AI Alert Questions

[BradGray]

Hey Passengers!

In looking at many of the taxi listings I've noticed the following:

**AI ALERT:** This company does not accept or sign music that has been lyrically or melodically composed using generative AI, nor does it accept backing tracks, full instrumental tracks, or lead vocal performances created with AI.

Does this mean that no music created with AI, in any respect? I.E. Using it for ideas, and adjusting, modifying, creating from there, or just not using a finished AI generated track.

This is not suggesting that I'm doing this, but I was curious if its come about because of the recent lawsuits from the big labels against Suno and others. That said, the labels are also working on creating their own AI models, and with complete ownership of the works, this will be a new frontier for music.

To summarize my questions:
1) Is there any capacity where AI is acceptable in the creation process for listings?
2) How are people, other Taxi passengers, using these tools today as part of their creative process? If so, what tools, in what way - assuming a very limited extent?
3) Other thoughts on all of this?

Overall, this comes from curiosity more than I want to go create a bunch of AI tracks. In the tech world, which is where I spend my days professionally, AI has really enhanced what we can do on the technical front, and provides a faster path for many things, and lets you "stretch out" your abilities to other areas outside of your expertise. So its become a great augmentation, versus a skill replacement. For music, and even for the Film and TV industry, it appears to represent a potential "extinction level event" as the AI can generate the content faster than any human writing or production team could; and its only the beginning. While it's not 100% there yet, it will be soon enough - hence the lawsuits. But the lawsuits (in my opinion) are really a power-play of sorts, as many famous artists have sold their entire catalogue to various companies, who (by my speculation) own their work and soon those companies will have their own AI model where they can generate unlimited content based on what they own.

Anyway, sorry for the lengthy post. Its a bit rambling, but not generated by AI.

Happy Sunday!
B

[AlanHall]

Good questions! There has been some discussion in a few recent Taxi TV episodes, but I don't think ML has published a canon of what AI can be used for and what must be avoided. I think we are to stay away from AI-generated FINAL tracks, whether vocal or instrumental.

Also up until the pushback against AI, manufacturers used the term "AI" in their ad copy the same way the words "lite" "natural" etc. were plastered all over food products for the last few decades - mostly without any real hard science to back up the claims. AI has been used as a buzzword for long enough that it may not mean anything when used in reference to a tool.

The real generative AI products are rare enough that they stand apart. It would be nice to see a history of AI lyric- and music- creation programs; both to see what to stay away from, and also to see where the output is substandard and where the output is good enough to be a true threat.

[cosmicdolphin]

I think the AI Alert is fairly self explainatory.

If you're using it to generate any of the words or any of the musical parts to your track then they don't want it. Seems pretty black and white to me. There's an inherent risk because some of these models have been trained on copyrighted works without permission, and are likely to recreate fragments of these works in their output and infringing. No library wants to be dealing with that.

I use chat GPT for researching and helping come up with cue title suggestions, I've had a quick mess with Udio & Suno but once the novelty wore off I don't think it would be much genuine use for sync instrumental cues that I make. I got into the industry because I enjoy working with my instuments and being creative - Not to be lazy and have an algorithm do the fun part for me. What's the point.

Mark

[BradGray]

Good questions! There has been some discussion in a few recent Taxi TV episodes, but I don't think ML has published a canon of what AI can be used for and what must be avoided. I think we are to stay away from AI-generated FINAL tracks, whether vocal or instrumental.

Also up until the pushback against AI, manufacturers used the term "AI" in their ad copy the same way the words "lite" "natural" etc. were plastered all over food products for the last few decades - mostly without any real hard science to back up the claims. AI has been used as a buzzword for long enough that it may not mean anything when used in reference to a tool.

The real generative AI products are rare enough that they stand apart. It would be nice to see a history of AI lyric- and music- creation programs; both to see what to stay away from, and also to see where the output is substandard and where the output is good enough to be a true threat.

Thanks Alan, I did a search, but will look again for those AI inspired episodes. I do agree that like anything new, it can quickly be used as a buzzword as its been incorporated in many tech products for several years already; not in the generative sense though. As for the history of these programs, were currently living it and its a new frontier in many ways. The power of these tools will likely follow Moore's Law, which he stated that "transistors on an integrated circuit will double every two years with minimal rise in cost." This is typically, and "loosely" applied to all tech as well, as it too tends to be true; generally speaking. The AI models of today are only the beginning, and what is blatantly obvious AI creation today, will likely be far more difficult to determine any difference in the future. Especially to the end consumer.

Fascinating times

[BradGray]

I think the AI Alert is fairly self explainatory.

If you're using it to generate any of the words or any of the musical parts to your track then they don't want it. Seems pretty black and white to me. There's an inherent risk because some of these models have been trained on copyrighted works without permission, and are likely to recreate fragments of these works in their output and infringing. No library wants to be dealing with that.

I use chat GPT for researching and helping come up with cue title suggestions, I've had a quick mess with Udio & Suno but once the novelty wore off I don't think it would be much genuine use for sync instrumental cues that I make. I got into the industry because I enjoy working with my instuments and being creative - Not to be lazy and have an algorithm do the fun part for me. What's the point.

Mark

Thanks Mark,

Yes my assumption is what you noted above, but curious as to if it was that binary or not. The reality is that we all learned from copyrighted works, without permission as well, and what we create tends to come from that amalgam of knowledge and experience of music, in both composition and performance. Trial and error. Learning, evolving, and slowly getting better. Learning Models (AI) also uses this process, but obviously faster and at a grater speed and scale than what we could do individually. I'm not arguing for "the machine" but really its doing the same thing that we all did - learning patterns, putting together what works and understanding what doesn't, and so on...based on feedback.

I too have messed around with some of these tools, to a limited extent, more for initial inspiration than anything. My personal perspective is that at some point it can be leveraged as another tool in the toolkit, versus something to create music on our behalf. I agree there wouldn't be much point of buying instruments and equipment to do all the recording, if you can just put in a few prompts on your behalf to do it for you. Its a creative art, which is a wonderful outlet for so many of us.

Cheers,
Brad

[cosmicdolphin]

Yes my assumption is what you noted above, but curious as to if it was that binary or not.

From what libraries tell me, yes it's that binary. They don't want any grey areas so it would be wise not to use it at this point.

The reality is that we all learned from copyrighted works, without permission as well, and what we create tends to come from that amalgam of knowledge and experience of music, in both composition and performance. Trial and error. Learning, evolving, and slowly getting better. Learning Models (AI) also uses this process, but obviously faster and at a grater speed and scale than what we could do individually.

I see this mentioned a lot ( usually by people who think Suno & Udio will win the lawsuit from the Record Companies ), but human learning and AI learning are not comparable. Human brains learn through experience, emotion, and sensory input, allowing us to adapt and understand context deeply. We bring all that uniqueness to bear on our creative expression which is why everyone's music is different.

AI, on the other hand, crunches numbers and patterns but misses out on the real human stuff like intuition and context. It's just brute force paint by numbers type stuff - it doesn't ' learn ' per se. It just knows what comes next as part of the typical pattern. It can't do originality, just remixing what it knows. Personally I think the AI companies will lose if it actually gets to court, and we are heading towards an era where you can't train AI's on copyrighted materials at scale without licensing it.

Mark

[Casey H]

Obviously nothing that's generated directly by AI such as any music or vocal track. For lyrics, using a tool like Chat GPT for general ideas as to a direction to go or even a word or common phrase you haven't thought of is fine. Sometimes I'll go there to help me think of what my storyline could be to fit my title. I won't use their lyrics but it's always great to stimulate ideas.

BTW, a lot of people are using tools where they sing a vocal and AI changes it such as from male vocal to female vocal. AFAIK, this will not fly with libraries that say no AI. I've heard 1-2 people say, "Well they'll never know". Not cool. Always be honest, treating people as you want them to treat you. And risking destroying a relationship with a library is definitely not worth it.

Casey