After watching a video from Simon Colton which basically explained the concepts and premises behind computational creativity my initial reactions harked back to that wonderful scene in Jurassic Park where the charismatic Ian Malcolm states ‘Your scientists were so preoccupied with whether they could, they didn’t stop to think if they should.‘
The thought processes behind this just seemed so alien to me, talking about create a code base that could eventually analyse itself, correct itself and write new code for itself without human input? Sorry but does anyone not remember Terminator? Also my business side then kicks in and begins to realise ‘but if this game can design its own content then it does not need me, therefore I am redundant, who is going to pay my mortgage?‘
Obviously artificial intelligence in games is a massive subject and something that has been established and developed since the inception of video games. This has also extended past creating game content, with several notable examples of machine learning being applied to not just creation but also the playing of video games. We have the famous example of the Deepmind project, where researchers created a computer program, AlphaGo, which was the first piece of software to defeat a Go World Champion.
But while I started this post with a precursor the dystopian thoughts of a world run by machines, funnily enough so commonly used as a theme within the games industry, it seems to be a while before I need to be worrying about who is going to pay my next electric bill.
But what can computational creativity do for games?
As stated by Bowling, Fürnkranz, Graepel and Musick, (2006) in their journal article Machine Learning for Games there are several topics of particular importance that machine learning can aid us with, ‘learning how to play the game well, player modeling, adaptivity, model interpretation and of course performance.‘
The most interesting area for me from this subset would be player modelling and model interpretation. Learning a players habits and being able to capture and correct their errors for them has been the realm of usability for a long time. I have spent countless hours watching consumers play my products trying to understand just why they cannot see the big red button in the middle of the screen (when of course its not that big actually, neither is it red). Human observation could do with a helping hand in this area and machine learning seems like the perfect opportunity to be able to absorb and interpret the amount of data that would be required to really take any benefit from this process.
But what if that modelling would allow the software to adapt itself to suit the play style of the consumer? Would that allow for more ease of access and therefore more engagement with that player? Would I need to design the systems to handle this? Create the content from which the system itself learns from before it can create the content for itself?
This is a big subject and something that would be interesting to delve further in to.
References
Colton, S (2021) ‘Computational Creativity: Short’ [online] Available at [https://flex.falmouth.ac.uk/courses/912/pages/week-2-computational-creativity?module_item_id=54081] (Accessed 9th June 2021)
Deepmind. 2021. AlphaGo: The story so far. [online] Available at: [https://deepmind.com/research/case-studies/alphago-the-story-so-far] (Accessed 12 June 2021).
Bowling, M., Fürnkranz, J., Graepel, T. and Musick, R., 2006. Machine learning and games. Machine Learning, 63(3), pp.211-215.
Figures
Feature Image: Deadline, 2017. Jeff Goldblum in Jurassic Park. [image] Available at: <https://deadline.com/wp-content/uploads/2017/04/rexfeatures_5886121s.jpg> [Accessed 9 June 2021].
0 Comments