-
Asked by seem1carp on 28 Mar 2024.
Question: Have you used AI using Chatgpt
- Keywords:
-
Neil Barnby answered on 28 Mar 2024: last edited 3 Jun 2024 2:55 pm
I have used ChatGPT and I often use it now to help me to make my writing sound better. I sometimes use it to help explain complicated technology, but I have to be careful because ChatGPT, like other LLMs (Large Language Models) is not always correct. AI can suffer from ‘hallucinations’, which is basically where they make something up when they don’t know the answer or if they don’t understand the question. Before you use an answer from ChatGPT you should check it is correct from another place.
-
Ravindu Ranaweera answered on 29 Mar 2024:
Yes of course. Also Google’s Gemini. The way I see it is, they are tools to make your life easier. Like a calculator. At the end of the day, you need to know how to tweak, apply and use the answers it gives. I use it a lot to get first versions of code and to summarise lengthy reports to find the topics I need.
-
Grant MacLean answered on 4 Jun 2024:
I’ve used ChatGPT, it’s variant for coding, alternatives such and Llama and Gemini. They’re good for generating ideas, and really good for frustrating you when they make something up and run with it like it’s the truth.
More recently when testing the new Gemini, I was pleased with its ability to interpret text in a video and extrapolate information from that. But was disappointed that it failed to correlate multiple feeds from the same time to provide more information. So, I went on to test it with playing the Journaling RPG “Thousand Year Old Vampire”. That was much more satisfactory.
To get the best out of AI, you need to be really good at Prompt Engineering. I have little patience for this, I want to ask it the direct question I want answered, I don’t want it to assume that because I didn’t mention something that I might still be interested in that.
-
Kirsty Ross answered on 19 Jun 2024:
Personally, I tend not to use it. The datasets that underpin ChatGPT and similar models can be biased, and this affects the answers that it gives. It is also very confident at presenting the wrong answer as the right one, and not everyone can spot when it is getting it deeply wrong. It also makes up references!
I may be behind the times though!
Comments
Cliff W commented on :
I have used ChatGPT and other LLM based AI solutions.
In practice I am not won over to their effectiveness. These tools can be great for having some fun and for producing some fairly basic, less important tasks but for anything serious I would have some reservations.
One issue is that I have found is that you need to learn how to phrase a question/request to the AI to get the best output. It can take several refinements to the input to get a suitable output. Also as Neil says you need to check the correctness of the output. And it is not just ‘hallucinations’ that can cause doubtful outputs but also the nature of the data that has been used for training the AI.
I will sometimes use AI to get a draft that I would then heavily amend or rewrite but IMHO the AI models are still not good enough yet. No doubt all of this will improve in the future.
stephaniebayne commented on :
Yes I have, It’s useful if you have an idea or some knowledge in an area and you want to find different language or explore wider knowledge. Always re-check the information though – AI ‘learns’ from information that is readily accessible and if it’s not sure it can make things up, or even output some very dubious ‘facts’.