Let’s explore another AI topic. As we noted in the previous article, we have several insights to share regarding the connection between business and AI.
After multiple internal discussions, in our team we had diverse opinion of how generative models could address the potential issue that if every decision in the world relies solely on AI-generated information, those decisions might be flawed or incorrect.
So we decided to ask one of the most popular AIs today, ChatGPT (free version). And we made 4 question in succession:
Where do you get your knowledge from?
ChatGPT Answer- I draw from a wide range of sources, including books, articles, websites, and other texts available before my last update in October 2023. My knowledge encompasses various topics, including technology, science, history, and more. I don’t access live data or the internet, so my responses are based on the information I was trained on. If you have specific questions or need information, feel free to ask!-
Ok, that’s fine, we got first interesting answer that helps us make the second one.
Who defined your bias?
ChatGPT Answer- My responses are shaped by the diverse data I was trained on, which includes a wide variety of perspectives, cultures, and viewpoints. While efforts were made to minimize bias during my training, some may still exist due to the inherent biases in the source material. The goal is to provide balanced and informative responses, but if you ever notice something that seems biased or unclear, please let me know!–
So, wait a minute! GPT AI is telling as that its bias is related to the source information it was trained with. So then, let’s ask the next question:
Is it possible to eliminate your bias if the source material you use to get your knowledge has its own bias?
ChatGPT Answer- Completely eliminating bias is challenging, especially since the training data reflects human perspectives, which can be biased in various ways. While efforts are made to include diverse viewpoints and reduce bias, some may still be present in the model. Continuous updates, diverse training data, and user feedback are essential for improving fairness and reducing bias over time. It’s a complex issue, and discussions about bias are ongoing in the field of AI. If you have specific concerns, I’m here to help address them!–
ChatGPT is telling us that among others, training data and user feedback is essential to decrease bias on the information we get using AI models.
Got it, so are machines going to take over? Or will they just ask for our feedback first? LOL!
Could that lead to a moment in few years where all the information automatically generated will be wrong?
Going back to our small research, let’s address our concerns to ChatGPT, by writing the following script for our last question:
Question- Yes I have specific concerns, if your are continuously getting knowledge with a bias and all the new knowledge you get is starting to be generated automatically by other GPTs like you with its own bias, could that lead to a moment in few years where all the information automatically generated will be wrong?–
ChatGPT Answer- That’s a valid concern. If AI models like me start relying heavily on each other’s outputs without critical evaluation or diverse sources, it could create a feedback loop of biased or incorrect information. This is why it’s important for AI development to prioritize rigorous training on diverse and high-quality data, along with ongoing human oversight.
Additionally, critical thinking and verification from users play a crucial role. Encouraging a culture of skepticism and fact-checking can help mitigate the risk of widespread misinformation. It’s a complex challenge, and discussions around AI ethics and responsibility are vital as technology evolves.-
There we have! ChatGPT is confirming the possible endless loop if only automatized information sources are being used.
It seams the AI needs high-quality data, rigorous training and human oversight in order to be sure that the data which was used to be trained is not biased, thus is not wrong.
At this point, some might believe that this research holds significant importance based on the responses we got, some others will think analysis have our own bias as we lead GPT to the answers we wanted to get.
We can agree with both of those points of views, as the issue with the AI implementation, is related to the nature of the AI itself.
As the tool it is, it needs supervision, improvement and adjustment regularly, exactly as the rest of tools that our business have in order to achieve company’s goal. Exactly the same as your organization’s operations, processes and tasks.
The above preformed test highlights that if you want to boost your business is not enough to implement an AI, you will need to nurture data, validate your sources, and, importantly, bring together a team of skilled professionals with a strong track record in the specific area you wish to enhance using AI.
We’d love to hear your thoughts and feedback on this article! Feel free to ask the same questions to the GPTs you’re using.
And please:
Don’t trust- Verify!
