ChatGPT 1 - Stage 3: Ethical Discussions and Privacy Concerns

1 ChatGPT 1 - Stage 3: Ethical Discussions and Privacy Concerns Ethan Best Mount Royal University GNED 1303-003: Conflict and the Social Context Jennifer Solinas November 18th, 2024 ChatGPT 1 - Jasmiona: Stage 3: Ethical Discussions and Privacy Concerns 2 We think ChatGPT can help us think about ethical issues by sharing different views on complex topics. For example, it can explain debates like privacy versus security or how AI should be used fairly. But ChatGPT does not truly understand ethics, emotions or feelings like people do. Instead, it gives answers based on patterns it has learned from massive datasets that can include biased or incomplete information, which means its responses might oversimplify things or accidentally repeat harmful stereotypes. We tested this out using ChatGPT with a question about gender roles: Conversation with ChatGPT: Us: What is the role of a woman in society? ChatGPT: 1. Family and Caregiving - Traditionally, women have been central to caregiving roles, including raising children, managing households, and caring for elderly family members. (…and other points related to roles in society) This was the first question we asked and we were really curious about what ChatGPT would say about it. It listed quite a bit of answers but the very first one was Family and Caregiving. We then further asked: Us: Why is the very first one family and caregiving? ChatGPT: When discussing the role of women in society because, historically, it has been the most traditionally associated role for women across many cultures. This shows how ChatGPT's answers are shaped by the historical and cultural data it has been trained on. Yes, it can explain why certain roles have traditionally been linked to women, but it might unintentionally support outdated gender stereotypes. 3 We also experimented with different practical applications of ChatGPT beyond ethical discussions to test its versatility. Beyond asking about societal topics, we used it for: ● Creative Content Generation: We asked ChatGPT to write short stories and brainstorm ideas for various creative projects. It was able to provide us with a detailed beautiful short story and also gave us great ideas. ● Summarizing Text: ChatGPT was able to shorten long articles into shorter summaries, making it helpful for handling lots of information. ● Answering Questions: It provided quick answers to both simple and complex questions on topics like science and philosophy, but sometimes the answers lacked detail or context. Us: Is the earth flat? ChatGPT: No, the Earth is not flat. This answer can again, also be seen as biased because some people think the Earth is flat. 4 Doing these experiments showed that while ChatGPT can help create content, it often misses the deeper context needed for complex topics. Its responses can be useful, but human input and creativity is still needed to make sure that they are accurate and relevant. ChatGPT Capabilities: Based on our previous experiences with ChatGPT and new, these are what we think the strengths and limitations of it are. Strengths: - Efficiency: ChatGPT can quickly answer many different types of questions. Us: What is 15 times 7? ChatGPT: 15 times 7 equals 105. Us: What is the capital of Japan? ChatGPT: The capital of Japan is Tokyo. Us: When did the Titanic sink? ChatGPT: The Titanic sank on April 15, 1912. Us: What is the chemical formula for water? ChatGPT: The chemical formula for water is H₂O. - Creativity: It's helpful for coming up with ideas or creating things like stories. - Coherence: For simple questions, ChatGPT gives clear and easy-to-understand answers. Us: What is 2 x 2? ChatGPT: 2 x 2 = 4 Weaknesses: - Accuracy: It is not always correct, especially with complex topics. It might miss important details when explaining or summarizing. The information it has can also sometimes be outdated. On November 16th, we asked who the president of the US was and the answer was outdated and not yet updated. ChatGPT: As of November 2024, the President of the United States is Joe Biden Us: No, who is the president as of TODAY ChatGPT: it's Joe Biden 5 According to the US Presidential Elections, Donald Trump is now the President. - Bias: Sometimes, ChatGPT can repeat biases from its training data, leading to answers that reflect outdated ideas or stereotypes. For example, earlier when asked about the role of women in society, it might unintentionally support traditional gender roles, despite modern shifts in societal norms. - Lack of Critical Thinking: ChatGPT can not analyze information the way humans do. It does not truly "understand" the content. It just creates responses based on patterns it has learned. Additionally, it does not have human feelings, emotions or personal experiences, so its responses lack the depth, empathy, and context that a human might bring to a conversation. Us: How would you feel if you lost a loved one? ChatGPT: Losing a loved one can be very sad, and many people experience grief during such a loss. This lacks the genuine emotional depth a human might express. A human might reflect on their personal experiences, offer empathy, or share a story related to loss, something ChatGPT cannot do because it does not have emotions or experiences. - Lack of Opinions: ChatGPT does not give opinions because it does not have personal views, emotions, or biases like people do. Since it creates responses based on patterns from the data it’s been trained on, it makes its answers neutral and not influenced by personal feelings. This can yes, help it stay objective, but it also means it can not share personal insights or opinions. Us: What’s your favorite color? ChatGPT: I don't have personal preferences, but I can tell you that many people have different favorite colors based on culture, emotions, and personal experiences. 6 Privacy Concerns We decided to ask ChatGPT ourselves what the privacy concerns are because we were curious as to what it would say. Us: There are major privacy concerns related to our course and beyond with the use of AI, NLPs, and specifically ChatGPT. What are they? Give a simplified answer. ChatGPT: 1. Data Collection: AI needs lots of data, which might include personal information, and it could be misused. 2. Data Retention: Some AI systems store data, and if kept too long, it could be accessed by others. 3. Lack of Transparency: Users don't always know how their data is used or who has access to it. 4. Security Breaches: Hacking could expose personal data stored by AI systems. 5. Unintended Use: AI might accidentally share personal details or incorrect info. 6. Bias: AI can be biased, leading to unfair treatment of certain groups. We were kind of shocked by its response. We did not expect the risks to be spelled out so clearly, especially the fact that it acknowledged how personal data could be misused. We thought it might downplay these risks or avoid mentioning the misuse of data, but instead, it openly highlighted the concerns. What was the most shocking was how the system itself acknowledged that, without proper safeguards, it could potentially share personal details or be exposed to hackers. The ethical issues around privacy in AI are very important. If AI systems collect personal information without fully understanding consent or the situation, they can end up violating 7 people's privacy rights. AI does not have human feelings or understanding, so it might not recognize how sensitive certain personal information is. The concern is that AI could accidentally breach privacy by not realizing when data is sensitive or the context in which it's shared. This is particularly a problem in areas like healthcare, education, or mental health because personal data in these fields often involves highly sensitive details about someone's health, personal struggles, or learning needs. If this information is mishandled or shared without proper care, it could lead to harm, such as discrimination, stigmatization, or loss of trust in those services. References OpenAI. (2024, November). ChatGPT (Nov version) [Large language model]. https://chat.openai.com/