02-21-2023 06:33 AM
I wonder what ChatGPT's confidence was on these examples. I wonder if its threshold for deciding doesn't have a lower limit so, even if it has a very low confidence, it will return the best thing it could come up with. I work with people like that...
02-21-2023 08:41 AM
@jcarmody wrote:
I wonder what ChatGPT's confidence was on these examples. I wonder if its threshold for deciding doesn't have a lower limit so, even if it has a very low confidence, it will return the best thing it could come up with. I work with people like that...
Usually people want me to return an answer - any answer - so I will give them an answer and a confidence level. The confidence level has been as low as 10 percent.
02-21-2023 10:58 AM
@jcarmody wrote:
I wonder what ChatGPT's confidence was on these examples. I wonder if its threshold for deciding doesn't have a lower limit so, even if it has a very low confidence, it will return the best thing it could come up with. I work with people like that...
To the people who provided the transcript link, it does give an answer confidence metric.
(Side note: my confidence of this answer is 0% 😁)
02-21-2023 11:00 AM
@StevenD wrote:
That's really interesting. My girlfriend was listening to a podcast about Chat GPT and the expert (I don't know who) said Chat GPT was really good at giving convincing answers that are wrong. That answer is a perfect example of that.
So what you're saying is that I have competition?
02-21-2023 12:25 PM
I have been using github copilot (https://github.com/features/copilot) which uses GPT3 to write code and it is a truly disruptive technology. If you are a developer you can now write passable code in any language that GPT3 supports (most text based languages) even if you have never written a line of code in that language. If you are writing in a language you know, you get a major productivity boost. If you have never written code, you can now learn to write code in any GPT3 supported language very, very fast.
It will be interesting to see what happens in the next few years. Hopefully LabVIEW will get on the bus pretty quickly.
02-23-2023 12:41 AM
Screenshot of the 4th point from my previous post
All points are correct, except the last one, slow and fast regulator are just function names, not actual circuitry.
Unfortunately i did not screenshot the circuitry it drew.
05-30-2023 02:03 AM
@Kiwi-wires wrote:
I have been playing with ChatGPT too at https://chat.openai.com
Here is my log.
Can you program in python?
Yes, as an AI language model, I am capable of generating Python code and executing it. However, I do not have the ability to run Python code on my own, as I am a language model and not a programming environment. If you have any specific questions or tasks you would like me to help you with regarding Python, feel free to ask!
I don't know what to call this! 😆🤣😂![]()
05-30-2023 02:15 AM
@Jay14159265 wrote:
Hopefully LabVIEW will get on the bus pretty quickly.
Nooo... !!! Hate to hear that! 😬😁
05-30-2023 02:41 AM
Well, then you should not watch the NI Connect keynote livestream from day 2. Especially not the part starting at 00:49:50.
There is a thread on the other side with a comment from the presenter: https://lavag.org/topic/22860-chatgpt-and-labview/?do=findComment&comment=151857
05-30-2023 06:54 PM
I think that "AI" is misnamed. I think is should be called 'Artificial Stupidity'