It seems like to answer this question we first need to define creativity in machine learning language. E.g., it is clear that clustering or classification in pre-defined classes/clusters are not creative tasks per say. But what about prediction using learned representation for input data not seen before? What about formation of new features using combinations/operations of existing features. And what about nonparameteric clustering/classification machines that create new classes based on input data.

If these tasks can be thought of as creative, then are machines already being creative?

The value of creative machines could probably be noticed only after we would map our mental representations with 'their representations', as in the case of the non-human animals. Wittgenstein said: 'if a lion could speak, we could not understand him'. probably we need a shareable semantic world or a mapping between these two worlds.

It seems like to answer this question we first need to define creativity in machine learning language. E.g., it is clear that clustering or classification in pre-defined classes/clusters are not creative tasks per say. But what about prediction using learned representation for input data not seen before? What about formation of new features using combinations/operations of existing features. And what about nonparameteric clustering/classification machines that create new classes based on input data.

If these tasks can be thought of as creative, then are machines already being creative?

See my opinion at::

http://www.intelligent-systems.com.ar/intsyst/proposedBrain.htm

and

http://www.intelligent-systems.com.ar/intsyst/effects.htm

I definitely believe that artificial brains will be very useful for everybody.

Thanks for sharing your views and specific links relating to this topic.

@Girish Chowdhary:

"prediction using learned representation for input data not seen before...", well this is a bit interesting since you are indicating as of how algorithms would be able to deal with uncertainty, or events and data not perceived before. I think the only way to counter such a problem is to increase the capability of the code/program to handle and experience a greater number of probable and unpredictable events, and also, back to the basic as of how human mind deals with unpredictability.

@Walter Fritz:

Your link points to interesting inferences. Learning is important but instruction design is more important, to be highly efficient to enable machines to learn and concept building required for intelligent perception.

But Fritz, could you be more specific, I am not able to understand this phrase you have written: I apologize for not being specific;

"Here is an example that shows that a person normally acts reactive:

She sees a clock, it is eight a.m. (the situation) So she grabs her things and goes to the door (action). The door is closed (situation), so she gets out her keys and opens it (reaction).

During all this the thought that she is going to work in order to gain some money, does not enter her mind (not goal oriented). She reacts "automatically", without "thinking". Goal oriented activity is learned and sometimes used."

The point is, she has already developed a memory about her routine tasks, well and good. But how does it relate to one saying that this task that she performed is not goal oriented? Then what for she is opening the door and getting her keys? How does that make sense? Does it mean conscious goal and conditioned subgoal?

How does command originate from the brain? and why does it so-we have an answer-to make decision, whether lifting a hand or reacting to someone else. Indeed stimulus perception can be composed into rules, but intentional states, like "she grabs her things and goes to the door..." is an intentional state of the mind, causality being her goal to reach the office in time.

And interestingly, you have mentioned "By the way, here the action concepts are identical to the concepts in the future situation".

Does it relate to intentional states underlying concepts of goal, since again, the future situation for "her" is to accomplish her

1. primary goal; to reach office by time

2. to finish her work related tasks etc.

3. do her job good in order to earn money (main goal)

She must have her goal in mind to be able to accomplish these small subgoals; opening door etc.

so, how does it relate to intentional states with the goals or rules?

Could you explain a bit more.

Much Thanks,

Regards,

Sidharta

Reading this discussion, I am thinking about "all this old stuff". Researchgate is the place for new ideas. We have done a long way through cybernetics, learning level I and II, self-organizing systems, Turing Machine and Cellular Automata up to the point we recognized that classical artificial intelligence is a nice but not real fairy tale. Since than we are beginning to understand what brain conscious might be. We learn that simple explanation like those of Walter fritz don't even scratch the main issues. So this discussion is really unnecessary at all. Sorry to say this, but as we are now searching on Quantum process level, things are getting a total new perspective, looking at the micro-world of neurons, microtubules an tubulins and what they have to do with brain quantum computing. But this doesn't mean that we will be ever able to transfer quantum process on machines.

Dr. Kreitman,

Very much Welcome. I suppose that my intention was not to ridicule or undermine someone's past efforts, however basic that might be, as you are correct to infer nevertheless that RG is a place for new ideas, not old stuffs. Indeed, that's what we are here for, to generate new ideas based old ones with new thinking. I was just mentioning that Walter Fritz missed some essential elements in his modeling, which is bit archaic in that respect.

Indeed you are speaking of Quantum process level to understand consciousness, Penrose-Godel and beyond. Quantum world microprocessing is the new tool, and we envision such a future for quantum computing, brain-machine interface, or to develop entirely artificial minds capable of embedding human level consciousness. But how far have we gone to understand the subjective nature of consciousness, in the light of Quantum science? How far do we need to go to be able to model such in machines? and how far can ANN take us beyond software-hardware correlates of artificial mental processes? How do we model such artificial parallel correlates of consciousness (APCC) which would support consciousness in machines, and by what models, Quantum processing or adaptive algorithms, or some other or combination of both? Well, it is for this why we are here, to understand what is really happening out there, and what is indeed really happening inside our mind.

My specific question is: How can we bridge the gap between classical AI and future synthetic intelligence(SI)?

And what does it mean when you say that such quantum process will never be able to transferred in machines? Then for what purpose do we attempt to model such mental processes in the light of Quantum systems? Do we really need that if we are doubtful about such procedure? Where is the present working model lacking, so we need to go for QP?

These are just some of the questions which generally arise in an inquisitive mind, sorry for asking you so many questions all at the same time.

Best Regards,

Sidharta Chatterjee

Dear Sidharta,

I agree fully that correct instruction is most important. On "how does it relate to one saying that this task she performed is not goal oriented": I am looking on AI processes from an engineering point of view. (so all processes shouild be veryfied by experiment). This means what exactly the program is doing at any time.

Here , at the moment, she opens the door, no other thought is present. There is only a present situation, a response rule and a present action.

It is quite possible that previously she had a higher level goal, but that is not present at the moment she opens the door.

Dear Walter,

When you are looking AI processes from the engineering side, don't you think that the machine should be conscious about what it is doing at a particular moment, or why it is doing so? The moment she opens the door, she is happy that she starts to think that she will be reaching the office in time and not get fired...

I think that programming alone is not enough to shape the future of AI, or the merging of minds and machines. We need deferential paradigms of thought, design approach complemented with programming, intentional state design and perhaps Quantum processing, as Dr. Kreitman above has mentioned, and as Dr.Chowdhary delineated the need for development of creative task clusters in robots.

First it is required what things should a robot perceive, and to redefine the perception processes in robots, we need better technology. Then,how such inputs should be processed and reused, recalled and create relationship between what a robot makes understanding of it perceptional inputs, and further, enhance its experiential capacities. This require design approach. Then perhaps, how to deal with errors of perception, and to develop creative cluster settings.

It is essential first to understand how our brain handles and processes input data, right from perception to experiencing such patterns, and code, decode and recode such patterns stored via indexing. We need more efficient data compression technology and better relational databases for machines, since we already have the best in human. This require programming, and code-decode-recode require far better technology. But there is still something essential beyond software-hardware paradigm. And that is the artificial correlates of consciousness as much as understanding what consciousness really is at the first instance. More discussion in this space will open our frontiers of understanding how to deal with the problems encompassing machine creativity. But opinions may differ, these are just my views.

I hope this explains.

Best, Sidharta

I have an internet site, about 200 pages long, were I try to explain the processes of the human and electronic mind. What and how it can sense and learn.

www.intelligent-systems.com.ar/intsyst/

This is difficult to compress into a few lines of message.

Best regards, Walter

Dear Walter,

thanks for your link. [...]i went through it . nice compilation about robby...wish you all the best!

Regards,

Sidharta