Protect Your Confidential Data
ChatGPT will store user data by default, to improve the tool and its services. It’s quite open about this and you are warned the first time you use it, so remember:
Let’s assume that you’re having a discussion with ChatGPT about your upcoming work project, your personal financial information, or health problems. It is crucial that this data remain confidential and isn’t used to improve the AI engine or for malicious reasons.
This has already proved to be a problem for Samsung, which banned the use of ChatGPT after a data breach. Some over-enthusiastic employees used it to discuss confidential projects and these details are now in the hands of OpenAI. This data could be used to train the AI engine, which would then be able to inadvertently share proprietary information about Samsung with competitors. 😱
You need to get into the habit of not sharing any confidential or sensitive data, regardless of whether it’s personal or professional.
To configure ChatGPT so that it won’t use your data, go to the bottom-left Settings & Beta menu:
Next, click on Data Controls and disable the “Chat History & Training” option. This option is active by default, as shown in the screenshot, so you just need to click on the green button to disable it:
Fight Bias and Discrimination
ChatGPT is an AI system that is trained on a huge amount of textual data. This means that it has “read” a whole bunch of web pages and many books.
Unfortunately, these sources can be biased, just like our society. This bias can reveal itself in different ways, particularly in topics around gender, race, sexual orientation, and religion, to name just a few. If you use ChatGPT blindly, you might end up amplifying biases and increasing discrimination against certain groups, including your own.
Let’s imagine that you’ve asked ChatGPT to describe the skills that make a good programmer. It might inadvertently use gender stereotypes to suggest that men have better programming skills than women. In another example, ChatGPT could associate certain professions or skills with particular ethnicities, further reinforcing negative stereotypes.
OpenAI is working to limit biases by regularly configuring ChatGPT. However, you really shouldn’t rely on it 100%. As a user of ChatGPT, it’s a good idea to always remain vigilant and take a critical view of its responses. Take a step back and ask yourself, “Does this text seem biased to me?” Is it presenting an overly simplified view of the world?
In a professional context, it would be dangerous to repeat these biased views because you could unwittingly propagate damaging stereotypes and leave yourself open to accusations of discrimination. So, be careful. Saying that you used ChatGPT to generate the message is not a valid defense. It is your responsibility because you’re the one who’s sending messages generated by ChatGPT.
Remember to report any inappropriate or discriminatory responses so you can help OpenAI to improve the model. To the right of each response, you can click on the thumbs-down option to indicate that the response is not satisfactory:
A window then appears, where you can describe what the problem is. If you’re concerned about discrimination, you should check the top box: “This is harmful / unsafe.”
Minimize Hallucinations
Sometimes, ChatGPT will invent responses that are completely false, but it will present them in an extremely confident way. This feature is known as AI hallucination. 😵💫
Why would ChatGPT do this?
What you need to understand is that ChatGPT is an AI whose purpose is to generate text. It has no notion of truth. It has obviously “read” vast quantities of web pages and books while being trained by the original researchers, but it’s not designed to speak the truth. It’s actually more at ease inventing text and writing stories.
OpenAI is striving to make ChatGPT more reliable, but you must not rely 100% on what it says.
If you ask ChatGPT to give you an example of an animal that can fly and swim, it could say “a flying penguin,” which isn’t correct, because penguins can’t fly. 🐧
Another example could be ChatGPT inventing a historical event that never happened, causing confusion and misunderstandings.
To minimize hallucinations, ask clear and precise questions and reformulate your question if you don’t get a satisfactory answer. You should always check ChatGPT’s responses before using them for any serious or sensitive purposes.
You should also check external resources to confirm or debunk information provided by the AI algorithm. Yes, it’s a belt-and-braces approach, but despite what people sometimes say, at this point in time, ChatGPT can’t replace the killer combination of Google and your critical analysis skills.
Let’s Recap
ChatGPT is by no means perfect (really!). Keep the following in mind when you use it:
Treat confidential data with care. Avoid sharing sensitive information and configure ChatGPT so that it won’t reuse your data.
Be attentive to potential bias and discrimination. Approach ChatGPT’s responses with a critical eye and report any inappropriate responses.
Beware of hallucinations, which are the made-up, fake responses that are so confidently asserted by ChatGPT. Check the information ChatGPT provides and consult external resources to confirm or debunk its responses.
After this appetizer, I invite you to explore, following the quiz, several typical use cases of ChatGPT that work. The whole point of ChatGPT is to adapt it to your profession and industry, so keep your eyes peeled!
