A review of the article 'Generative artificial intelligence, human creativity, and art' by Eric Zhou, Dokyun Lee Review by Nana Ofori-Atta Oguntola
- Nana Oguntola
- Oct 3, 2024
- 5 min read
The researchers used a ‘dataset of over 4 million artworks from more than 50,000 unique users’ to reach their findings presented in the article.,
The article starts by acknowledging that AI tools can now produce content originally considered creative and the domain of humans. Tools such as Midjourney Dall E 3, Stable diffusion and ChatGPT enable human participation in the creative process through text or visual input.
The article expects AI to become even more effective in tackling creative tasks and will bring significant economic value.
The writers understand some of the issues which have been raised around AI that it has the potential to steal work away from humans, or that creativity will become stifled and generic thus the focus of the article is to discover the impact of GenAI on creativity including creative production and the product themselves.
Creative productivity
They found by utilising GenAI tools productivity increased 50% and stayed constant,in comparison to those not using the tools
Creative value
They found that peers responded more positively to work generated with GenAI and saw more value in the work over time than in work created by non adopters.
Content Novelty
This covers the subjective output of GenAI which the research showed was decreasing on one hand as the end product seemed to be similar and generic. The research however also showed that this was changing and increasing as artists made more creative demands on AI requiring more creative and innovative outputs.
Visual Novelty
There appears to be a homogeneity in visual content as AI seems to produce specific content from prompts which appear similar across users. This is probably due to the formulaic nature of AI.
Role of human creativity in AI-assisted value capture
Artists less proficient in artistic creation will develop some form of creativity with AI tools. Artists already proficient with creating artistic products can push the boundaries further with AI tools and can create content more adventurous or exploratory pushing visual boundaries.
Interestingly the research showed that content novelty increased in that concepts and ideas generation by the artists increased to produce the required image., indicating once more that content is king.
Platform-level value capture
Peer feedback provides information on effective platforms to use.
Robustness checks and sensitivity analyses
AI tools can serve as inspirational drivers for people to try or create art and the researchers believe this inspiration is more focused on the results of the creation than on the specific AI tools themselves.
Discussion
Current research shows that using GenAI tools can improve output in ‘coding, ideation, and written assignments’ though there are concerns around risks such as ‘potential disinformation and stagnation of knowledge creation’.
GenAI is viewed more favourably with an increase in content novelty but reducing visual novelty.
As GenAI reaches equilibrium where they all seem familiar as it is fed by past content, images seem to look the same and generic over time, thus developers need to start to improve input data to ensure more diverse content.
They suggest the theory of ‘Blind variation and selective retention (BSVR) may be applicable to the use of generative AI by artists as they generate new ideas and select the most promising ones based on their selection criteria.
There is a process of filtration and alerting which goes into the generative process before the final selection. Thus people with less theoretical or artistic abilities will produce less coherent work whilst people with artistic ability will produce more valuable work that will be better received.
‘This phenomenon in which AI-assisted artistic creation is driven by ideas and filtering is what we term “generative synesthesia”—the harmonization of human exploration and AI exploitation to discover new creative workflows’.
Thus the focus is on what ideas artists want to generate and not how and thus AI is not the source of the idea but the collaborator.
‘utilizing our research shows that over time, text-to-image AI significantly enhances human creative productivity by 25% and increases the value as measured by the likelihood of receiving a favourite per view by 50%’. This such an interesting statistic especially against the backdrop of resistance to AI generated content by slow adopters and policy makers.
By analysing ‘53,000 artists and 5,800 known AI adopters’ they discovered that creative production and artwork value ‘known as favourites per view’ were greatly increased.
Conclusion/Recommendations
This article by Eric Zhou and Dokyun Lee, 'Generative artificial intelligence, human creativity, and art' was not on my list of literature I wanted to review, but when I saw the title, I had to stop and read it and I am glad I did.
This article provided the relevant data required to understand the impact of using GenAI in creative practice in the current environment where most literature is focused on risk and the threats AI would have on creativity.
Their ability to assess over 53,000 artists provide a robust data focused basis for their results.
Their research indicated that humans who are more creative and able to effectively filter and prompt AI generated content will be more effective and successful in utilising AI. That is such an interesting finding as it does counter the narrative that the use of AI will take human jobs away and is a threat to human creativity.
The research also confirms that the use of GenAI is not a threat to human creativity but will encourage people not considered artists to engage in creative activity but will also allow people who are already artists to push the boundaries of AI and demand more of the tools they use.
These findings are supported by the article written by by Alex Frohlick entitled 'Artificial Intelligence and Contemporary Film Production: A Preliminary Survey' (You can read the review on my blog) which looked at the value of getting an arts degree in the age of GenAI, and, found that achieving an arts degree enabled adopters to create better content with GenAI tools as they had a better understanding of the theoretical frameworks around the production of content such as depth, focus, angles etc.
Moving forward, it will be useful to have research which looks at how viewers of GenAI inspired art respond to the content in comparison to non-Gen-AI inspired art.
Course/Workshop
My course 'How to use AI in your Creative Practice' is available herehttps://www.famk.co.uk/challenges
My one workshops can also be booked here: https://www.nanaoguntola.me/contact-8-2
Reference List
Alex Frohlick, Artificial Intelligence and Contemporary Film Production: A Preliminary Survey, July 2020,
Eric Zhou, Dokyun Lee
Generative artificial intelligence, human creativity, and art
PNAS Nexus, Volume 3, Issue 3, March 2024, pgae052, https://doi.org/10.1093/pnasnexus/pgae052
Kommentare