How AI is changing creativity: A physicist weighs in

At the recent Rakuten Optimism 2023 business conference, one topic made its way into almost every discussion: the meteoric rise of artificial intelligence. The event even served as an opportune occasion for Rakuten CEO Mickey Mikitani to announce an exciting new collaboration with OpenAI.

One session, titled AI × Creative: Unlocking the Potential of Innovation for the Future, invited Kyoto University physicist Professor Koji Hashimoto to share his perspectives on how AI tools are transforming science and human creativity.

Hashimoto has published a number of popular books about physics, and is currently working to combine physics and AI to create a new field of research he calls machine learning physics.

Creative uses for AI in science

One application that Hashimoto and his colleagues are exploring is AI as a tool to engage with published research.

“I’m having the AI read papers, and engaging in discussions with it about their contents,” he told the audience. “In traditional physics research, your research results are published as papers. Those published papers are read and understood by other scientists, giving them new insights. That’s the way it’s been done for centuries, but it’s changing: Now, AI can read the papers and serve as something like a grad student, and we can engage in dialogue about the paper’s contents and better our understanding.”

“You can make AI create all sorts of wild ideas, connecting disparate elements.”

Professor Koji Hashimoto, Professor of Physics at Kyoto University

Thanks to AI, the very user interface of scientific research is changing, Hashimoto proposed.

“If this continues, you could have an AI that is familiar with quantum field theory, or astrophysics, or any field,” he said. “We’re heading towards that future right now… Diverse AIs, diverse creators will be able to talk and stimulate new ideas.”

Hashimoto also sees AI as a source of unintuitive ideas.

“You can make AI create all sorts of wild ideas, connecting disparate elements. You could give that to a bunch of scientists and let them pick out the interesting ones – let those ideas develop from there. I think even the AI we have today can be used for this purpose.”

Is creativity in danger?

The session was also joined virtually by creative director Kashiwa Sato – the designer behind countless iconic Japanese logos and brands, including Rakuten’s own.

“I’m actually looking forward to it,” Sato revealed. “The most important thing for creative work is inspiration. The moment you see a new perspective, your existing values come crashing down, and you’re able to produce something new. I yearn for that kind of stimulation. I’m really excited to see if AI can give me ideas I’ve never had, show me things I’ve never seen.”

Hashimoto was largely in agreement, but also warned of some potential pitfalls in the direction existing AI tools such as ChatGPT are taking.

“The AI we have now is very safe. If you ask it something, it will give you an opinion, but it’s a very boring one.”

Our current AI tools are trained to stay within certain parameters, but Hashimoto believes that this limited approach is leaving serious potential on the table.

“In scientific fields like physics, we all want to tackle ideas that other scientists haven’t thought of. This culture is a very efficient way to churn out new ideas,” he said. “It’s a culture in which we also share our failures. Test all sorts of new ideas. And from there, the very interesting ideas are nurtured, and they end up revolutionizing everything.”

The implementation of overly conservative AI tools in the research evaluation process carries a danger of diluting this important culture, Hashimoto ventured.

“Research in physics and other scientific fields is evaluated by how many citations it gets. Personally, I don’t think this is ideal, but that’s how it is,” he explained. “One application of AI could be to judge for you exactly what kind of research would earn the most citations. AI tools could potentially even be implemented into the review process. I think this would bring about a dire situation and lead to us losing our culture of testing new scientific ideas.”

“We’re welcoming a new definition of creativity, and that doesn’t mean that we’ll lose what we had, but rather that new territory will be born in addition to what we had.”

Professor Koji Hashimoto

Hashimoto considers himself a “creator” in his own field. To continue stoking scientific innovation, maintaining a culture of creative failure will be key.

“I think creativity is something that you feel. Say a young student brings me an idea, something super interesting that I’d never thought about before – for me, this eureka moment is the source of creativity,” he said. “We need to build an environment in which motivated students can test all of their ideas freely, using cutting-edge technology. And it needs to be a culture that forgives failure. You might test and test and have 99% of your experiments fail. But those tests stack up, and it’s the remaining 1% that changes the world.”

Will AI take our jobs?

One topic dominating the recent AI discourse is the threat it presents to the job market. Hashimoto compared it with the rise of the computer and its effect on physics research.

“In the 1990s, people started owning PCs in their homes. Supercomputers were capable of doing massive calculations, well beyond the ability of humans. Until then, physics had been done largely by hand – humans solving things with formulas was regarded as an important part of the field,” he explained. “With the arrival of supercomputers and PCs, anyone could solve these problems. I think people were concerned that this might mean the end of the physics profession. But what actually happened was that a new field suddenly emerged, called computational physics. Traditional physics was able to work alongside computational physics to create the physics we know today.”

Hashimoto sees AI having a similar effect going forward – including for creative professions.

“The rise of new technologies has an effect on fields such as science… It’s not a destructive effect, but rather one that will change what it means to be creative,” he said. “We’re welcoming a new definition of creativity, and that doesn’t mean that we’ll lose what we had, but rather that new territory will be born in addition to what we had.”