Posted on 07/01/2024 10:08:20 AM PDT by dware
Now that, my FRiend, is a stunning example of the power of AI/ChatGPT. Far from perfect, but it’s going to revolutionize so many different sectors within the tech industry.
As I said upthread, I use it all the time and am just now getting the hang of how to ask the right questions.
It's like giving instructions to a human assistant except that it can handle very long and complex instructions if they are absolutely precise.
The great thing about AI is that it never loses patience or gets discouraged. I regularly find myself telling ChatGPT that it's a genius (I forget it's not a person).
BTW, I am a subscriber to ChatGPT Pro.
Same!
I’ve been having fun exercising ChatGPT 4o. It is programmed with many biases, some it can’t get past. It will apologize once a lie or incorrect answer is given. One frustrating observation is that I can ask a question, several, get an answer that it agrees with, then open a new session, ask again and get a different answer. It doesn’t learn. Overall, it is like a 5 year old where everything is black and white and nothing is final or remembered.
Whatever. I used ChatGPT a couple of days ago. Needed a BASIC programming language script. Gave it w website address that showed the subset of the language that could be used and it kept using statements outside of that subset. Went round and round with it and at one point, it gave me back MicroPython to be used on a Raspberry Pico. wtf?
How can an AI computer, trained with the internet as a dataset, not know every programming language and also have poor reading comprehension?
I did notice it picked up a profile pic that I used once or twice several years ago and I’m sure it did so by way of my email address, so I wonder if it’s been programmed to offer junk answers to “far-right” people like me.
Interesting, Are you using chatgpt to day trade or for longer term market analysis?
I’ll like to hear more about this, dm if you’re interested in sharing. Thanks
I use AI to analyze/locate predictable patterns (to the extent they exist) that improve my odds of making a profit buying Nvidia calls when it takes a steep-ish downturn right after hitting a new high.
To do that, I used ChatGPT to help me write the scripts to download various price intervals (minute-to-minute, 5,10,15-minute, 1 hour, and daily) from the ThinkorSwim trading platform. I download them into Excel and use ChatGPT to help me write VBA code to find and analyze predictable behavior.
If you trade options, you know that finding predictability isn't easy.;-)
What I have been most successful at so far is assigning odds to Nvidia downturns and recoveries (which, as said, almost always occur after a new high is reached) over specific periods (days, not minute-to-minute, etc.) that so far have help me make money on OTM calls.
The stock predictably recovers within a week or 10 days MOST of the time. Of course, there are many times when the stock turns down sharply after hitting a new high and then wallows for weeks on end.
As to using ChatGPT, it is incredibly good at writing code if you give it VERY specific and precise instructions. It can also provide good input as to how to develop a system to perform statistic analyses. My learning curve with ChatGPT has been incredibly steep because it is only as good as my questions. I sometimes spend hours thinking about and writing the questions. And then more hours testing those questions with ChatGPT.
No.
Or at least not in my experience.
You have to learn to make your questions very narrow and precise to get good answers from ChatGPT. The broader and more general the questions, the worse the answers.
Imagine that you want a piece of software developed by a programmer. You plan to write out the logic or a script for software and then give those instructions to a programmer who will then turn your instructions into a finished product.
If you don't provide perfect instructions to the programmer, the process won't work. The programmer can't read your mind.
Or imagine that you are a lawyer with a brilliant paralegal/assistant. You ask your assistant to develop your defense strategy but you don't share with that assistant the minute details of the case or your thinking.
You'll get back garbage.
As I said upthread, ChatGPT is like a brilliant but inexperienced assistant. You have to help by giving it very careful instructions. If it fails, that's on you.
If you ask dumb questions or give it dumb instructions, you get back dumb answers.
When I started out using ChatGPT, I asked very dumb questions. The more I stuck with it, the more I learned how to manipulate it to get what I wanted.
Think of it like a machine. Yelling at it won't work.;-) You have to learn to ride it like a motorcycle or surfboard.
It you don't use it correctly, it will just spit out garbage.
Yes, how you ask the question determines the result. Most of the time. I have spent quite a few sessions trying out questions on climate change and it seems that written in bias limits how it answers, it goes back to the approved answer even if you work it through admitting error on individual items, put together it basic says ‘despite that...’
BTW, AI operates quantitatively in that it assigns quantitative values to language and data. Ask a qualitative question and you'll get a poor response.
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.