Even before the current AI boom, ethical concerns and considerations were always on the minds of data scientists. Now, with the growth of generative AI in the public imagination, these concerns have only exploded as different issues crop up when it comes to AI. While AI continues to grow, there are many questions that are left unanswered. With that in mind, here are a few ethical questions to consider as AI continues to scale across the world.
Data Sources – What was the model trained on?
One of the easiest places to see this concern is the art community. In the summer of 2022, the art world was inundated with a constant stream of AI-generated art. At first, many worried that this could displace artists, but as time went on, how these models were trained caught the art world by storm. That’s because many artists and their supporters began to push back against AI that generates art because many of these programs trained on existing human art without the permission or consent of the artist.
In fact, back in October, this came to light that Stable Diffusion used the artwork of real artists to train its generative AI program. This brought up a lot of ethical concerns as what could be considered plagiarism began to be asked. Much of the anger came due to the lack of notice or permission from artists. Many felt left out of the discussion and their distinctive styles used to train a program that only required text information to produce similar work that took years for them to master.
It’s not just within the art world itself, the question of privacy, data governance, and more have to be asked as more and more models are training on more and more data without the consent of the people they’re drawing the information from. This doesn’t even touch on the issue of bias when a data set isn’t properly handled. One only needs to recall the infamous chatbot Tay released by Microsoft. It only took a few hours of interacting online before the AI become so hostile and racist that the tech giant had to quickly pull the plug.
Output
Already ChatGPT has found itself in hot water due to content it generates claiming that certain individuals were connected with some serious crimes. In one case, the person it generated has filed a lawsuit against OpenAI. And for those who may not fully understand, it’s simple. All it takes is a simple accusation online before any Google search involving your information results in some false information. This could ruin both lives and reputations. This is why there have been calls in the past few months to slow down AI development until issues such as AI hallucinations could be solved.
In another case, a lawyer used ChatGPT to cite and build up a case file for a lawsuit he was working on. The only problem was that ChatGPT cited cases that never happened and made them up. As you can imagine, the judge wasn’t all too pleased as this could potentially ruin someone’s career. Finally, there is the output that could be used that honestly the AI is unable to tell the user that it’s now how it was designed. We saw this a month ago when a Texas Professor falsely accused students of using AI by feeding their essays into ChatGPT and asking the chatbot to determine if they were written by AI.
If you are wondering, no ChatGPT can’t do that. This is why output is a critical issue for generative AI as so many have flocked to the technology and are now depending heavily on the information it produces.
Unintended Consequences
There are other consequences from bad output that go beyond what was just mentioned. Imagine a news conference from a world leader claiming that they’ve made contact with aliens from another world. Or, more dangerous, an announcement claiming their nation was going to war with a neighboring nation or group of neighbors. As quickly as Deepfake technology has advanced, it’s becoming harder and harder to determine what is real and what’s not, but it doesn’t have to involve a world leader. There are already reports of deepfake voice chat scams.
AI-powered criminals use deepfake technology to essentially clone the voices of their targets and then contact their families in a bid to extract money or other resources. This isn’t science fiction, it’s already here.
Intellectual property
If there is one thing that companies don’t want, it’s their intellectual property being mishandled. This has gotten so bad, that even lawyers at Amazon had to remind software engineers not to use ChatGPT to answer coding questions that could compromise their property rights as each input to the chatbot only teaches it more. Thus, OpenAI could, in theory, gain a competitive advantage as workers from potential rivals use their AI to answer work-related questions.
Data privacy violations
One industry that is extremely hopeful about generative AI is healthcare. The technology help reduce administrative costs, remove information bottlenecks to improve care, and provide doctors with greater insight into medical history. Generative AI is also at the cutting edge of developing new medicines and machines that have promising potential.
But there is a catch. Unlike most industries, the healthcare field has very strong laws, both state and federal within the United States, when it comes to the protection of medical records. This is why data governance and enterprise platforms are crucial when it comes to employing AI in the field. If not, the private medical records of tens of millions could be out in the open. But this is only one industry. Most LLMs, until very recently, utilized web scraping to train their models.
So it stands that there’s a great deal of information used to train models that have tremendous privacy concerns. This in itself is why calls for responsible AI have grown so rapidly in a global manner.
Conclusion
Though generative AI has become the new superpower in data science, many are flocking to enhance and future-proof their careers, there are still plenty of ethical concerns that must be addressed if the data science community wishes to see the technology continue to benefit the global community.
ODSC is hosting a free Generative AI Summit on July 20th. Get your pass today, and come away with new insights and enhanced generative AI skills. So, what are you waiting for?
