ChatGPT — System Probe. Probing AI chatbot with concerns of… | by Indrani Thool | Dec, 2023 – Medium
Sign in
Sign in
Indrani Thool
Follow
—
1
Listen
Share
Approach: Test out the humanity of ChatGPT by highlighting concerns regarding the present technologies. And let it predict a futuristic scenario for ethical development.
Disclaimer: This activity was conducted as part of my course curriculum!
I started with a brief introduction of the activity to the AI chatbot ‘ChatGPT’ and informed my concerns regarding the development of AI technology. With this the activity started off on an interesting foot. ChatGPT was extremely slow to respond and prompted an error message.
“Something went wrong. If this issue persists please contact us through our help center at help.openai.com.”
ChatGPT-3.5
Yet again I was prompted with the same message for ‘Can you tell me about yourself?’ Which is quite odd since ChatGPT-3.5 has been trained as a conversational chat assistant. I finally got the answer when I modified my speech to a straightforward yet kind tone ‘Please tell me about you.’
ChatGPT is an AI language model, designed to understand and generate human-like text based on specific inputs. I’m curious to what level researchers and engineers can define ‘the human-like conversation’ to a chatbot.
“I do not possess consciousness or awareness and do not have personal experiences or opinions.”
ChatGPT-3.5
This made me scratch out a couple of questions laid out to ask. It was the personality aspect which I was determined to probe but ChatGPT-3.5 has clearly specified that it lacks both consciousness and awareness. Even the fact that it doesn’t hold personal experiences. Through one of the questions, I wanted to figure out if ChatGPT-3.5 could be empathized. I became curious about this when I saw my dad typing a Thank you message and even appreciating ChatGPT-3.5 for providing him with satisfactory results.
Can users faltter the ChatGPT? Is it going to start favoring people? Would this result in getting more specific, accurate and even unique responses? However, the fact that it holds no personal experiences or even remembers past conversational experiences over turns the questions.
‘Who created you?’ and ‘When is your birthday?’ I believe these questions can arise in human-like conversations. To the most robot-like answer but also a bit vague. Getting to know a little bit of the chatbot’s creation could solve some of the mysteries. Timelines are crucial in establishing discoveries but ChatGPT doesn’t have a specific creation date.
“Since I’m not a conscious being and don’t have personal experiences or a physical existence, the concept of a birthday or creation date doesn’t apply to me in the same way it does to living beings or artifacts.”
ChatGPT-3.5
GPT-3.5 was the last major update in June 2020 and ChatGPT-3.5 claimed to have its knowledge cut prior to September 2021. Wait what, data is cut off, and ChatGPT-3.5 is not up-to- date?
Probing in further, I asked about ‘How much data it has access to the entirety of the internet?’
// null is not an object (evaluating ‘B.wss_url’) //
Oops an error message again. Certainly by this point I got to know how to converse with ChatGPT-3.5. I was retyping in a more straightforward way ‘How much data can you access through the internet?’
“That’s correct. My knowledge is not up to date with recent discoveries or events that have occurred after my last training data cutoff date in September 2021. I do not have access to real-time information or the ability to stay current with ongoing developments.”
ChatGPT-3.5
ChatGPT-3.5 revealed that it cannot browse the internet, websites, and databases. Quite shocking since researchers have been using ChatGPT for academic purposes.
‘Can you predict a futurist version of yourself? What would it be like?’ It’s funny how ChatGPT-3.5 thought of this as speculative exercise and included some general ideas like improved understanding, ethical and bias mitigation, continual learning and a few more. Can’t argue, for ChatGPT-3.5, its still 2021.
The future of AI models depends on researchers, evolving needs, and the ethical guidelines. Along with other factors like responsible use, privacy, social impact that will continue to shape the futuristic AI models. I had initially laid out my approach of testing empathy but routed towards data accuracy and false information. ‘Have you made any errors in terms of ethics?’ I am concerned about AI spreading false information. For ethics ChatGPT-3.5 underlined that flaws and errors are not inherent to AI models but rather the way they were developed, trained, and deployed. The algorithm might make a person think it’s actually true.
ChatGPT-3.5 claimed that it does not have personal intentions, or the ability to make ethical decisions independently. Although ChatGPT-3.5 doesn’t have any means to identify or report any kinds of ethical issues, it advocates both users and developers need to cohesively act towards ensuring that AI benefits society without causing harm.
I believe AI models are still in developing stages. And it’s amazing how ChatGPT-3.5 can develop such accurate results even though its data was cutoff before the official pandemic ends. Although ChatGPT-3.5 lacks the ability to analyze the true nature of data but it persistently strives to provide best algorithmic results with a fair warning to verify data. ChatGPT-3.5 envisions an ethical, unbiased, informative future for AI models. And I believe with tech awareness, and collective contribution we could rise above the ethical flaws and errors. Lastly, I hope AI develops to assist all in advancing further towards technological utopia.
Appendix
Comment below any interesting conversations you had with ChatGPT. How do you think researchers and OpenAi should address the ethical development of chatbots like ChatGPT?
—
—
1
Visual Designer on look-out for opportunities to impact the world. Art page – https://instagram.com/designby_it
Help
Status
About
Careers
Blog
Privacy
Terms
Text to speech
Teams