The Dark Side of AI

Generative AI. It’s probably something we are all familiar with, living in this modern-day world with new technologies such as ChatGPT, GenesisAI, DALL-E, and even video-making resources such as Sora. These days, it seems as if AI is involved in our day-to-day lives, with some even saying it is starting to govern our lives. For example, an engineer might use ChatGPT to help generate code ideas; an architect might use DALL-E to generate some possible designs for a house… So it seems as if this new type of AI would be beneficial in the advancement of our society. However, this is just one side of the story.

Generative AI has huge potential, along with some equally large risks. For example, while it can be used to design, say, a waterpark, it can be easily hacked into and manipulated (with a variety of methods including inserting fake training data, exploiting AI biases, prompt manipulation, and more), and since some Generative AIs cannot think for themselves, this could easily lead to problems (we’ll get more into that later).

While some would say AI is helping to advance our society, others say it is doing more harm than good. For example, if AI were to do some jobs even better than we humans could, it could potentially knock millions of jobs off the market. To give a little context, in a recent report of 750 business owners, 44% reported that there would be layoffs in 2024 because of AI. This would obviously be bad for society and would also plunge our world into an economic crisis, causing the infamous “AI takes over the world, and no one can do anything without the help of AI” trope. Additionally, influential people such as Elon Musk have made bold statements that AI will take over all the jobs on the market. Although this can be seen as good for humanity, it also begs the question: How will many households survive like this?

Generative AI is also not as good at “keeping secrets” as one might think. Companies such as OpenAI have explicitly stated that any input their bot takes will be used as training data for the Generative AI. For instance, a while ago, Amazon sounded the alarm to stop sharing code with ChatGPT, as they have noticed their code being used as training data for ChatGPT. This sharing had many harmful impacts, such as ChatGPT spreading the code to other users, and using the code itself as training data illegally. Obviously, this kind of code-stealing isn’t legally or morally allowed, but what are we going to do about it? 

Another potential reason for harm is that AI could be used for malicious purposes, such as hacking and posting harmful messages. Since AI doesn’t have a sapient mind, it cannot think for itself. Instead, it bases its responses off real-life information, including the internet, which certainly harbors some bad things. This is another problem: AI that can’t think for itself cannot determine what is right or wrong, and so far there is no good solution.

One example is the Microsoft chatbot Tay, released a couple of years ago. At first, this chatbot seemed to be interacting well with users, promoting the Microsoft interface and pulling more users in to use Microsoft, but after some time, Tay started posting racist comments, transphobic insults, and generating other types of harmful text. This brings to light another issue of Generative AI – it generates based on the user’s prompt and tries to respond directly to it. While this can be incredibly useful sometimes, it can also lead to malicious manipulation, as we can see with Tay. Furthermore, this also shows up in our current generation, with ChatGPT and other chatbots such as Gemini. Although these major companies have implemented guardrails, it is still possible to hack into the code, or feed it incorrect/racist information, leading to more harm.

However, AI can also be a priceless resource that helps us with our jobs, helps us gain information, and stores information that can be saved and transferred easily without human interference to the next generation of AI, which can analyze it and save it for other services to use. Ultimately, it comes down to a philosophical debate. Should we keep this invaluable resource that can save us countless years of work, or should we banish it because of the potential mass harm it could cause?