Rahim Rahemtulla: Hello and welcome to Silicon Valley Innovation Center. Here at SVIC, we promote discussion on technology and business innovation through our Executive Immersion Programs and online events. I’m Rahim Rahemtulla, SVIC Brand Ambassador. Today I’m talking to Dr. Maya Ackerman about Artificial Intelligence and Irreplaceable Human Characteristics. Don’t forget, if you have a question for Dr. Ackerman, you can send that to us in real time. Just use the comments section of whichever platform you’re watching us through. So our guest today is the CEO and Co-founder at Wave AI. Wave AI are the company behind ALYSIA, an artificial intelligence platform which can help in the songwriting process. So Dr. Ackerman is with us now to say more. Dr. Ackerman, thank you so much for being with us today.
Maya Ackerman: It’s a pleasure. Thanks for having me.
Rahim Rahemtulla: So, ALYSIA, it seems like with that you’ve set out to solve a very interesting problem. AI, I think, when we think about that, a lot of us would just assume that’s for solving repetitive tasks, things that we would probably like a machine to take care of and that we wouldn’t have to deal with ourselves, but ALYSIA’s been made to write songs. So, it sounds like you’re sort of challenging these notions of what AI can do through this platform.
Maya Ackerman: That’s a really excellent question. Yeah, indeed, we’re taught to believe that machines are here to do the boring work for us – stuff that’s very repetitive and dull – and that we are the creative agents. But when you look at the history of creativity, you discover that the definition has actually been adjusted in parallel to the developments in the Industrial Revolution and with the emergence of technology. So, for example, if you look at the earliest definitions of intelligence, they include things like speed and accuracy, whereas today’s definitions are a lot more about higher order problem-solving and creativity. So as soon as a machine is able to do something, we move the bar towards something that it still can’t do. So, in that light, it’s perhaps unsurprising that eventually we get to – as machines become more and more complicated – these notions of what makes us human and what makes our abilities unique inevitably ending up getting challenged, in a way. Now, I’m not here to claim that machines are as creative as us or even creative in the same way, but definitely their creative capacity is a lot broader and wider than sometimes given credit for. And ALYSIA is part of that, is part of computer creativity, part of computational creativity, but it actually more neatly falls under helping us be more creative. So, in that sense, it feels more integrated with our creativity.
Rahim Rahemtulla: And I was wondering, I read one of your research papers about ALYSIA and you have some songs available online as well which were made with the platform and I listened to those too and they’re all very good, especially “I’m Always Chasing Rainbows”, that was one which I particularly enjoyed, in case that helps. But I was wondering, when we listen to songs which have been crafted with the help of ALYSIA or some other AI platform, do we need to listen to them any differently? Should we be evaluating them differently? If we hear them on the radio set against a song that was written without help of AI, is this something that we should be aware of in any way?
Maya Ackerman: That’s very interesting. You know, in computational creativity, we often wonder, how should we be evaluating stuff created with a machine? Or perhaps even independently by a machine, because that’s also a line of research. And there is this thing called the Turing Test, where the evaluators are blind to whether it was made by a machine or by a human. And sometimes you get more authentic evaluation of the artefacts, if that information is occluded.
Another way to consider this question would be to look at how technology has been used in the creation of music for decades. So, perhaps the most prominent example are digital audio workstations like GarageBand. And nowadays it is very common to hear songs on the radio and, in fact, the default is to assume that sounds were produced on a computer using a digital audio workstation. And that’s no longer considered cheating, that’s no longer considered not human because it’s a tool that we’ve accepted and integrated. So I very much believe that the same trajectory is going to go for AI-based composition because the role that a human plays remains important and still allows us to give credit to the human creator.
Rahim Rahemtulla: Do you think the human role will be always the primary role in that relationship?
Maya Ackerman: I think we’re likely, as a society, to place emphasis on the human role whenever any sort of computer system is involved. Even if what the computer is doing is perhaps substantial, it is the humans expressing themselves that’s often interesting in a work of art.
Rahim Rahemtulla: We’ve seen in other parts of the world, especially coming out of Japan, we have a sort of voice-synthesizing software out there and that actually gives concerts. I mean, they have a hologram which is up on stage, lot of people – thousands, I suppose – come to these concerts, they love it. In that sense, I suppose we could argue that the human role there has really taken a backseat to the AI. It’s the AI, maybe, which is what people are coming to look at. Although humans ultimately are the ones behind it and creating the lyrics for those songs, so what do you think on that one?
Maya Ackerman: Humans are creating almost everything in these concerts. I love, by the way, Yamaha’s VOCALOIDS. They are fabulous. I was just playing with it a few days ago. I think it’s fantastic for a computer to be able to sing and it’s perhaps notable that me, as a singer myself – because that’s my artistic training in addition to all the computer science expertise – still, I really enjoy seeing how far does voice synthesis go. But, as far as the performances, it’s a hologram and it’s a synthesized voice, but everything else is human-made. Everything.
Rahim Rahemtulla: Yes.
Maya Ackerman: The composition, the production.
Rahim Rahemtulla: Yeah.
Maya Ackerman: In fact, VOCALOID allows you to control the automated voice down to the most minute details. So, it’s fascinating technology. I think it’s a fascinating social phenomenon that this is so popular, but you have human songwriters, you have human composers, human producers, in the background and it’s only the voice synthesis and the hologram that’s computer-made. And that’s almost always the case when it comes to anything computer-made – that it’s partially computer-made and then largely human-made.
Rahim Rahemtulla: Absolutely, yes. And what about all the performance aspects? As an opera singer yourself, I suppose this is perhaps particularly close to your heart. I mean, watching you perform on stage, for example, I’m sure is a very wonderful and joyful and emotional experience for those who are in the audience. But if there were a hologram which could sing just as well as you could, if we could reproduce your voice in a hologram, I think we would still rather see you perform, wouldn’t we? Because I feel like there’s something there, just in that human factor of seeing a person onstage channeling that emotion and the passion of a song, which I guess, I just feel a machine will never get there. A hologram will never do it.
Maya Ackerman: I mean, you’re right. We don’t have any way to authentically give computers emotions. The closest we’ve come is to simulate it a little bit, to let them say, “I feel happy because I read a happy article.” But that’s all completely simulated, not even realistically. And I think that even with the popularity of AI or computing in all aspects of our lives, there’s something that’s always going to remain that’s really special about human connections. I’m one of these old souls, if you will, who still loves meeting people in person because there’s a certain connection that you get having a person in the same room that’s irreplaceable for me and I think for a lot of other people. And that generalizes it, that for a lot of us, even if hologramic performances become very popular, for a lot of us the meaning behind a human performer will always be there and will always be special.
Rahim Rahemtulla: And so, if the human is always going to be primary, let’s assume, in these sorts of human-AI relationships and so ALYSIA is, for example, really just there as a tool, as a creative collaborator, as you put it, I’m just wondering, do you think that can also apply in other fields as well? Because with ALYSIA you don’t necessarily have to be musically-trained but you can still produce songs which could be played on the radio perfectly respectably, for example. So, I’m wondering, do you think that can apply in other fields? Is it too much of a stretch to say, for example, I don’t know a lot about cars but maybe one day there’s going to be an AI platform which can help me to fix a car?
Maya Ackerman: Yeah, yeah, that’s a very good question. If anything, the creative applications are more emerging and different. Computers have been helping us, especially in recent years, in so many different ways. For example, to admit another weakness, I have a terrible sense of direction and yet I have no problems navigating new areas right now because of Google Maps. So a lot of our weaknesses have been fixed through technology for a long time. Another kind of common weakness that a lot of us share is I wear contact lenses. Nobody even knows anymore because we’ve gotten so good at fixing human vision. So it’s a balance. At its best, technology helps us fill in weaknesses, become good at things that we wouldn’t otherwise be good at. Of course, it’s not the only side of technology; it’s complicated and its social impact can be complex and sort of wide-reaching. But I think that humanity is best to focus at the kind of technological developments that really makes us more powerful, make us more free, gives us more abilities.
Rahim Rahemtulla: I see. And speaking of the social impact, that’s obviously a topic which at the moment is very newsworthy and there’s a lot of talk – we hear in the public sphere about how AI is going to impact us as a society, on our work lives and so on – and I feel like with ALYSIA, what you’ve gone and done is something that, again, like I mentioned earlier, something which maybe people would not have expected that AI would be capable of doing. So, is this, in some way, something that we should be more worried about, AI? In the sense that more jobs than we think may be at risk from the rise of the machines?
Maya Ackerman: That’s a really good question and that’s a really, really important topic and a social topic, a topic that can connect to government policy. But first, let me comment on the fact that it’s unexpected because I couldn’t agree more. I’ve been studying machine learning, well, my whole adult life and it was only when this concept of a computer collaborator was presented to me that I even realized that it’s possible to create ALYSIA. I’d been struggling to write songs and at the same time I knew machine learning very well for a long time at this point and still it took a long time to make that connection because there is something just so surprising and unusual, to have a computer help you write music. So, first of all, I completely agree there.
Yes, so I suppose it does, in a way, suggest that there might be many other applications that we’re not thinking about, but I think there are always a lot of details with AI. Just because songwriting turned out to be possible doesn’t necessarily mean that a whole bunch of other stuff is possible because each problem – at least the way machine learning works right now, the way that the edge of AI is functioning – you really have to do it one problem at a time, and each problem is very deep and complicated. And so, I don’t think it necessarily means that instantly we’re going to have a whole bunch of solutions that we didn’t have previously because we’re nowhere near being able to create general AI.
That being said, I do think that, as a society, as individual AI creators, we do need to take into consideration how our creations and how AI in general is impacting the world. And, in fact, for example, AI researchers, for the most part, feel very strongly against AI-based weapons. So there is an awareness among researchers that it’s not necessarily always all positive, that we do need to think about how we apply our artificial intelligence. Now, that’s the case with everything. Chemical warfare, right? So almost anything that’s really powerful can be used in a negative light and there’s a certain amount of responsibility both for the researchers and for the society, and for government, to create the structures in which these really, really powerful tools are used in a way that benefit people and doesn’t hurt us. So, yeah.
Rahim Rahemtulla: It sounds as though there’s obviously a big discussion that still probably needs to be had then, where as a society we come together and we decide where is AI acceptable, where we are happy for it go, where we want to keep it under wraps, so to speak.
Maya Ackerman: Yeah, yeah, that’s right, that’s right. So, definitely when it comes to weapons, we don’t want to be applying AI to warfare just because we can. And I think that with certain forms of AI that impact the jobs of millions of people, it also makes sense to pause and think, Okay, so we have this amazing technology – let’s say that it can allow us to relocate goods from one location to another with minimal human participation – how do we want to bring this technology into the world? How do we do it in a way that has positive impact on all of us, as much as possible? Does that involve, perhaps, social programs, re-education? In what speed do we bring it on so that it doesn’t shutter too much in our society? So it’s not simple. There are no simple answers to any of this. But I really think it’s critical that we solve it as a society together – not as isolated pockets.
Rahim Rahemtulla: And with ALYSIA, I suppose you’re not worried, then, that it’s going to suddenly put a whole lot of music producers out of work. And the other side of that of course, as you’ve said it, is it opens up the world of musical creation to a whole audience and so that’s a major plus. And if there are any minuses, there’s a balance to be struck there.
Maya Ackerman: There is not a day that goes by that I don’t think about these questions. It’s very, very important for me that the technology which comes with ALYSIA, again, that’s it’s positive to as many people as possible. I believe it’s most similar to the bringing on of photography to everybody. It used to be a very rare specialization and now it’s commonplace. We have it in our phones! We have excellent cameras in our phones. It’s mind-blowing to me! This was definitely not the case as I was growing up. And so that’s how I view ALYSIA. I think overall we’re bringing in this massively positive, empowering, healing form of self-expression. We all know that music is healing and when I use my own system – which has been, of course, been getting better and better as we’ve been developing it – it’s amazing to me that right now I can create a song that really expresses myself in fifteen minutes. It’s like being able to see when you couldn’t before. It’s amazing. That being said, we also think about professionals. How do we make this a positive movement for professionals? So let me just note that producers are not impacted by it because it’s on the composition site.
Rahim Rahemtulla: Sure. Sure.
Maya Ackerman: But for songwriters – and we have been working with songwriters – we want to create a tool that’s specific to them, that’s a lot more nuanced, that has a lot more degrees of control. In that way, we’re helping them as well. We’re helping them make their job easier, faster. And I think of ALYSIA as one more composer, one more songwriter. And I think that human songwriters are so diverse and so different from each other that for a long time they are going to have so much to express that perhaps ALYSIA would just help them do it faster and better but couldn’t do it for them. ALYSIA is not an autonomous songwriter.
Rahim Rahemtulla: Sure, sure. There’s plenty of positives there. And I wonder what you say about, for example, the fact that everyone now has a smartphone and can take photos and it sounds like everyone is a photographer now. But I think that’s kind of these two things. On one side, the real professionals, the ones who are really good at it, become more valued. And, on the other side – and I wonder if this is the case with AI – people could just become very good, not necessarily with music, to talk about ALYSIA, but just very good at using the AI. And I wonder if going forward as well we’ll be looking at workplaces where we have more AI inside them. Could jobs just become something where you’re actually just very good at using the AI? You know how it works, you know how to get the best out of it, although you’re not actually necessarily very good at the job it’s doing. Do you see that as a scenario?
Maya Ackerman: I think it’s been happening to a very large degree. Animation has become all about using automated tools. Granted, not AI-heavy, at least they did not start off AI-heavy, as other disciplines. But animation used be by hand so it used to need a lot of stroke accuracy and the multitude of skills that go with being a painter and being able to draw very well and right now it’s all about using systems like Maya. Amazingly, it has the same name as me. So I think that it only stands to reason that a similar revolution might happen in other spaces. Essentially, letting computers take over what they’re genuinely amazing at and learning, us humans, to do what we’re best at. So, for example, humans are very, very good at telling whether they like something. All of us are great judges. We know if we like a song, we know if we like a musical phrase, we know if we like a picture – even if we can’t create it. And computers are much better at exploring the creative space. They can give you a whole bunch of variations that you would never think about. And then from those, you can make choices. That is what ALYSIA is taking advantage of. Yeah, so I completely agree with your prediction about our roles changing as technology is able to provide as with more, because I’ve been seeing a lot of evidence of that.
Rahim Rahemtulla: Excellent. And, Maya, we’re almost out of time, unfortunately. So just to close up our conversation, I’d like to ask, given all that we’ve said, whether we’re talking about somewhere in the workplace, maybe in a music studio or wherever it is – and I supposed you’ve answered this, really – but you do believe, then, that there are limits to artificial intelligence? And there will be always some human characteristics which, even if we could replace them by machines, we just won’t want to necessarily?
Maya Ackerman: Yeah. So even what computers are good at today, the more we learn about what they can do, it tends to be different from humans. There are exceptions, of course. They are just better at a multiplication, at crunching large numbers. But, for the most part, when it comes to more complicated tasks like creativity, like intelligence of a higher order, they are just different. And they’re not better, they’re not worse – they’re not directly comparable, which is why these interactions with computers make the most sense. So definitely for the foreseeable future, for the future as it were able to accurately predict, I believe that computers are not about to overtake us, but to really just become progressively more and more useful for us.
Rahim Rahemtulla: Fantastic. So it sounds like a good message: we shouldn’t be afraid of the coming-age of AI. We should embrace it, we should see it as partners in collaboration. Fantastic. Well, Dr. Ackerman, thank you so much for joining us today.
Maya Ackerman: A pleasure.
Rahim Rahemtulla: And to our viewers and listeners out there, many thanks for watching. If you’re interested in trying out ALYSIA for yourself, the app is coming out soon and testing is going to be starting in a few weeks. If you go to withalysia.com, you can sign up to be a part of that. And, don’t forget, if you’ve enjoyed my conversation today, make sure you join us on September 13th. Dr. Maya Ackerman is going to be joining us for a panel here at SVIC on the future of the workplace, shaping up to be a great event full of insights about what the world of work is going to look like in the future. You’ve already had a sneak preview today, so join us for more. Check our website siliconvalley.center for all the details. I hope you’ll be able to join us then and we’ll see you. For today, for now we’ll have to end it, but until next time. Good bye.