Since its debut, generative artificial intelligence (AI) has been surrounded by rumors that its creativity has surpassed that of humans. A recent study published in the International Journal of Human-Computer Interaction, however, offers a more nuanced look. The authors found that while AI excels at generating rich interpretations of ambiguous images, humans still perceive human-made work as more creative, suggesting that systems like OpenAI’s ChatGPT, Meta’s Llama, and Google’s Bard are still inferior to us. However, their rapid development and adoption have fueled a debate about the impact of AI tools on society, the internet, and work. That’s why, as AI continues to advance and take on tasks traditionally reserved for humans, concerns are growing about the displacement of human labor, the implications for education, and the ethics of AI-generated content.
AI systems are getting better and better, rapidly changing the Internet and the approach to creativity. Image: media.wired.com
Contents
- 1 AI is changing the Internet
- 2 AI and creativity
- 3 How to evaluate creativity?
- 4 Limitations and conclusions
AI is changing the Internet
For decades, searching for information on the Internet meant searching search engines like Google and Yandex and following the links they suggested. Search was the main tool for finding information, so much so that few people even thought about it. Needless to say, the situation has changed after the triumph of generative AI systems – searching for information on the network is now controlled by AI, which greatly simplifies the process, since you no longer need to look for answers simply by following links.
It sounds and looks great, but meanwhile the Internet is filled with content of dubious quality, created using artificial intelligence. This, in turn, degrades search results and makes traditional online searches less useful and effective. The consequences of this shift could be serious, as traditional search engines could be almost completely replaced by new ones with embedded AI tools.
The Internet is in crisis, and artificial intelligence is to blame.
Another problem associated with the arrival of AI on the global web, as mentioned above, is the production and quality of content – with the help of ChatGPT, as you know, today you can do anything – create blog posts, write music, songs, poetry, etc. even communicate with friends, family members and the opposite sex.
More on the topic: How to find love using ChatGPT and artificial intelligence?
However, we cannot deny the obvious – modern AI is impressively good at writing content, even if it lacks personality, character and personal experience. As you can see, AI is already actively changing the Internet – from search engines to social networks and work applications and will continue to do so in both the near and distant future.
AI and creativity
Any conversation about the quality of content on the Internet is not complete without artificial intelligence, and the debate that smart systems have already surpassed humans in areas such as art continues. But what about the creativity of the same ChatGPT? The researchers note that due to the rapid development of generative AI systems, they are placing a lot of emphasis on the current capabilities of this know-how compared to the performance of humans.
In a paper recently published in the International Journal of Human-Computer Interaction scientists sought to empirically examine the current state of AI creativity compared to humans.
Generative AI has already had an impact on creative professions. But what is it? Image: deep-image.ai/
More generally, we were interested in understanding how modern AI systems will change people's perception of creativity and what social impact they might have on society and culture,” explains one of the authors of the study, Simon Grassini from the University of Bergen.
To study the comparative creativity of artificial intelligence and humans, the authors of the work used a «Figure Interpretation Task» (FIQ), which aims to interpret ambiguous abstract figures. This task is designed to assess divergent thinking, a cognitive process that involves generating multiple unique solutions or ideas in response to a problem (which is different from convergent thinking, which focuses on finding a single correct solution).
Read also: Microsoft's neural network declared itself superintelligence and demands worship from usersIt is important to note that the FIQ is a newly developed test, which means that the artificial intelligence system tested in the study (ChatGPT-4) could not obtain optimal results on training data until 2021. But because the FIQ is a multimodal test that requires both perceiving ambiguous visual images and generating creative textual interpretations of images, it allowed scientists to evaluate the AI's ability to «come up» original creative solutions, rather than simply extracting answers from existing memory.
AI comes up with creative solutions, but are they really that good? Image: scads.ai
The study involved 279 native English speakers recruited via an online platform, of whom 256 completed the test. These participants were asked to provide two different interpretations of each of four ambiguous figures. The goal was to provide interpretations that were as semantically distinct from each other as possible. Participants were given 30 seconds to describe their interpretations of each image, ensuring that they responded within the time limit that was consistent with real-life creative writing tasks.
In parallel with this, a chatbot with artificial intelligence ChatGPT-4 was tested, interpreting the same data as the subjects. The AI was asked to generate responses of varying lengths (one, two or three words) to mimic human responses (this was done to ensure a fair comparison of human and AI responses). AI sessions were conducted several times to collect a reliable data set of interpretations generated by the system.
Do you want to always be aware of the latest news from the world of science and high technology? Subscribe to our channel on Telegram – so you definitely won’t miss anything interesting!
How to evaluate creativity?
To assess creativity, a panel of eleven judges, who had no idea whose interpretations they were assessing (subjects' or AI's), rated each response on a scale of 0 to 2. A score of 0 indicated a basic or uncreative response, 1 indicated a creative response. answer, and 2 is for purely creative interpretation. The judges were psychology students who rated responses as part of their research training or in exchange for credit.
As the authors later discovered, ChatGPT-4 was adept at generating semantically diverse ideas and demonstrated its ability to think in differently in different contexts. “AI has demonstrated a higher level of flexibility in generating interpretations compared to humans,” the article says.
One surprising aspect was how well the AI language model performed in terms of generating semantically diverse interpretations of ambiguous visual images. The AI's flexibility scores, based on semantic distance analysis, were on average higher than those of humans, highlighting the capabilities of large language models, Grassini told PsyPost.
ChaGPT-4 does its job much better than its previous version. Image: cfotech.com.au
However, despite the AI's abilities,the judges generally perceived the answers given by humans as more creative This suggests that there are aspects of creativity that go beyond mere semantic diversity and are likely related to the richness of human experience and the nuances of symbolic interpretation.
This is interesting: What article on Hi-News.ru was written by the ChatGPT neural network
In addition, the answers of people who scored the highest scores exceeded the best answers of the AI - while ChatGPT produced a narrower range of answers, people demonstrated a wider range of ideas, which contributed to their higher creativity scores.
AI excelled at creating semantically diverse interpretations of ambiguous visual images, but when judges subjectively rated the creativity of the AI's output compared to human responses, they tended to perceive human responses as more creative overall, explains Grassini.
< /blockquote>
AIs are great at their tasks. Image: f8n-ipfs-production.imgix.net
The results of the work suggest that although artificial intelligence has developed the ability to generate a variety of creative ideas and is very good at following given instructions, human creativity still wins, at least in how it is perceived subjective.
Don't miss: How will technology change in 70 years? Forecasts of scientists and ChatGPT-4
The study also highlights the significant progress in the development of large language models and their current limitations compared to human cognitive abilities when it comes to creative tasks requiring multimodal processing and unlimited imagination.
Limitations and Conclusions
The new study, as groundbreaking as it may seem, has limitations and focuses on only one type of creative task, associated with ambiguous visual images that may not extend to other areas of creativity. Additionally, the AI's training data shutdown in September 2021 means it may not have been fully optimized for the FIQ task that was developed later.
Note that previous studies also showed interesting results – in one of the studies, AI outperformed the average person in a test of creative thinking. Another study found that artificial intelligence scored the highest in creative thinking using a divergent thinking task that measured the number and uniqueness of ideas generated in response to prompts.
AI currently cannot beat us in creativity. Image: dq8l4o3au0fto.cloudfront.net
This type of task emphasizes the ability to generate a wide variety of ideas, an area in which AI's ability to use vast data sets and the efficiency of its algorithms give it an advantage over the average human.
< p>You may be interested in: The “dark side” of chatbots: from declarations of love to conversations with the dead
One way or another, the long-term goal of researchers in this field is to better understand the cognitive mechanisms that underlie creativity of a person and how it relates to the implementation of creativity in artificial intelligence systems. Further research may shed light on the unique advantages of human cognition compared to AI approaches (and vice versa).