unfair-use-ethics-genai-music?r=3ht54r
FYI, we're being agile about executing on our publication plan for the article series ;) We've evolved it a bit since this Introduction was published. Parts 2 and 3 are now swapped.
- Part 2, the next article in the series, will cover ethical risks and guidelines that impact all three stakeholder groups we identify in Part 1.
- Part 3 will cover the generative AI products being developed for music.