I’m currently studying AI and music for my final year project next year at uni (studying Creative Music Production). Had a pretty interesting talk with my professor the other day who seems to think it’s impossible for machines to be truly creative, especially in regard to music as there’s so many variables and because it’s so subjective. I think he’s getting confused with current ANI and what AGI could bring but even when I tried to explain the difference he seemed to still think it was impossible for machines to ever be truly creative. Could ANI even be enough to unlock creativity though? AlphaGo seemed to be onto something.
Cambridge dictionary meaning of creativity: the ability to produce original and unusual ideas, or to make something new or imaginative.
Seems like that’s what this move was to me, especially since brute force logic doesn’t work with Go. Although of course music is much harder still, it seems like we’re going in that direction much faster than thought to many. Go seemed impossible for at least another 100 years to some notable people in the late 90s for example.
Any thoughts on the topic would be very helpful and I’m very interested to hear other opinions in general in regard to AI and creativity.
This is very interesting topic. I am also music composer and I work in music industry for 5 years. I make music for commercials, TV jingles also a lot of my own projects. I made about 600 tracks.
I am currently developing kind of automatic composing program in Ableton. I was analyzing my steps when I am composing what sounds I am choosing and what melodies I am making and how.
I started program it. I have to say I am suprised how good it already works. I selected hundereds of sounds that i like and patterns in melodies that randomly changing by many modules.
I am not finnished with the program but I have to say I think creativity can be programmed. Music can be made by computer by itself and you would likely not reconize that it ia made by computer. If you put enough good data and you analyze the chords progressions key of the track and many other things, you can kind of programm your music I.
True, but that brings up the question of what level of AI would you need to be able to do what we do in the area of true creativity. Some argue that even AGI might not be “truly” creative, since the key to that may lie in how chaotic our brains are.
Sure we’ll reverse engineer the whole brain eventually but the time scale on that is very, very subjective, and I’m more thinking what AI might be able to do in the next 5 to 10 years in terms of this subject to try to avoid extreme speculation.
Hi mate, damn that’s really impressive! congrats! And yeah I know of AI composing based on theory and logic, currently playing around with some software called Orb composer which is pretty cool but still far off true creativity and needs human input for most of the tasks. I’d be interested to see the software you’re designing though! And my professors point was more on the side of can AI be truly creative by creating something new that it wasn’t programmed to do. So it’s more could a deep reinforcement learning algo learn to be truly creative, with the ability to “change its mind” for example? Of course there’s the argument that once we have AGI it’ll do literally everything we can do by definition. But it’s more the argument of can a machine, even AGI, really have free will? I made the point that the question of free will is a bit of an illusion for even us humans, since we can’t decide our genetics of which a lot of our fundamental decisions and desires are based on, but our brains are a ridiculously spectacular mess and some question that you may need this for true creativity? I’m not too sure myself so would love to hear your thoughts.
I’d actually argue the other way around - that you don’t even need AGI for AI to have creativity. Like your professor said and the previous replies alluded to, creativity boils down to additional variables for a machine to learn. How do we know that? Because that’s exactly how our brains work. After all, don’t forget that neural networks were inspired by the human brain, so, in theory, they can be made to learn anything we learn.
And again, as your professor said, creativity is very subjective, which means that it’s much easier to “fool” people into thinking that a machine is creative. I’m using quotation marks because it really depends on the person judging whether something is true art or not. But isn’t that the case with almost any piece of art? If someone randomly splashes color onto a canvas, is that art? To some people it might be.
And I really like the example you posted. AlphaGO is the very definition of narrow AI, as it can only play Go and nothing else. And yet, like you said, it made moves that no one could have predicted, thus developing some form of creativity. So in the end, I think it’s not a question whether they can be truly creative or when will they be, because there are multiple examples of narrow AI which are already doing some very creative “thinking”.
What is “creativity”, as our understanding is based on our experiences, in general, most AI would classify as having “creativity”?
But I would find it very probable that creativity is one of the things that a human-level intelligence would possess.
we have a saying in my country that goes there’s nothing new in the world except for new eyes . it’s to mean that nothing new ever exists in this world except when someone looks at things differently and we are all able to look at things differently because we have been through different experiences (genetics and environmental) So in my opinion, that is creativity; being able to imagine things none other could. If we can accept this, then it doesn’t matter if it’s a human, a rhinoceros or a computer software, if it conceived and created something new from this redundant world based on it’s experience, then it’s creative.
If this neural network library was given data to train on some images and the image generator was called then it would output NEW images based on it’s experience. I don’t see any reason to uphold the honour of being “creative” only to humans… after all, we have put a urinal in an art museum
That’s a really good point, a new way of looking at something is creative regardless of its origin, exactly! I was trying to make that point to my prof but couldn’t think how to word it, thanks for the help man!
I think that’s right. It’s about creating your own programming lol in a sense. Like I said, nature programmed us to be creative, to connect ideas and thought-forms and bring them together to manifest something in corporeal reality. Why can’t we tell AI to do the same!? Teehee.
Wow, great experience. The community is putting together commercials to help get the word out about sNet since the devs can’t legally pump it (as it is a utility). Message me if you are ever interested in getting involved!
I like to think of it this way: whether they can be creative seems like the wrong question. I prefer to think of it as, can they have a life experience they can draw from for insight for their artistic process.
In this respect, they can’t right now. But this can easily change on a dime. Plus creativity, defined in the sense of a subjective appreciation for ones life story, I think is pretty much an inevitable part of evolution of any sort of intelligence. For me, what seems prudent, is whether society will be ready for a poly-creative world.
And my answer to that is, yes eventually. Hopefully soon.
For sake of full transparency, this is a view I’ve grown into. I feel like thinking they can’t be creative is inconsistent with views about human / robot equality.
And yeah I know of AI composing based on theory and logic, currently playing around with some software called Orb composer which is pretty cool but still far off true creativity and needs human input for most of the tasks. I’d be interested to see the software you’re designing though! And my professors point was more on the side of can AI be truly creative by creating something new that it wasn’t programmed to do. So it’s more could a deep reinforcement learning algo learn to be truly creative, with the ability to “change its mind” for example? Of course there’s the argument that once we have AGI it’ll do literally everything we can do by definition. But it’s more the argument of can a machine, even AGI, really have free will? I made the point that the question of free will is a bit of an illusion for even us humans, since we can’t decide our genetics of which a lot of our fundamental decisions and desires are based on, but our brains are a ridiculously spectacular mess and some question that you may need this for true creativity? I’m not too sure myself so would love to hear your thoughts.
