[ 3 / biz / cgl / ck / diy / fa / ic / jp / lit / sci / vr / vt ] [ index / top / reports ] [ become a patron ] [ status ]
2023-11: Warosu is now out of extended maintenance.

/sci/ - Science & Math

View post   
File: 72 KB, 800x450, shutterstock_artificial%20intelligence.jpg [View same] [iqdb] [saucenao] [google]
5582939 No.5582939 [Reply] [Original]

Can robots be creative?

>> No.5582944

Define "creative"

>> No.5582963

>>5582944

Playful. Curious.

You create a robot with some sensors and actors. It first has to figure out how its actors affect its sensory inputs (and/or the environment).
Then it decides that doing certain things are 'interesting'. i.e. not only it gains new information from them but also they maximize the probability of it detecting a pattern.

Is that practical?

>> No.5582972

Creativity arises when atleast to concepts are connected together in an uncommon form. A machine could very well take two ideas in a database and make relations between them.

>> No.5582984

The issue with everything robot is that they are truly a blank slate. Whereas human brains come with a development hardwiring from our brains.

So the human brain have numerous routines and libraries and external influences and communcation and whatnot else that drives and manipulates it.
Whereas a robot got nothing at all except for the code and hardware that we added to it.

So yeah, robots can be creative, if we tell them to, or give them the right initial algorithm sets.

>> No.5582983

bumping for interest. been thinking about this recently

>> No.5582985

I've heard about machines composing music based on patterns of other pieces of music. That's a kind of creativity in itself I guess.

>> No.5583035

bump

>> No.5583071

If they use quantum chips, they could. They can do random decisions and random solutions

>> No.5583100
File: 378 KB, 560x733, artificial-intelligence-AI.jpg [View same] [iqdb] [saucenao] [google]
5583100

>>5583071
But creativity shouldn't be random. There should be meaning to it. In the same way that the evolution of human intelligence and thus our creative decisions and aesthetic perspectives aren't random.

Just a thought.

>> No.5583109

>>5583100
>creativity should have meaning behind it
So analytic type creativity? Then the robot doesn't even need QM chipset. Current robots can do it

>> No.5583122

>>5583109

Well yeah, the issue isn't acquiring specific chips or other parts. It is only the algorithm. Can we program a robot to be creative?
Is there even an algorithm for creativity and playfulness?
How did we evolve a sense of playfulness in terms of our neuronal system and can that be replicated in an artificial neural network program?

>> No.5583131

Whatever it is, it'll probably have to be based on reinforcement learning and particularly genetic algorithm.

>> No.5583127

Also what kind of things do we expect a creative robot to do. And I don't mean a robot who sits there and draws. This could just be a piece of strings with LEDs and light sensors which eventually figures out to do some kind of an interesting pattern with its lights maybe.

Or a robotic arm that autonomously learns to do something interesting with its fingers after a few thousand iterations.

>> No.5583139

>>5583122
It all depends on how we define or accept as creative. We definitely can create a creative robot, its just that we know how the code was created that prevents us from seeing it as creative. If the robot was shown to a person 40 years ago, they could think the robot was supersmart/supercreative/etc. Or even showing the robot to a technologically impaired person today would say the robot is creative. But to us, the robot would not be creative because we know its a program.

>> No.5583149

>>5583139
In a similar case, if we knew all the codings/programs behind human brain/body, then we will not see any creative works. All the actions can simply be explained via logic

>> No.5583151

>>5583139

Ah, but you may be missing the point. If we were to program an expected outcome, then yes, we would know that the robot is not being creative. BUT if we gave it some sensors, some actors, and algorithm that essentially says: "do something with your actors so that your sensors will receive some information, and then use that information to make your actors do something 'interesting' and keep making that 'interesting'." then we won't expect a specific outcome.

All we need to do is define what the robot should consider 'interesting', as in some kind of a reward-based process.
The robot will figure out WHAT to do, as long as it's interesting.
That's creativity.

>> No.5583156

>>5583149

But I don't think creativity and logic are necessarily mutually exclusive.
Creativity is not meant to be this magical thing that comes out of nowhere. It's all brain chemistry, and evolution, and a collection of synaptic impulses. But we still consider it creative, as long as it creates a concept that is not solely imitative.

>> No.5583171

>>5583156
>But I don't think creativity and logic are necessarily mutually exclusive.
Thats what I was getting at. Creativity is logical. Its only that we lack a total understanding of that logic that we see creative things. Creativity belongs to the same class as mysterious powers/phenomemas of the past. Until science explains it. Understanding logic completely will remove the creative part of it. Just like how magicians don't like to give away their secret to the audience as they believe the magic becomes "dull". So too is the creative.

>> No.5583173

>>5583151
But by following the programming and routine and what variables were introduced you should be able to recreate outcomes. That's the good and bad of programming an AI. Creativity in machines are simply bugs. If the source of the bugs cannot be sourced they are ghosts until better information is collected on each ghosts and eventually the ghosts are removed.

Creativity is to lose an objective standpoint and to allow a larger error tolerance.

>> No.5583180

>Creativity is to lose an objective standpoint and to allow a larger error tolerance.

This.

>> No.5583191

>>5583071
our brains are pretty insensitive to quantum phenomena so it's not necessary for quantum processing for things to be creative.

>> No.5583193

With a proper understanding of purpose and technical understanding a robot could be creative.

>> No.5583194

>>5582944

Qualia soul magic, obviously. Faggot.

>> No.5583199

Isn't creativity just taking two different things and making a link?
For example, for example: robots knows snow and sand are kind of similar, it knows you can make sand castles, so it decides to try and make a snow castle

>> No.5583201

>>5583193
A robot solves a problem. It does not create outside of a relative information to that problem. If it does that is in error because the information is not relative to the issue at hand and defeats the purpose of the robot. Stop anthropomorphizing tools. Even human understanding is simply an ever expanding umbrella of understanding and not random inheritance of knowledge. We are an objective organic database of relational information set to a trial and error query.

>> No.5583203

>>5583199
sand and snow are malleable materials.

Sand can be used to make sandcastles
Inductive reasoning would suggest snow can do the same.

Inductive reasoning is a skill not taught to people these days.

>> No.5583233

>>5583201

You're assuming that creativity doesn't solve a problem.

>> No.5583251

>>5583233
I explained a situation where a tool has to dig a hole and it instead paints a flower.

If it solves a problem in a logical and routine manner it isn't creative and is predictable even if it is in hindsight. Creativity is a 100% error tolerance with no objectivity. The presence of creativity would philosophically mean it has no purpose.

>> No.5583265

>>5583199
originally it just meant of or related to the act of creation

>> No.5583329
File: 238 KB, 1360x768, rob.jpg [View same] [iqdb] [saucenao] [google]
5583329

>>5583251

Creativity does not mean 100% error tolerance.
A painter can be creative even if he/she paints a picture of a flower.

The whole point of researching this is to show that creativity was evolved in our brains to serve a purpose. A rational goal for that matter. This could be replicated in machines.

If it is logical, then it is not necessarily 'routine manner' nor predictable.

It's not supposed to be a robot that was told to dig a hole that instead paints a flower.

It could be a robot that is told:
"here's a pen in your hand. here's a sensory input telling you what happens when you roll this pen on paper. oh and by the way, do something interesting that maximizes your successful complex pattern recognition."

The first few hundred iterations are spent on the robot getting the hang of how its movements of the arm and the pen affect its sensory inputs (i.e. learns to use the pen).

Then it starts drawing things and its pattern rec system recognizes that as a random pattern (not good).
It continues drawing, iteration after iteration, to maximize some kind of a pattern.
It ends up drawing a circle, with a circle, within a circle, with horizontal lines going through it. That's creativity.

You didn't tell it what to draw. You didn't even tell it that drawing is a good thing. It figures it out.

This kind of research is being done right now. Check out Jürgen Schmidhuber.

>> No.5583407

bumping for interest

>> No.5583419

>>5583329
So you set parameters and measures that focuses on a specific task and find the efficiency of the learning algorithm for hopefully a narrow set of results? Seems to be object driven and not particularly creative.

>> No.5583433

>>5583419

Yes but the target isn't a specific goal/object.
It is an 'interesting' action or pattern. Nothing more specific than that. That's creativity.

By your logic, human creativity is object driven too unless it's completely 'random' (which doesn't happen).

>> No.5583441

>>5583433
That's what I'm saying. There is no such thing as creativity.

>> No.5583452

>>5583441

Oh.
Well, clearly there is. It's the thing that we call creativity. If you want to call it something else then go ahead. But there is 'something' which the norm tend to identify as creativity, and the question is whether we can replicate that 'thing' (which most people would identify as creative) in machines.

>> No.5583460

>>5583452
So it's another Qualia concept.

>> No.5583558
File: 28 KB, 226x321, boschetar5.jpg [View same] [iqdb] [saucenao] [google]
5583558

>>5583329
>robot learns
>robot has self

Laughinghobo.mvk
>

>> No.5583578

>>5583558
Shh, let him believe in ghosts. We all did once.

>> No.5583609

>>5583558
>implying robots cant learn

>> No.5583669

>>5583460
>>5583609
>>5583578
>>5583558

lrn2AI

>> No.5583718
File: 27 KB, 320x320, robots-adam-and-eve-ai_1.jpg [View same] [iqdb] [saucenao] [google]
5583718

>>5582939
We already have one, thanks for asking though.

http://www.scientificamerican.com/article.cfm?id=robots-adam-and-eve-ai

>> No.5583722

>>5583669
Explain creativity in a learning format.

>> No.5583755

>>5583722

I already did, numerous times.

ctrl + f reinforcement learning

>> No.5583815

>>5583722

Also see >>5583329

>> No.5583834

>>5583755
Information gains are not creative. Efficiency would mean there is a measure to the pattern learning and if there is a pattern there is predictability. You keep talking about a concept that is not applicable in an objective model.

>> No.5584935
File: 63 KB, 565x600, 1357462084052.jpg [View same] [iqdb] [saucenao] [google]
5584935

Well, this kind of thread can be relatively good when someone isn't yelling "go back to /x/" everytime anyone say anything.

>> No.5584943

>>5582939
Your brain is just electrical signals.
The only thing a robot needs to be creative is the software.

>> No.5584956

pigeons and monkeys can use primitive tools to get food. thats pretty creative.
have the program associate and aggregate novel data based on observation and exploration and come up with similar plans. successful ones (with high reward value) are incorporated into even more complex actions.

>> No.5584973

Robot

Noun

A machine capable of carrying out a complex series of actions automatically.

(esp. in science fiction) A machine resembling a human being and able to replicate certain human movements and functions.

You meant AI instead of robot, right?

I think it's going to be interesting when one day someone creates and AI which is capable of noticing patterns about human behavior that is far beyond the capability of actual humans. That AI could then observe someone for a long enough period of time to figure out the things they find artistically appealing and create a piece of art for them that almost perfectly appeals to their sensibilities and taste. I bet that when that happens, that person will think "What a creative robot!"

>> No.5584982

>creativity
>consciousness
>>>/x/

>> No.5584998

>>5584973
No human engineered machine can commit complex actions, a robot is built from simple parts and runs on a program. Complex action implies results that are unpredicable from their initial state, a robot will never do this IMO

>> No.5585011

>>5584943
i love you

>> No.5585013

>>5583329
Just checked out Jürgen Schmidhuber and the development of genetic programing is a interesting way to set up a "robot brain."
The name drop is the only part of your post I approve of though

>> No.5585014

>>5584998
>Complex action implies
There's no implying without lying. You either back up that statement or you're a faggot.

>> No.5585022

There's an issue here, that you actually have no fucking clue what you are asking.

If you ever phrase the question in such a way it actually makes sense, the answer is either yes or no but people don't do that either.

>> No.5585137

A robot can seem creative if it was programmed that way. But it can't try things out that it doesn't know. And it can't do things in ways in which it wasn't taught. I'd say no.

>> No.5585159

>>5585014
Kk, that statement did seem to come out of left field. I'll retract that and instead say that a programs output will always be predicable while a system with a large number of interacting parts can produce the emergent phenomenon of complexity. I'm dissapointed that I can't truly back up my statements but there you go.

>> No.5585163

machines can't be creative. they only follow a set of commands that we assign. anything "creative" is just a derative of that.

>> No.5585166

to answer this in the most complicated jargon possible - yes.