This series is designed to help people to understand modern technology, and become more confident in using computing devices. It is not designed to educate experts.
The author is involved in tutoring older students at SeniorNet, a New Zealand wide organisation. SeniorNet hopes that students will feel more confident in using their computing devices as a result of the learning opportunities offered. This series of articles shares that hope.
There has been a lot of talk about ChatGPT and other artificial intelligence machines of late. I have an account with openAI which allows me to use ChatGPT3. ChatGPT4 has lately been released and is said to be much improved, and less likely to hallucinate (the term used for ChatGPT making stuff-ups).
I thought I would run some tests to show the readers here what can be achieved with this tool. Yes, to me it’s just a tool. I don’t believe it’s sentient, or ever likely to be alive in any sense I can understand.
I do believe there is some very clever programming involved, and the system can “learn” to a degree, but there are no synapses learning in the sense we understand it.
So, what form did my test take? I decided to ask it to write an article on a subject I have some knowledge of, and interest in, to see what was produced.
My text input was: “Write a 500-700 word article explaining why Libre Office word processor is better than Microsoft Word.”
This article will be published as Tech Talk: GPT Writes For Me #1
I then asked the system to rewrite the article. This is achieved simply by clicking the button that says “Regenerate response”.
I haven’t reproduced the regenerated response. It was similar in content but set out differently.
I then rephrased my text input to ask the opposite to see what would be presented. The new text input was: “Write a 500-700 word article explaining why Microsoft Word is better than Libre Office word processor.” By doing this I was asking ChatGPT to prostitute itself, rather like a lawyer who is asked to defend with their best efforts whoever comes through the door, notwithstanding their own personal beliefs about the client.
The second article will be published as Tech Talk: GPT Writes For Me #2
You can read both versions to get an idea of how well the system writes.
A couple of notes here.
Firstly, I didn’t ask the system to write in a New Zealand style, with spelling as used here. So it has spewed out Americanised spelling. That was my bad. It will “spell” New Zealand if asked.
Secondly, I know that ChatGPT3 training is about two years old, so I was careful not to phrase my query in any way that would be time sensitive. The articles should be considered to be “old” as the research is all at least two years old.
As an example of this time cut off, I asked ChatGPT “what is different between ChatGPTt3 and ChatGPT4”. It answered:
As of my knowledge cutoff date of 2021, there is no such thing as “ChatGPT4” as OpenAI has not released any new versions of the GPT-3 language model after GPT-3. However, it is worth noting that OpenAI has continuously improved and updated GPT-3 since its initial release in 2020. These updates have included increasing the model’s parameter count, improving its performance on various language tasks, and releasing new API features for developers to use.
It is possible that OpenAI may release a new version of the GPT language model in the future, but at the moment, there is no information available about such a release or what differences it might have compared to GPT-3.
For a fuller description of the differences between the two versions see this article published on Techopedia in April of this year. ChatGPT4 is not yet fully available, and being a paid product will not be as fully explored as is the free version 3.
What do I think of the output? Would I use this and put my name to it as the author?
Good questions.
I feel that the articles are factually OK, but I have the sense that the system rather toes the corporate line and would use Word for no other reason than that’s the accepted norm. As Microsoft has a large investment in OpenAI, the company that develops ChatGPT, perhaps this is another reason that I feel the bias.
Being an open source advocate I have my own bias and can’t ever see myself putting pen to paper to laud Microsoft Word. Others not so invested in open source as I am may feel differently.
I thought the writing style was pretty bland. It had no sparkle; it was the sort of thing that would emanate from a corporate office. Had I asked for a different style then maybe that could be improved. But this series already consisting of three articles is long enough.
As for the question about passing this off as my own work: no, I don’t think so. I was always taught not to cheat, and I feel that to pass off someone (or something) else’s work as my own is dishonest. Also, as I said, the writing is bland.
If you want to play with this yourself you need to create an account. To do so, go to the login page here. You will need to supply an email address and create a password. From memory, they also asked for a phone number.
Each time I go to use the system there are a couple of pages to click through. Then I just type in my text prompt. The system will not (apparently) produce anything illegal, so please don’t try. Also, remember they know your email address so you could be located by the long arm of the law if you do transgress.
Your queries will be remembered, along with the responses, but can be erased. The free version is used by OpenAI for training purposes (they don’t say who is being trained; it may well be me).
OpenAI is a misnomer. Initially, the business was designed to be open source, but has since become proprietary and is financed in large part by Microsoft. You should therefore consider it in this light.
There are a number of YouTube videos about using this tool. Search for how to use ChatGPT3, or perhaps ChatGPT3 tutorial. Or if you want to go directly for the moneypot search for money making with ChatGPT 3.
I hope you will get some idea of the possibilities of this tool, and its limitations. I trust you will find the coming articles interesting.