What do you use AI for? (a discussion, not a question)

The Emperor's Lounge
#107

I’m still new to AI and how to get the most out of it. What intrigues me the most is the prompts and how that can determine if you get the data you want or not.

I’m using PrivateGPT. I uploaded a csv of all my relevant journals. Still messing around with what usefulness I can get out of it. Was thinking about having it ingest the product descriptions/objectives and see if it cross reference and see how many of them I hit in certain topics. But that might be above this models capabilities.

Still super powerful stuff. Pretty amazing what technology can do nowadays. Just having a chat bot assume a different perspective and have that influence data retrieval.

1 Like
#108

My team currently uses it for marketing purposes.

Research, building avatars, creating first drafts, creatives etc.

It’s a force multiplier. With it, I can keep my team smaller. And that leads to them being paid a lot more for their work.

I believe this year (2023) was just an introduction to AI.

Next year we will start seeing more complex use cases for AI.

2 Likes
#109

I hope it keeps advancing in a direction that empowers people vs a “why do we even need to hire them?” type situation.

Have you seen any problematic things that pop up at times when your team uses it? I think the capabilities are amazing but I’m very interested in hearing about application and what shortcomings there are.

3 Likes
#110

With text, everything is just a first draft. There’s still manual “human” work involved.

You also need to understand the topic you are working on. So you can throw the crap out.

We did an analysis about what we could use the AI for in the business. The only thing we couldn’t implement it for yet (if we had the resources, of course), is fixing the machinery. And I believe that will also be possible in a few years.

2 Likes
#111

Lol. I got my first taste of this querying my journals spreadsheet the other day. The model was just making up random experiences I had with the wanted sub. I could definitely see some more incompetent businesses think they can cut out the spot checkers and totally rely on AI to increase profits and inevitably tank their company.

But yeah totally agree, there has to be knowledge or understanding going into it. It’s just high level computer use to me at this point. Crap in, crap out.

1 Like
#112

Yes, that is the point.

Designing the prompts that make it easier for the algorithm to make your life easier. Helping it to help you.

And you need to understand the difference between knowledge mastery and language/communication mastery. I think that before anything else, ChatGPT is a master communicator. That’s the precise reason that it improvises and simply confabulates false data.

It’s sort of like, “I’ll be wrong…but I’ll do it in style!”

If you provide it with the right prompts, guardrails, guidelines, data, and quality controls, it’s an excellent tool. Also, you have to be very sensitive to its working memory limitations. It’s currently designed to put 90% of its computing power on communications-related computation. It’s not so big on data storage, or adapting to new, untrained inputs.

(I’m just talking about the generic ChatGPT version.)

3 Likes
#113

It’s funny because I still communicate with it like a human being lol. Trying to gently recalibrate it towards what I’m looking for. Kind of like when one of my team members at my job is off on the wrong path and I try to drop hints as to how to get back on track and solve the problem.

1 Like
#114

I do as well; for ChatGPT I say good morning and goodnight, I say please when I ask it to do something, and I thank it when it does what I ask.

I’ve had people tell me I’m foolish for being polite to something that isn’t even human, and I tell them point blank I’d rather be foolish than be such an a$$hole that being rude is my default option.

4 Likes
#115

That’s funny. I’ve noticed something like this happening organically as well.

Not good morning or good night, in my case; but definitely statements of appreciation.

It just feels right for some reason. I’m guessing it’s probably for reasons similar to what you’ve expressed.

One thing’s for sure, I much prefer working with the versions that have much larger token capacity. Like ChatGPT-32K and so on.

I use the feature a lot where you can attach a pdf and have it use that text as a source of information or a template/model for the style of thinking you’re looking for in its responses.

For a while now, I’ve also been using it to get very insightful I Ching interpretations. I provide my original query as well as the I Ching’s original text in archaic Classical Chinese. ChatGPT has no problem translating and applying Chinese as it was written 3000 to 2400 years ago. That’s incredible.

2 Likes
#116

One thing I’ve been using that feature for is in the form of a custom GPT for my book; I wrote 9 chapters (~35k words) but then decided to redo pretty much all of it, by giving it access to the original work, it’s helping me recreate plotlines and characters that retain much of the original essence, just reimagined… no way I’d have been able to remember all of those details without endlessly scrolling back and forth through 100+ pages of text.

1 Like
#117

Awesome.

Excellent that you’re writing that.

2 Likes
#118

It’s a fictional story exploring some of my fears about the darker side of what will be possible if authoritarian forces use AI for absolute control… it remains to be seen what thinking about this is doing to my state of mind long-term though. haha.

1 Like
#119

Well i guess if he is happy with that we cant complain :slight_smile:

1 Like
#120

Imagine if AI was suddenly taken away from everyone overnight.

#121

I use AI to both write emails and write skeleton code in bash or powershell.

1 Like
#122

I think this is pretty much a certainty.

I was nervous when it looked like the OpenAI team might all go over to Microsoft; I had no idea what would happen with GPT-4 in that case…

1 Like
#123

One of the clients i use to work with one of their policies was to never use chatgpt to create code as it might create something malicious.

#124

Dang, if only humans were capable of reviewing code before running it… :wink:

The CEO of my company doesn’t understand AI whatsoever, therefore it must be evil… :roll_eyes:

2 Likes
#125

LOL yup no logic in what the policy stipulates.

#126

It’s been my experience that that is the defining concept behind almost all bureaucracy lol.