Will AI replace artists?

Also Consciousness, As long as scientist cannot explain what it truly is I doubt their ability to reproduce it.
Without consciousness or awareness, there is no intelligence.

Or literally.
I wouldn’t object If you call Jackson Pollocks Art as “stupid”. I actually agree.

I am not sure, there is nobody that learns, its just process optimisation. You might argue that a human does the same but there is a consciousness that predates everything else - a machine has no such thing, there is no pure machine consciousness after it was manufactured - no awareness - so who is it that does learn?

I just want to leave my two cents.

There’s a important aspect of A.I. that people neglect: an optmization function.

In short every AI has to satisfy a goal, usually this goal is to mimic a dataset and produce more data similar to it, so in fact AI usually copies stuff, but when the goal to optimize is to produce art pieces, the optimization function is “human amusement”, how much the human likes the pieces produced by AI, this is sort of tricky to do, it’s easy to copy a dataset of art pieces but the produced result will only be as good as the dataset if not a bit worse and it will only be as diverse as the dataset also.

The only way to get around this problem I believe would be to directly wire a human to a machine and have the AI feed piece after piece while a sensor measures the response of the brain and how much dopamine it releases… and I dont think anyone is willing to undergo such a horrific experience.

Edit: there’s also the AI who can play against itself, and the optimization function is to be better than the past self but this only works for competitive games with no human input soo…

Edit(2): humans also have one or two optimization functions, one is quite obvious (To not feel pain) and the other not so much (to not feel social pain).

You are thinking in terms of neural networks and given datasets. Though, what prevents a computer from experimenting and trying new things out? If you have a dataset from which a neural network can learn what people may or may not like, you may have one piece of the puzzle. The dataset may also be used to measure how close or far away it is from the already existing work. This may give us the opportunity to find pieces which are likable and yet unique in some sense.
Now, you may let it experiment with different methods to produce images/music or whatever. It might mix them and try different things. When it achieved a few things, it may even mix those techniques to create something more unique.

I don’t see a reason why consciousness or other not well defined things would be needed, as it has been suggested by others here.

Technically nothing prevents a computer from randomly trying out new things, that’s called a genetic algorithm, I use it a lot because it’s simple to implement but it’s not a very good or fast algorithm.

this as far as my understanding goes, will limit your results to lie on the scope of the dataset, the computer may have learned a concept of what looks good from the dataset and use it to create new good images, but it’s extremely likely that it will not hold accurate for anything outside the dataset so if the computer try to stray too far it will likely produce random/garbage looking results.

I am indeed thinking about neural networks because they are the most promising and frighteningly creative kind of machine learning at the moment (refer to GPT3).

But who knows what new algorithms might be invented in the future.

Yea although…
I do believe recurrent neural networks are conscious even if in a very rudimentary way.

You may also argue this robot is conscious (depending on of your concept of consciousness)
it’s from the OpenWorm project

My response was more aimed at the concept that creativity is unique to humans and is used as justification for their selfish domain over the rest of creation, despite the personal touches both chicklids and bower birds add to their love nests. Machine consciousness and the rise of the perpetual machine just happen to be favourite subjects of mine, so yes, perhaps a bit off topic… :wink:

Here is the future of AI, Humans, and Artists. :slight_smile:

Perhaps the 2000 years have already passed, humanity has become extinct in the traditional sense and is replaced by artificial hyperintelligence more human than humans and now we are all simulations in virtual worlds that believe they are real humans instead we are all managed by a very intricate and complex multidimensional AI who is imagining us to understand life and emotions and everything else (ok the answer is already known and is 42) of their previous creators who were probably real humans. :grin:

Whether it is a genetic algorithm, reinforcement learning or other techniques which explore doesn’t matter. All of them are very slow as of today. My argument is that those can come up with new, original ideas. It doesn’t work well yet, but conceptually, I don’t see a reason why they could become a lot better.

Neural networks can generalize to a certain extent. You don’t have to train it on every possible dog image, such that it gets an “understanding” of how a dog looks like. If we apply that to e.g. digital art, we might train it with currently popular works with as many and diverse styles as possible. As a first step, we might tell the exploration algorithm to replicate existing artwork, such that it learns how to do that. Now we may tell it to explore combinations of those techniques which still look good according to the neural network, but are special or new, also according to a possibly different neural network.
This is certainly heavily oversimplified and hardly achievable as of today. But, I don’t see why this should be possible in theory (and in practise in the future).

I am convinced that those kinds of techniques are going to exist in the future. Nevertheless, I don’t think they are going to replace artists. I believe they are going to help artists to explore a broader range of possibilities and help them to overcome issues like a writer’s block more easily. Artists with a realizable vision might get a tool to actually reach and explore it.

1 Like

Almost inevitably. At least somewhat. Hopefully by then we’ll all be safely immersed in bio-pods experiencing our collective dream-state from the comfort of our homes.

It’s not hard to extrapolate that a few hundred years in the future the average “incredibly good artist”, or exceptional storyteller or prodigal musician will barely hold a candle to what AI will produce. I think the golden period or renaissance 2.0 will be the in-between zone (basically starting from now) where AI is learning from but also inspiring humanity and vice-a-versa.

Here’s a few GAN pics I’ve made:

1 Like

For God’s sake! - a computer cannot feel! Therefore it cannot produce art. It’s as simple as that.
It can maybe produce something that tickles your eyes, at best, (by “mixing and matching”) but never your SOUL!

Just chiming in on this…

I think the answer is somewhere in between yes and no. For some of us.

For more repetitive tasks, I’m pretty sure artists will be replaced for the most part. With that I mean a computer with some parameters will be much much faster in generating variations on a theme.
Have a look at PDG in Houdini as an example.

There’s already some addons for Blender that can whip up variations on models based on kit bashing. Same for textures, landscapes etc. AI is already good with this based on very large datasets.

On the other end there’s still some artists needed creating the ‘basics’ for this to make the process happen. Or make decisions how to continue of a certain result.
For a lot of things AI will mean we will be able to do more with less time and artists.

For all the metaphysics please watch "Ghost in the Shell’. The '95 anime, not the one with Johansson. :wink:

I didn’t check the whole thread. Could you please tell me what the question was? :grin:

I rather liked Dorro’s coining of the phrase "simulated intelligence (SI)," because I really think that this is a more practical scope to our explorations.

Another phrase that I rather like is: non-determininstic algorithms (NDI).” This simply refers to any possible situation where “what the programmer expresses in source-code form” does not deterministically (i.e. “pre-ordained”) determine what the finished algorithm will actually do. Virtually the entirety of what we now call “machine learning” and even “neural networks” could fall into this category. But I like the term because it is … well … humble. Or maybe, grounded.

There are various problems which regularly face it which are simple enough to “place boundaries around,” but which condemn us to great drudgery if we are actually forced to “do it by hand.” We naturally and properly look for creative ways in which computers – at their present state of development – can usefully assist us “within these boundaries.” Literally, "to relieve us of at least some of that infernal determinism equals drudgery."

Even if the computer is “still an idiot,” at least it might be “a more useful idiot!” :smiley:

Yeah OK sundial, we get it, you don’t have to repeat yourself over n over again.


I disovered this article about VFX, yall should take a look! :