Do you want discussions limited to human interaction, without AI distraction?

[redbaron]

Kevin:: maybe we have invalid poll results due to my wrong edit of your title.

I believe users can change their vote to match the current phrasing.

 

[Neal]

I deleted the poll, it's irrelevant and people that voted will be inaccurate now that it changed.

 

[Rosy]

Honestly, as we do research on topics and things we plan and plan to do, that’s using AI. So how do you not use in some way?
So a response someone has to a question, may be an AI response.

Asking for a friend.

I don't use AI in my research...

 

[Kevin D Pem]

The uncomfortable truth!

AI will only get better,it has a place that will dominate the work place in the future.

This has to happen for the emerging financial world.

I don't fear the direction! But it should not replace, and can't replace human interaction. If it could the back orders for robot wifes would be beyond unbelievable.

Let's face it, human interaction is messy, not always honest, and sometimes wrong.

But it's the only reason I still show up here!

We don't need a poll to voice how we feel about this topic. I didn't think it would make it past the first hour!

Hopefully, however, it will lead to those that use AI for leverage the understanding that they are not only hurting their cause, but their reputation as humans.

 

[John&Andrea]

I don't use AI in my research...

A google search these days is basically AI. It seems.

 

[Rosy]

A google search these days is basically AI. It seems.

That's why I switched to Duck Duck Go.

 

[Rollin Ollen]

A google search these days is basically AI. It seems.

Sure a Google search probably seeks AI but If I'm looking for advice or information from a live user, I won't be searching AI will I?

 

[Boat Bum]

AI has been with us, in one form or another, for decades.

 

[TheLooks]

3 times recently I have used AI to solve a problem I knew the answer to..mostly. It took over 30 minutes each time to corner the AI and have it admit MULTIPLE times it was wrong....It will tell you what the prevailing wisdom is or what the manufacturer of a product wants you to hear, not necessarily the truth you are looking for.

 

[redbaron]

Here is a fun one, ask each model to count to 10 from 11.

Some will count to 20, others get it correct and go 11, 10.

 

[MemoriesByTheMile]

AI will present everything as fact and show complete confidence.
This is what I do to help alleviate risk of AI making up answers:
1) Always ask it to provide sources, sources need to be recognized as trustworthy. Provide examples of sources to trust and not trust. Instruct it can look at other sources but will need to highlight these as potential risk.
2) instruct it to point out areas of contention or areas where there is disagreement.
3) tell it to provide a confidence score between 0 and 100
4) ask it to provide summary reasoning for it’s overall response
5) after your iterations of questions, take the final result and feed it back to AI in a NEW conversation and ask it to check it for errors and provide the above 4 steps in its error checking

The above steps have been very effective for me in getting results that align with my manual checks/validations. This takes work, not a simple question.

 

[Underboss]

I tried AI for travel information, exits, rest area on my trip and it was helpful.

 

[turbopilot]

AI will present everything as fact and show complete confidence.
This is what I do to help alleviate risk of AI making up answers:
1) Always ask it to provide sources, sources need to be recognized as trustworthy. Provide examples of sources to trust and not trust. Instruct it can look at other sources but will need to highlight these as potential risk.
2) instruct it to point out areas of contention or areas where there is disagreement.
3) tell it to provide a confidence score between 0 and 100
4) ask it to provide summary reasoning for it’s overall response
5) after your iterations of questions, take the final result and feed it back to AI in a NEW conversation and ask it to check it for errors and provide the above 4 steps in its error checking

The above steps have been very effective for me in getting results that align with my manual checks/validations. This takes work, not a simple question.

Excellent points. Exactly how I use the AI tool. A human who is knowledgeable would not put up with that degree of scrutiny unless you paid them. I would also add ask the question in different ways. Also provide as much detail as possible in the question.

 

[redbaron]

One more thing that I find helpful is I ask AI to generate a prompt for the given results.

Reviewing that prompt will expose critical flaws in either the question or the parameters. Then I start a fresh session and reload the modified prompt and often get a different result.

Finally I feed that prompt to several Models and compare the results based on their strengths.

Using the same prompt really does help with the process