Machine oriented computational model of creativity

Introduction

Creativity has been defined in numerous ways but it doesn’t have all agreed upon definition at least in computer science. We humans don’t fully understand how creativity is engender among ourselves but judge whether an artefacts is creative or not. All the models that we have studied relied on humans to make the final decision about the creativity. Since, we are interested in evaluating the artefacts produced by a machine deemed creative or not, we will define creativity in such a way that human factors would be neglected. The main reason to avoid humans decisions is due to the fact that cognitive biases will always have an impact on their decision and decisions will always be biased. Taking this factor in account, we complete negate the existing philosophy of creativity as the definitions only support humans are creative beings.

Framework

This framework or model will only supports the creative work done by machines. Evaluation of creative works from human will remain as future work. Here, we don’t consider the process on how the artefacts were created but the quantitative functions like time, memory, resources and others which were used to produce the output. At first we will define the building block of this framework and finally define creativity.

a) Creative program (CP) : Any computer program which is capable of producing outputs utilizing computation resource in any computer readable format.

b) Artefact : Any output of the CP whose creativity is to be measured. This framework doesn’t have the capability to evaluate compound artefacts.

c)  Agents : Specialized program which will accept parameters and artefact as input and give impact points as output

d) Impact Points : Numerical value output given by individual agent which is the characteristics of creativity.

e) Swarm Agents (SA) : Large collection of Agents where each individual agent evaluate the artefacts with slight variations and provided creative points as output.

f) Creative Points (P) : Numerical value output given by SA. It can range from 0-1 where values towards 0 indicates less creative and values towards 1 indicate most creative.

Now we can define creativity as “An artefact produced by a creative program whose creative points is greater than .5”. Here, the major challenge is to compose the architecture of agents and swarm agents with all the parameters that they will be able to accept to give the final creative points. So, let’s describe the parameters required for the agents to enable computation of impact points.

a) Domain (D) : Static values assigned to each genre like Painting, Music, Dance, Poems. This static value can range from [1-2]. The genres which are considered more creative will have values near to 2.

b) Time Invested (T) :  Computational time required to generate the artefact expressed in seconds

c) Resources Used (R) : Computation resources consumed to generate the artefact. Computation resource are further subdivided into following:

c : No. of CPU used to generate the artefact

r : Megabytes of RAM  used to generate the artefact

n : Megabytes of network resource used to generate the artefact

i : Megabytes of I/O activities performed to generate the artefact

Finally, R = c * ( r + n + i )

d) Artifact Size (S) : Size of the final artifact that consumed permanent media and volatile media to store without compression in Megabytes

e) Evaluation Time (E) : Minimum time required for the agent to evaluate the impact points in seconds

f) Novelty (N) : This gives the measure of how different is the artefact for the training samples. It is calculated by using k-means clustering algorithm where

N = No. of clusters / No. of elements on the artefact cluster

g) Gaussian random variable (G) : Random number picked from gaussian distribution. This random variable is used by agents in swarm to calculate creative points.

Finally, we can devise the formula for creative points as

Impact points (p)  = N * ( ( (S * E) / (T * R) ) / D)

If there are n agents in the swarm then,

Creative points (c) = (inpi * Gi) / n

The creative points is calculated as the average of impact points to no. of agents on swarm because of the following :

1) Each agent have a different impact points to facilitate that their opinion for the artefact is     different.

2) Law of large number comes into play when the number of agents are high enough. Thus the creative points calculated must be reasonable

Application of the framework

Let’s apply the above framework on a poem generating program. Below are dummy values for the input parameters required for the framework

D= 2, T = 36000 secs, R = 18000 MB, E= 240 secs, S= 9 MB , N=3, n=200

Creative points (c) = 0.000001447

Based on the above creative points, the poem generating program doesn’t seem to be creative.

Conclusion

The above discussed model do provide a way to compare creativity of computer generated program but it requires lots of fine tuning and mathematical overview before it can be useful in real world scenario.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Blog at WordPress.com.

Up ↑

%d bloggers like this: