Am I just an "angry optimist"?
I recently watched an incredible video from the CEO of Patreon, Jack Conte. I knew him as a musician before I knew of his connection to Patreon and his bands and the musicians he has on his channels are all incredible talents.
Jack's Talk
My Takeaways
I work in software engineering and technology and I am a musician myself. The implications of the speed that we are introducing new models and new capabilities is both exhilarating and terrifying for me as a software engineer. I will admit that I do feel some of the same hesitations about generative AI with regards to making music, but I find Jack's take about human creativity to be more inspiring than anything.
The Software Side
I finished my MS in Computer Science with a specialization in Artificial Intelligence, and the speed that these models are capable of solving coding-based problems are incredible. I am senior enough in my career that I recognize that my judgement and rounded experience is what I am paid for, but it still does trigger that existential dread knowing that my coding skills will never be able to keep up with these models.
I have been shifting my personal projects deeper and deeper into AI by working on local models and common sense guardrails for my own use. My time at work is a very pragmatic approach to AI adoption with a focus on safety and security. Just a few months ago (at most a year ago) I would have said that these models and coding tools are really just shifting our focus as engineers further along in the Software Development Life Cycle (SDLC). If you are already past the architecture portion of the life cycle, you used to focus on the implementation. Writing clean, well documented, and maintainable code were the hallmarks of great engineers. Now we must double down on our code review cycle to ensure the same standards are upheld, but the volume of code we have to review will continue to skyrocket.
In my day-to-day work, I have been documenting and teaching my team how to work with these AI tools through the lens of opportunistic productivity gains. We are writing up skill, prompt, and instruction files for our AI agents that are also serving dual purpose for human engineers. We can rapidly get up to speed in a codebase that is not our normal working directory through these files as well as leaving your agents in ask mode and going crazy with all the questions you can think of. We have been tightly scoping our agents to portions of the codebase that have well maintained and clearly articulated patterns and documentation, and it is a judgement call of the human engineer where there is too much code rot to let an agent loose.
The Music Side
I know there are examples of really good (scary good) generated songs. Some of these have been playing music in a different style, some are generating lyrics out of absurdity (for the humor), and some of these are just tonal background noise generated songs. I have never been a song writer myself, but I have had horrendous experience with non-specialized multi-modal LLMs trying to create songs. I once even asked ChatGPT what the next chord in a given progression should be starting from the chord names and guitar tablature... It. Did not. Go well.
Here in this topic of music creation vs generation (I am not convinced that we should consider a statistical prediction of existing patterns to be "creation"), I found Jack's talk particularly inspiring. I was aware of many of the music history milestones listed in the video and dealt with my own fears when I was studying trombone performance all those years ago. The analogies made with film and video production gave me a whole new appreciation of that side of the business, but I will admit I still struggle with the "social media" influence on our society (although I love that Jack does not like "content creators").
Closing Thoughts
It can be paralyzing to think about keeping up with technological change, but separating out these leaps into a "new medium" or "more powerful tools" gives me hope that musicians, creators, engineers, and humans in general can still survive. I want to be more optimistic about the future AI can bring, but boy does it feel validating that anger is an acceptable feeling too.