Hey guys -
Thought this might be upping the ante a bit.
https://openai.com/blog/musenet/
I still don't finding these pieces to show a subtle hand in the carving, but there are, it seems, some interesting twists and turns - as it relates to the promt.
What do you think?
I recently used 'midjourney' - an AI for visual imaging - for the cover of a book. (This was after being a little dissatisfied with a work an artist offered for the project.)
After some intricate prompting - 'mix cubism with Escher in a jazzy style..with piano with worn ebony keys....etc'
or -'Kandinsky style with ascending keyboard with ...etc. " - Im sad to say I have mixed feelings - about how much I liked how it turned out.
Replies
Somebody posted a similar discussion recently, but for some reason I just can't find it anymore.
But anyway... I still believe that the value in art comes from the human touch behind whatever outward trappings a work may have (that can be imitated mechanically, with various difficulty/ease). I recently listened to the AI "completion" of Beethoven's 10th (or something like that), but found it lacking in Beethoven's "spirit" so to speak. It sounds like Beethoven and is written in a Beethovenian style, even down to the style of orchestration, but lacks the characteristic vehemence and maverick of a genuine Beethoven work. The sudden unexpected wrench that Beethoven loved to throw in just when the listener's expectations were setup a certain way, and, most of all, the large-scale overarching structural drama that Beethoven is famous for. At the bar-to-bar level it's sometimes genuinely difficult to discern the AI output from real Beethoven, but at the larger scale of the overall structure of a piece, the difference is IMO pretty obvious.
Hi Gregorio, thanks for discussing this, it's an important topic. AI is advancing rapidly on many fronts and the more it is used the better it gets. The Chat GPT software is in the headlines because it is good enough that material from it is finding its way into academic and professional publications. Of course like self-driving cars it makes mistakes and I think we all like hearing about that.
I know artists feel threatened by this and react negatively to it but it's here to stay and it's only going to get more prevalent. We have to adapt to this. I'm not saying to necessarily start using it; keep using a pencil and paper if that's what works for you. But to me it is just another tool and there are plenty of examples of people using it creatively. We all adapted to computers; albeit reluctantly in many cases.
As I see (or hear) it, AI can only copy.
It can't create until it has reason to self-express for which it needs its own emotions and I think that's a long way off.
It'll do the hacks out of business, the ones who compose drizzel backing TV commercials and documentaries. And that wretched "music" we get hanging on the telephone awaiting a human. It could probably rattle off contemporary atonal music just as some composers seem to do when they really don't know what they've composed. Maybe action film soundtracks which are very cut-and-paste these days.
.
Yep. At the moment, from what I understand, AI copies, it does not create. When it can create a new style of form, then I'll take a look. But as music is a living breathing art form, AI will never be able to create human articulation. We are just too diverse, and unpredictiable, machines are not.
AI will certainly soon be able to compose the vast majority of what passes for "music" these days. What it will not be able to do, as Dane and H.S.correctly point out, is bring the emotional need to express something specific. It's this, and not technical ability, that can make music interesting. And I'd much rather listen to a flawed work which has something to say than a competent one which has nothing. Contemporary serious music has tended to move on from purely technical exercises to an interest in instrumental colouring but there are still rather few composers who try to continue the great humanist tradition. Something which computers by definition also cannot do.
That anyone now can generate AI composed music may raise some surprising ethical and legal issues, such as:
As has happened with many other changes computer technology has brought about, many people will try to answer such questions by saying, "The situation with AI created compositions is just like ..." But the problem with that, as with many of those other changes, is that the new entity isn't just like anything that existed before.
There will no doubt be many other issues. The only thing we can be sure of is that lawyers will make money.
Several members have commented that AI doesn't at this time have emotions or 'spirit' and that this quality is critical to the creation of meaningful art.
I think that this opinion leads to a philosophical discussion and there is no obvious right or wrong answer at the present time. Part of the problem is that the standards for success are subjective. I think however that there are three possibilities to consider.
If human beings are endowed by their creator with divine qualities then AI will never have those qualities unless the creator gives them to it also.
If humans are instead just an example of highly sophisticated biochemical engineering then it is just a matter of time until we are able to synthesize more human qualities. But does software have to experience emotion in order to create an emotional and original work of art? I'm not sure.
The third possibility is that this process will be like the current effort to build driverless cars. That effort has become much more complicated than the developers had originally thought it would be and some of the projects have been curtailed. Some experts in the field question if completely driverless cars will ever be practical (worth the necessary investment). So maybe achieving emotional AI will remain an open question? Of course nobody will die in a car wreck from an unemotional piece of music so the standards are lower for the success of AI music. We'll just have to wait and see.
one quite interesting analogy is chess engines. Until quite recently they were regarded as playing boring, predictable chess which was good enough to beat humans because of the far greater calculational ability. But then Alpha Zero -- which taught itself how to play chess entirely through millions of games self-play and evaluating what worked and what didn't and judging a position solely on the chances of winning the game and not from any material consideration-- came along. Now this programme, and others which are following the same principles, are regarded as much more interesting and creative than humans could ever be by trying out things and taking risks that only the greatest geniuses would ever think of
I still believe it’s a long way off, Ingo. Perhaps one day AI will have reason to express itself but I can’t see why it should at present, still being just computer programming. All it can do is collect vast amounts of data and infer information / compositions from them. The musical impetus still comes from a human – the one pushing AI down a musical route. But without human intervention why would AI want to present its outpourings to an audience or even another AI machine?
Humans were creating compositions sometimes entirely intuitively: Mozart, the Beatles to think of two, or otherwise composed having (surprisingly) little knowledge: maybe some theory and ability to play an instrument. I posit that’s because they had reason to express themselves with something in their evolution pointing them at music rather than another medium, and that they had something to express. They develop as their technical skills grow which is probably in part by assimilating more music around them and as their life experiences compound.
I’m a big believer in inspiration – that spark that comes from somewhere – the Good Lord, the Soul or whatever is the driving force fundamental to an individual, that somehow impels one to awaken the secretary to translate it into organised sound, with a view to communicating it to others. It’s possible one is influenced by other music, perhaps even copy an idea but hopefully stamps one’s hallmark on the result.
To me, until musical AI has a reason even if subjective to communicate its alleged creations to others it’s just another toy, a complicated one. "Soulless," it might be usable commercially. Having said that, we can’t predict where this is going and whether humanity becomes redundant altogether.
Ha! Should I be looking for a job at the Umbrella Corporation?
AI seems to be all over the news these days, but I still have my reservations. For decades, the AI people have been promising marvelous achievements and the replacement of human agents within a short time, but each time it has turned out to be far more complex than anticipated and progress has been far slower than touted. These days things like ChatGPT are making headlines, but at the core it's really just an imitation, a toy model (and a greatly simplified one at that), of brain biochemistry. On the one hand it's rather impressive how far we've come in terms of copying some of the brain's functions, but on the other hand, objectively speaking, it's still a long way off from a self-aware, thinking entity. It has no reason to live, to fight and survive, to make its own path in the world, to push against its environment, the boundaries that box it in. To struggle and achieve. Without such a drive, there isn't really much meaningful emotion to speak of, and consequently nothing to express in art. It can imitate the way humans respond to queries by "learning" from hundreds or millions of "training data", which somewhat parallels how a human child might learn how to speak, but that's still a far cry from the inner drive of a human child that propels him to do things beyond what is fed to him by the environment. (As the father of a very independent child prone to rebel against what is set before him, I say this with a living piece of evidence before my very eyes.)
A question for the philosophical types to ponder: does an AI model that is fed with all the stimuli that a human child might face from birth till some arbitrary reference age N, and is therefore able to respond to queries like the child might, actually share the child's inner thoughts and feelings? Or is it just a philosophical zombie, a cleverly-made manniquinn that parrots how the child might respond, but inwardly devoid of the actual drives and impulses the child experiences that drives him to respond the way he does?
It's well-known that the human brain "hates the lack of stimuli", as some describe it, and in the absence of input (e.g., by being artifically confined to a place where there is no external stimuli) invents its own virtual stimuli to fill the gap. My own opinion on this, is that this is merely an expression of something deeper: the inner drive of a real living human being, that an AI model fundamentally lacks. The urge to live. Not just survive, but to "really live". To create, not just to consume. To seek out its own path in the world, not just to be content with what's presented to him by the present circumstances. The AI model has no such propulsion, because it wasn't created to forge its own path in the world. It was created to digest what's fed to it, and to reproduce a certain kind of behaviour that imitates the data presented to it. It wasn't created to struggle and fight for survival, to push against the environment that confines it, to overcome obstacles and turn around and bend them to its own purposes -- it has no purpose of its own. It doesn't need one, because the humans who created it has already provided it one; it doesn't have any reason to find or create one for itself. And without such a fundamental drive, why would it want to create anything? To express anything? It is sufficient merely to parrot what has already been fed to it, there's no motive to invent something new, from which true works of art emerge.
-
1
-
2
-
3
-
4
-
5
of 7 Next