The roadmap to ethical AI

Our daughter Nguyen, 15, is currently enrolled in a private high school in France. She also takes part in the school's art club. Here in France, kids not only have a chance to engage in hands-on art activities like painting and shaping, but also are frequently involved in discussions about painting-related topics.

Recently, teachers have been discussing AI in the arts with their classes. It's been lauded by some and criticized by others. Nguyen has publicly opposed this new technology, despite our close relationship and the fact that I am her biological father and an AI researcher in France. She is convinced that AI Art violates the copyrights of artists all over the world. She believes that nothing made by AI can ever be considered creative, and that it is unethical to create or produce content in this way. Nguyen even told me that she was only willing to reveal her shame to the anonymous online community. She did not dare reveal that her dad is an AI expert, despite the fact that she was working on an AI Art project for class. She has always thought of this as "embarrassing."

I was quite surprised by my daughter’s reaction. Until now, I still think she must be at least a little bit proud of her father, who is not only a university lecturer but also an expert in artificial intelligence, a quite popular field among young people today. But to be honest, I think her statement makes perfect sense, and indeed she is talking about the serious problem that AI technology is facing.

On February 6, 2019, Getty Images, the largest image and video provider in the world, filed a lawsuit against Stability AI, the first company to create AI Art, for "flagrant infringement of the intellectual property rights of Getty Images." According to Getty, Stability AI violates copyright and trademark laws, and it is accused of stealing over 12 million images from the Getty database without permission.

In order to train their algorithms, AI Art tools need a lot of images and artwork, which are frequently taken from websites without the owners’ permission. Then backlash ensued from artists when they realized this technology enabled the blatant theft of their ideas. Even for ordinary users, the use of their social media profile pictures as AI training data will cause a serious privacy intrusion. Who knows if they will one day discover "fake" pictures of themselves in an unexpected location?

That is what is happening in photography and painting, but what about content creation? ChatGPT has recently created quite a stir in Vietnam. This chat bot has the ability to write stories, essays, and poetry in just a matter of seconds. My colleague bragged that he would ask ChatGPT to create a fairy tale for him each night, and with a little editing, he would have a "new piece of story" to lull his kids to sleep. He can now tell new stories that cannot be found in any books and are, of course, much more interesting than the old ones that his kids did not love.

However, what we want to question here is the originality of the stories. "Well, it's just bedtime stories for my kids, why so serious?" Obviously, we were aware of the problem but ignored it, because it did not cause serious consequences. But on a different scale, piracy will not be justified by the saying "it doesn't matter."

Users are astounded by ChatGPT's capacity to provide seemingly accurate responses. No one, regardless of their intelligence, is able to retain such a vast amount of knowledge as the platform can. The technology behind ChatGPT can turn a massive amount of text-based knowledge into a computer-based probabilistic model. This enables quick information searches, summarization, comparison, and connection.

However, the main issue with ChatGPT is that academic integrity and transparency are not ensured. While citing the original author of the information is a fundamental requirement to prevent plagiarism, ChatGPT never does so.

Furthermore, ChatGPT can provide users with utterly sexist and racist information through its filter. In order to categorize people as likely to be outstanding scientists, Professor Piantadosi at UCBerkeley asked ChatGPT to write Python codes. It turned out that the platform believed a great scientist must be a white man. After this opinion was published, it learned how to fix errors and to learn from past mistakes, but these mistakes are just the tip of the iceberg with countless other potential risks.

So should we continue to use technologies like AI Art or ChatGPT? As an AI researcher, I still have to recognize ChatGPT or AI Art are useful products. If copyright and ethical issues are solved, they can become powerful tools to help users come up with more creative ideas and to work more efficiently.

However, the fight over copyright and ethics will continue for some time between established businesses and startups and content creators like musicians, authors and bloggers. I believe as AI continues to advance at a rapid pace, countries will need to create legal frameworks on the topic and give it an official status in order to avoid copyright and ethical disputes.

For users, when I polled my students about their experiences with ChatGPT, up to 70% said they felt it was important to verify the accuracy of the information by consulting alternative sources. The accuracy and completeness of the information returned by this chat bot are two of the factors that will be evaluated. Like Wikipedia, the current version of ChatGPT is a go-to resource for many people, but is not officially recognized by the scientific community. Users should exercise extra caution to avoid being held liable for any false or slanted information they submit.

As a lecturer, I have no objection to students using ChatGPT for learning. On the other hand, I seek to include this technology in my lectures as a tool to support and stimulate students' thinking skills. To prevent students from cheating by using ChatGPT, teachers can use the site's answers as prompts for new questions, encouraging students to think critically and avoid the site's trap. As a result, students who use ChatGPT will develop a greater depth of understanding and analytical prowess.

Since the development of AI is still in its infancy, blunders are to be expected. It is ultimately up to the user to decide whether ChatGPT or AI-Art will be a useful tool or a threat.

*Nguyen Nhu Van is an AI expert at the Orleans University in France.