Mike Vizard: Hello and welcome to the latest edition of the Techstrong.ai video podcast series. I’m your host, Mike Vizard. Today we’re talking with Sefi Carmel, who’s CEO for SphereTrax, they help people find ways to work with music using a little help from AI, and we’re going to be talking about how AI makes things more accessible to folks and will hopefully improve the level of diversity we have in the IT world because more people can participate. Hey Sefi, welcome to the show.
Sefi Carmel: Hey, Mike, how you doing? Thanks for having me.
Mike Vizard: There is, of course, a lot of fear and loathing of AI being generated everywhere we look, and that’s all probably got some legitimate concerns, but are we missing the larger point, do you think? Is there more to be gained than to be lost?
Sefi Carmel: Well, AI is here, it cannot be stopped. There are the daunting aspects of it that we all know, and Skynet is not yet knocking on the door, so the robots aren’t yet here to kill us, but there are positive effects as well. At SphereTrax, we make the distinction between generative AI and assistive AI. We use AI in the search engine of our platform, SphereTrax, to help find music tracks for your video or your trailer or commercial. And when you’re currently going onto a music licensing platform, you’ll experience an experience that is very list-based, very text-based, so you’ll see a long list of tracks that are offered for licensing, and we found that experience very tedious. So we thought that with the help of AI and some innovative approaches, we can make that search experience for the creatives looking for music for their product a lot quicker. So you can type in, for instance, a brief like ‘find me a track for an orange juice commercial’ and it’ll come up with something uplifting and happy and fun, or you can type ‘find me a track for my documentary about the Holocaust’ and it’ll come with something very dark and in a minor key with an Eastern European feel and strength.
Mike Vizard: If that is the case then, do you think that there’s a lot of folks, at least in my experience, that they may not have the greatest skills level in a particular area, whether it’s writing music or whatever it may happen to be, but they have an idea? So is this going to open up an opportunity for a lot of people who have an idea to execute on it because previously they would’ve just been too limited by their lack of tools?
Sefi Carmel: Absolutely. I think AI in general is still going to be executing, even generative AI is still going to be executing ideas generated by humans. If a human generates a brief and says, “This is what I need, help me find it,” then there is nothing in that that is taking away work from anyone else, especially since if we, again, look at the fact that, in SphereTrax’s case, the AI is finding tracks in a database of tracks where all the tracks were composed by humans. If we’re talking about generative music, generative AI generating music, and happily we’re not there yet, I don’t think that the music that AI is generating currently is on par with anything that humans are generating and can be used. But when we do reach that stage, there will be an issue there for composers, definitely.
Recently, a company called Waves launched an online AI based mastering service, and my friends that are mastering engineers are debating that this would AI out the profession of mastering engineers. On the other hand, you can say that it makes mastering more accessible to musicians that need their track to be mastered so they can just upload their track onto that website and get the track back mastered. There is that balance. Any way you look at it, AI is here to stay and us in the creative agencies, and humans in general, I think need to embrace the fact that it’s here, learn how it helps you, what are the benefits that you personally can gain from it? And how to work with it, and what shortcuts it offers you in your process. And people that won’t embrace AI will be replaced, not only by AI, but by people who will embrace AI and learn how to work with it.
Mike Vizard: We have seen some contention of late, there was strikes involving writers and lawsuits that are going on related to artwork, and authors are filing lawsuits about their copyrights being infringed. How do we navigate that ownership of content in an age of generative AI where a lot of content is just being, shall we say, hoovered up into a large language model with little regard for ownership?
Sefi Carmel: This is very true, and this is if an AI is trained using the works of humans, those humans should be remunerated for it. I asked Ian Penman, our lawyer that is a very prominent IP lawyer here in London, what his thoughts were on the matter, and he said, “AI, lots of work for lawyers, lots of work for lawyers,” which it is because it’s not a simple matter. But if you have an AI taking 50,000 songs and learning from that how to create stuff, how to generate various pieces of music, personally, I think, and I’m not a lawyer, but as a creator, I think that anyone that participated in the creation of those 50,000 pieces of music should be remunerated, at least in part, for anything that AI produces thereafter. And I think that maybe this time the deals will be negotiated in a better way than they were when, for instance, social media was launched and started monetizing music, or streaming services, where people get 0.00001 cent per time the track is streamed, I think better deals can be negotiated for that for the creatives, which the only way to ensure that is strong lobbies of the creatives to ensure that their interests are protected.
Mike Vizard: We also saw at least one case in the US where a judge ruled that the music or the art or whatever it is created by a machine cannot be copyrighted even though it was a human that asked them to copyright that. Is that reasonable or where is that fine line between what a machine [inaudible 00:08:28]?
Sefi Carmel: I do not think that’s reasonable. I’ve been using computers with my work as a composer and sound designer for the past nearly four decades, one could say that if you… When I type and I type full of spelling mistakes, the computer fix and correcting my spelling mistakes, that’s not the definition of AI, but it is the machine enhancing what I do, so should that not be possible to copyright because there is machine intervention in that? When I’m composing on my sequencer and getting the machine to correct my timing, is that intervention in the creative process that would make it impossible for me to copyright it? Of course not. So progressing from that in a linear fashion. I think if you prompt an AI to generate something that is a brainchild of a vision that you had and the AI helps you execute it, it’s a tool. It’s a tool like any other of the many pieces of technology that I have in my studio and have been working with, I definitely don’t agree with that. I think that I understand where that comes from because copyright needs to be attributed to a human and it can’t be attributed to a machine. But as the plaintiff in this case, as you said, prompted the machine to create the content using AI, I think that should be attributed to that person.
Mike Vizard: We talked a little bit about diversity at the start of this. What is your sense of the overall impact that AI will have on diversity? I think it’s fair to say that we all know that educational systems are uneven around the world, so that seems to be always at the root cause of a lot of the diversity issues, so connect the dots for me.
Sefi Carmel: Well, in the application that we are using it at SphereTrax, AI definitely has a positive impact in terms of diversity because one of the problems that musicians have getting their work monetized and licensed and synced to film and TV is discoverability. If they come from a distant part of the world, how do they get noticed? How do they get their music seen and how do they get their music synced, and therefore the remuneration that follows that? And one of the things that we’ve done with our AI model, at SphereTrax, so far we’ve had 48,000 musicians from all over the world submit music to the platform, and we find that in itself a very equalizing process, where we’ve had musicians from far afield, of course we’ve had musicians from North America and Europe and where you’d expect, but we’ve also had musicians from Mongolia and the Fiji and Virgin Islands. And when you’re searching for music, the AI will find music based on the musical parameters and the use that you’re looking for, it doesn’t care if the musician came from Mongolia or Fiji or Lebanon or New York or London, and it is a great equalizer in that sense and it is a contributor for diversity, equalizing people’s chances of getting monetized.
Mike Vizard: How do you think that will play out through the way we have music distribution these days? Because we have centralized services, whether it’s Spotify or something else, and that becomes something of a choke point maybe for distribution, depending on your point of view, so is AI just going to open everything up more?
Sefi Carmel: I think so. I think that assistive AI that searches and finds music that is agnostic to the origin of the music in terms of location on the globe and in terms of the ethnicity and cultural background of the creators is going to be a great supporter of diversity in this case.
Mike Vizard: So what’s your best advice to your fellow engineers in the sound space and, for that matter, engineers anywhere, and all the creative folks who create content as we go forward, is there some sort of relationship that needs to develop or evolve?
Sefi Carmel: Well, I would say that AI is a strong tool and will become a stronger tool. Learn it, befriend it. Keep your friends close, your enemies closer. So learn what the AI does, be the guy that knows how to use AI and is not afraid of it, rather than the guy who’s afraid of it and is fighting it. I still think that creatives should understand their rights, should understand when they’re signing contracts, read them with a fine tooth comb, get good legal advice, and see if you’re giving away any kind of rights in perpetuity. For instance, for if it’s a singer, for your voice to be cloned and replicated and used ad infinitum in AI uses, that’s something I wouldn’t sign, unless they pay you a bunch of money for that. So it’s be savvy. I think be savvy, this is always good advice. Be savvy of what your rights are, be savvy of what you’re signing. I think hence forth, anything you sign, be aware of, are you giving away any rights in perpetuity of something that you’re doing being cloned without you having residual rights in it? So be savvy legally. Be savvy technically. Learn the tools, make them your friend, make them a tool in your belt, in your tool belt, and bolt your doors when the robots come.
Mike Vizard: All right, folks, you heard it here. On the one hand, we have nothing to fear but fear itself as usual. On the other hand, the devil is in the fine print somewhere. Hey, Sefi, thanks for being on the show.
Sefi Carmel: Pleasure.
Mike Vizard: All right.
Sefi Carmel: Thanks a lot.
Mike Vizard: And thank you all for watching the latest episode of the Techstrong.ai video series. You can find this episode and others on our website, we invite you to check them all out. Until then, we’ll see you next time.