Computers Are Incapable Of Creatively Writing Music

Simple. Train a deep neural network to be a critical listener that can evaluate music for uniqueness and creativity. Train it with hundreds of sessions with music critics, until its output matches theirs almost exactly. (You will never get to 100% of course, just as two music critics will never agree 100%.)

Then train a second deep network to create music, using the first network to provide training feedback. Seed the second network with initial random weights. Then let it go and run millions of trials. Eventually the second network will create music that human critics will judge to be unique and creative.

Nope. No one would have programmed this.

 

Log in or Sign up to hide all adverts.

This may illustrate what will be possible in the future

 

Log in or Sign up to hide all adverts.

Oh I don't know

Perhaps a lump of red thingy appeared before my eyes and because the red receptors in my eyes were pre-tuned to send a particular coded type of impulses to my brain

Of course no name for this colour or thingy until told and memory takes up the mantel

Leading me to exclaim next time I see red thingy "Oh look there is that red, what's it called again? My memory not working so good"

Please Register or Log in to view the hidden image!

 

Log in or Sign up to hide all adverts.

You've just been told

Is there any confusion in Michael's post?

 

I have posted some of these before but they are great for purpose of visually explaining what produces colors in our brain.

 

An example of this, albeit not to do with music, is with AlphaGo Zero, a neural net designed to play the game Go. It only had the rules, and everything else it learnt it did so through playing games against itself. Admittedly millions of them. By all accounts it came up with unconventional strategies and creative new moves, and is now, arguably, the greatest ever player of the game, even all other AI.

And specific to music: https://blogs.scientificamerican.com/observations/creativity-and-ai-the-next-step/

 

Memory after the first experience is explained

15 minute talks about people with Total Recall. Most interesting part is in the last 5 minutes MRI and the size of a particular part of brain

Please Register or Log in to view the hidden image!

 

I assumed you would realize that the Programmers are doing all kinds of things to implement their Algorithms. Of course they are are Configuring Neural Nets as part of the Programming. Neural Nets are configured using Algorithms the the Programmers develop or using one of the many other already existing Algorithms that will adjust the Weights.

 

The Programmers actually wrote the Movie.

 

No Confusion, because it Completely and Clearly does not show how we Experience Redness.

 

Playing games like this is very Algorithmic and Mechanistic. Configuring the Neural Net, even when the Programmers let the Algorithm do it, is all still just Programming. Nothing Creative here, just hard core Algorithmic Processing. I should say the Computer is not doing anything Creative but the Programmers (Conscious Minds) are pretty Creative. It is Creative Programming by the Programmers.

 

No they didn't....
The programmers exposed the GPT3 to a bunch of movies and sketched a rough plot outline.
The program invented the entire movie , hence the occasional oddity which it guessed was appropriate under the circumstances.
Remember an artificial evolutionary learning period of a few months v a natural evolutionary learning period of some 13 billion years.

The question of "how" has not yet been answered. The question "if" we experience redness has been answered. It's an axiom.
The counterfactual is there are people who do not experience redness, because they are color blind.

btw. AI would have no problem in recognizing wavelength frequencies with exquisite sensitivity.

 

I have heard about this kind of thing, but it was still fascinating to watch. These people have a large amount of Memories to deal with. They seem to be able to construct Visual Memories, so the Problem remains as to how the Neural Plasticity encoding of the Memories leads to the Visual Conscious Experiences.

 

Who wrote the Movie is certainly open to interpretation but the Computer is not going to do anything without the Programs. And who wrote the Programs?

AI can only Detect Wavelengths. It works in the Realm of Wavelengths. AI cannot know what those Wavelengths are converted into by the Inter Mind and Conscious Mind. AI cannot work in the Realm of the Conscious Mind.

 

Programs like AlphaGo are algorithmically based, because games like Go and Chess and Checkers are purely mathematical games of which Go is by far the most probabilistic in available patterns. It is an abstract strategy game.

OTOH, GPT3 is text (language) based like humans, and only has a few logical linguistic algorithms which allows it make "best guesses", just like humans (Anil Seth).
It is mathematical in essence, but not numerical.

An AI can precisely tell you what frequency red is and convert it into descriptive language, as well as the descriptive symbolic names we have assigned to other frequencies that are outside human perception, but not outside some other animals, like insects.

You could even teach an AI to produce the color red on screen based on interpretation of wave frequency.
How would it do that? Maybe self-referential algorithms?

 

What people often forget is that it takes humans some 15-20 years to fully develop their brains (and fill their memory banks) with engrammed experiences. A learning computer has a much shorter learning curve and can do that in a much shorter time, but cannot get around the learning part.

 

I have programmed Computers to do that sort of thing. No Neural Net required, if I understand your premise. The important thing to remember is that nowhere in all the Programming and maybe Neural Net Configurations is there any possibility that the Computer is ever aware of or Experiencing Redness. Wavelengths of Electromagnetic Waves are Correlated with various Experiences of Color but the Wavelengths of the Electromagnetic Waves actually have nothing to do with the Experience of a Color. The Electromagnetic Waves don't Look like anything. The Brain/Mind maps Wavelengths to the Conscious Experiences of all the Colors. Wavelength is a Property of an Electromagnetic Phenomenon. Color is a Property of a Conscious Mind Phenomenon. These are two different things. The Color Phenomenon is a Surrogate for the Electromagnetic Phenomenon. Generalizing this to all Colors and Light in general we can say that you have never Seen an Electromagnetic Phenomenon but rather you have always been Seeing (Experiencing ) your internal Conscious Light Surrogate.

 

That depends on your definition of "experience".

Are they different in principle or do they have common denominators

Right, as does an AI. The process is the same;
1) incoming data
2) sensory perception of incoming data
3) translation and distribution of sensory perception to the central processor
4) retranslation of distributed data by the central processor and conversion from data into visual symbols (tokens)
5) display of token symbols on visual or text media.

Ask a GPT3 AI if it is conscious . It'll will answer in the affirmative, because as far as the GPT3 "brain" is concerned it is! Are going to argue with the AI that it isn't?

Is it possible that Tegmark is correct when he proposes that some physical patterns ARE conscious and produce chemical emotional responses and emergent self-referential conscious experiences?

If you don't know how consciousness emerges, how can you declare that it does not in some particular way? It clearly emerges in the human brain pattern! That is a "hard fact".

 

There are no common denominators between Electromagnetic Light and Conscious Light. One is a Physical World Phenomenon and the other is a Conscious World Phenomenon.

Haahhhh! Pretty funny on the "Arguing with the AI" thought.

 

How do you know that for a "hard fact"? Why and how does the universe process mathematical phenomena?
Tegmark does not propose that consciousness in the brain rests on different physics than any other form of mathematical patterns. He proposes it's all the same, just that there are hierarchical evolutionary orders of emergent self-referential conscious refinement, from the single celled paramecium to humans. Just as spiral galaxies display the same mathematical pattern as the petal arrangement in daisies and seed arrangement in sunflowers.

Well ???