I believe that text-to-image and image enhancing tools are fascinating. However, we all could recently see how negatively can be the impact of such tools in hand of unprepared people, like the authors that published the paper " Cellular functions of spermatogonial stem cells in relation to JAK/STAT signaling pathway" that was retracted by Frontiers.
While I reckon that, at the moment, the AI tools seem to be limited and often not being scientific accurate, I believe that this can be improved with prompt engineering. Most of users I see using this sort of tools are not educated enough in prompting. However, "AI hallucinations" seem to be more the rule than the exception when real scientific images are needed.
Any comments and ideas on how to improve AI image generation for the scientific context? I would welcome examples of failures and success cases, if any :-)
You’ve raised some important points about the use of AI in scientific image generation. The potential of these tools is immense, but as you’ve noted, there are significant challenges to overcome. Here are some thoughts on how to improve AI image generation in a scientific context:
Regarding examples of successes and failures, a study on nuclear power used 20 AI-powered text-to-image generators and found that models like DALL-E, DreamStudio, and Craiyon showed promising performance in generating relevant images from general-level text related to nuclear topics3. However, these models fell short in accurately representing technical details of energy systems, reproducing existing biases, and accurately representing indigenous landscapes3.
On the other hand, researchers are already using AI-generated images to illustrate methods in scientific papers and to promote papers in social-media posts2. However, there are concerns that these AI tools could spur an increase in fake data and inaccurate scientific imagery2.
1. https://roboticsandautomationnews.com/2023/05/08/the-challenges-and-potential-of-computer-generated-images/68137/
2. https://www.nature.com/articles/d41586-024-00659-8
3. https://arxiv.org/html/2312.01180v1
Dear Fabrício A Pamplona ,
AI tools can also improve the quality of images for researchers who find it hard to translate scientific concepts into visual aids, says Rodriguez. With generative AI, researchers still come up with the high-level idea for the image, but they can use the AI to refine it, he says.
Company Cisco sponsors group articles related to the Artificial Intelligence skill.:
Benefits of AI-generated images Challenges of AI-generated images Best practices for using AI-generated images Examples of using AI-generated images Future directions of using AI-generated images Here’s what else to consider
https://www.linkedin.com/advice/3/what-challenges-using-ai-generated-images-i1fic
Regards,
Shafagat
The retraction of the paper "Cellular Functions of Spermatogonial Stem Cells in Relation to JAK/STAT Signaling Pathway" by Frontiers highlights the challenges of integrating AI tools in scientific research. While AI text-to-image and image enhancing tools offer exciting possibilities, their current limitations in scientific accuracy can lead to misleading results, as seen in this case. Improving the accuracy of AI tools through rigorous engineering and providing education on their responsible use are essential steps to mitigate such risks and harness their potential for scientific advancement.
This will be a great event to researchers in this field
https://fllm2024.fllm-conference.org/Workshops/PromptEng2024/
- Badges
- Science topic
- Similar topics
- Phytochemistry
- Extracts