Using AI day to day

[10-Ring]

I've played with AI a bit and I'm actually starting to find it useful as opposed to a gimmick. Amazingly, just a couple of months ago I asked AI to write a program for work. It spit out a very generic program that would have required a lot of tweaking to be useful. I asked the same of it yesterday and what it produced would really only need vital things like the company name added. The technology is developing so fast that just two months ago it was rather useless, while I could complete what would normally be 4 hours of work in 20 minutes today. 

For those who use Excel, I used Chat GPT today to modify a very robust formula. The formula in question recognized and performed a function with the character "x", I needed it to perform the same function for "o" as it did for "x". Although it seems that would be simple to modify manually, it was not. I entered the formula into Chat GPT. Chat GPT accurately told me what the formula would do, then I simply instructed it to "modify the formula so that "o" yields the same result as "x". It took all of 3 seconds to return a formula that worked. I would have likely spent most of my workday tomorrow trying (probably unsuccessfully) to modify that formula. 

AI is quickly becoming part of my arsenal that I use to be more efficient and portray the illusion that I'm intelligent. Don't be scared of this stuff. I can certainly see how it could potentially be harmful. But it doesn't have to be. Use it for good and make your life easier. 

[MacGyver]

I use a couple private models and one public one.

Truthfully with the release of Claude Sonnet 3.6 in November - I went from my previous worldview of “this is kind of neat. I wonder what I could use this for?” to “oh… I should be using this a lot more.”

I have a lot of thoughts. I’ll save them for another post.

For now, I think it is sufficient to say that the future is already here - it’s just really unevenly distributed.  

[Erich]

agree,  AI ‘s ability to cut to the chase and deliver usable results has noticeably improved for the tech sector.

I was having some challenges connecting a third party client to Google BigQuery and it was helpful though it did not give a specific solution. The footnoted references saved a couple hours of scrolling thru search engine results.

I love that it can save big time on emails. I find work emails a pain point. PC concise emails without significant rewriting are a challenge. when I remember, I give highlights and voila! Just have to remember to use it up front versus realizing 20 minutes into writing one and debating word choices.

It probably would  have made this post better too 

[gregintenn]

This stuff sure seems to be moving fast!

[Luckyforward]

AI has helped me to organize treatment planning for my patients much faster.  I thought it wouldn't be useful at all, but I was wrong.  

[Defender]

My daughter uses it to compose case notes for her patients.

[BigK]

I can't deny having some reservations about the dumbing down of human intelligence as AI replaces them. I suppose it's no different that automating certain tasks on assembly lines though. The main difference I see is that programming AI can contain biases and ulterior motives in certain applications.

[gregintenn]

9 minutes ago, BigK said:

I can't deny having some reservations about the dumbing down of human intelligence as AI replaces them. I suppose it's no different that automating certain tasks on assembly lines though. The main difference I see is that programming AI can contain biases and ulterior motives in certain applications.

We’ve been skiing down the dumbing down slope for quite a while now. 

[MacGyver]

25 minutes ago, BigK said:

I can't deny having some reservations about the dumbing down of human intelligence as AI replaces them. I suppose it's no different that automating certain tasks on assembly lines though. The main difference I see is that programming AI can contain biases and ulterior motives in certain applications.

There are certainly cases where that will be the case.  Think of all the teachers who are thinking students will just use it to cheat on essays.

If someone wants to do that they can - and to be truthful - society already kind of caters to that anyway.  It’s not like we need AI’s help.

Where I find it insanely useful is at the other end of the spectrum - using AI to work alongside your particular expertise to create things that wouldn’t have been possible otherwise.

I’ll give a personal example.  I’ve worked alongside two private models for about 200 hours in the last month.  I have pretty deep subject matter expertise in some things - and am a mediocre programmer. The people who taught me to code were first generation computer scientists who thought computers might be useful in a lab one day.  Their worldview just didn’t foresee a day where increasing the GDP of the internet would be a thing.  I know what to do - but syntax can trip me up for weeks.  In the last month, working alongside AI as a sort of pair programmer, I’m confident in saying the work I’ve done would’ve reasonably taken three developers and a designer six months to accomplish.

It has literally messed with my sleep schedule as I’m working to maximize rate limits so I can ship faster.

While some people don’t even realize it’s a thing, and some people assume it’s just a gimmick, the doers of the world are going to be operating at a *much* faster pace.

To be honest, I don’t know that I’ve seen more opportunity in my lifetime. 

[RED333]

Can AI help me buy ammo at a better price? Or what find firearms at a better price?

I know the answer is yes, but yeah, to be honest with you AI kind of scares me.

I fix things with my hands, that is my job. I use screwdrivers and wrenches, I use volt meters and amp clamps. That's what I do for a living. I see no way AI could help me with that.

Edited yesterday at 04:44 PM by RED333
Because I can

[MacGyver]

12 minutes ago, RED333 said:

Can AI help me buy ammo at a better price? Or what find firearms at a better price?

I know the answer is yes, but yeah, to be honest with you AI kind of scares me.

I fix things with my hands, that is my job. I use screwdrivers and wrenches, I use volt meters and amp clamps. That's what I do for a living. I see no way AI could help me with that.

I actually saw a pretty cool refrigeration training setup the other day where the AI could "break" the system a hundred different ways and then work techs through training modules where it teaches them how to diagnose the problem methodically and then helps them learn what wrenches need turning.  

There will be voices that say, "oh, they're just dumbing techs down - they're just inputting what they're seeing on the gauges and being told what to do next."  But, that already exists in a host of ways.  At the same time there will be people who realize that they can bring new talent up to speed faster - and that will be really valuable.

The module I saw was pretty interesting.  It didn't just give them the next step - it walked them through diagnosing the problem in a methodical way.  It wasn't just do this next.  It was - "you're experiencing these conditions?" Here are the six things that could contribute to that - and let's work through diagnosing it further.  It had a lot of "why are you seeing this" that brought the classroom to a hands on lab if you will.

 

[TripleGGG]

I use AI to summarize client coverages.  I used to do this manually and thought I was fairly quick with it.  Now I can place all of the policies my client has into AI and ask for a summary of coverages.  Amazing what it returns and how fast.  I always verify the information.  But I do that even when I had secretaries and assistants.  

Another great feature of AI I have found is with operating manuals or pdf documents/books.  I love amateur radio but remembering every process to get to certain settings or how to change something can be hard especially with some of the latest radios, amplifiers, etc.  I upload manuals, books, etc. for everything I own if I can get it in a pdf.  If I need to know how to do something or try to remember something I read in a book or pdf all I do is ask.  It finds it and gives me the answer almost immediately.  

I can't tell you how much history and info I can pull on a firearm and ammo just by asking AI to search all of the documents I have uploaded.

AI is just another very useful tool to me.  Just like anything else, whether it is used for good or bad, is controlled by the user.  

Edited yesterday at 06:00 PM by TripleGGG

[NoBanStan]

57 minutes ago, TripleGGG said:

Just like anything else, whether it is used for good or bad, is controlled by the user.  

Agreed. Right now it's an infant. In fact it's more machine learning than AI. Effectively a better Google search. AGI is when it will start getting interesting. Especially if it links up with quantum. But, we're likely a ways away from that being mainstream.

[Hozzie]

We are in the microprocessor days of AI.  We can't even imagine where this will be even 10 years from now.  Think of where we are from the Apollo moon mission to now.   Now increase that by a factor of 100 and it may still be short of the potential.

 

[willki]

I have tried getting AI to write code but so far the code it has written is so bad that I spend more time fixing it than I would have spent writing it from scratch.  However, when I wonder "how do I do this in code?", AI has been useful showing me things in 5-10 minutes that I would have spent hours searching for online.  Essentially a much better search engine with relevant examples.

I have found that as long as you can ask it for simple things it's fine.  If you start giving it detailed instructions it gets confused and the output deteriorates quickly as complexity increases. 

It is getting better very quickly though and I see amazing and also terrifying potential in AI.

In the absolute best case scenario we need to find jobs for about 75% of people or deal with people only working 10 hours per week.  Because AI if it keeps progressing will be able to do most of our jobs if not all if you include advances in robotics in 10 years or so.

[MacGyver]

I am very much starting to think about this in terms of “how do we use this for human flourishing“ as opposed to just “how do we make more money off this?”

it very much feels like the early days of the microprocessor or the early days of the internet.  I pay a pretty decent chunk of money for some of the models I use and still hit rate limits at least three times a day.  It very much feels like the old days waiting to get processing time on a mainframe.  Or like waiting for data to come in over a modem. 

[Erich]

I’ve had the same kind of experience as you @willki. To get more out of it, you have to interact with interim result sets and direct it to get best results.

I’m unsure how someone can get a full program out it as referenced. Programs tend to be very environment and data model specific normally…. but likely down to the interaction and information shared with it. I am limited as i can’t share data models. 

After years of working in SQL Server environments and transitioned to  one with DB2 and straight ANSI sql, I love when it can translate to DB2 specific things with its set of misnomers and interesting attempts to emulate a SQL Svr method. 

I especially love throwing it a showy regex function that one of the IBM fans is super proud of and it tells me what it actually does with an example without having to look up its nonsensical control characters.

 Huge time saver. 

As far as the idea subs have of dumbing down. it’s no different than an encyclopedia on concept. The store of all human knowledge, but easier. Nothing wrong with readily accessible. 

My wish is it wasn’t censored. 

 

[MacGyver]

12 minutes ago, Erich said:

I’ve had the same kind of experience as you @willki. To get more out of it, you have to interact with interim result sets and direct it to get best results.

I’m unsure how someone can get a full program out it as referenced. Programs tend to be very environment and data model specific normally…. but likely down to the interaction and information shared with it. I am limited as i can’t share data models. 

After years of working in SQL Server environments and transitioned to  one with DB2 and straight ANSI sql, I love when it can translate to DB2 specific things with its set of misnomers and interesting attempts to emulate a SQL Svr method. 

I especially love throwing it a showy regex function that one of the IBM fans is super proud of and it tells me what it actually does with an example without having to look up its nonsensical control characters.

 Huge time saver. 

As far as the idea subs have of dumbing down. it’s no different than an encyclopedia on concept. The store of all human knowledge, but easier. Nothing wrong with readily accessible. 

My wish is it wasn’t censored. 

 

I think one of the best possible cases is to think of it sort of like a pair programmer who knows the API docs backwards and forwards.

It’s unlikely to be able to do anything well by itself.

But, sitting beside you it can really compliment your expertise.  Or in my case, it can be like, “hey idiot.  Don’t do it like that.”

[MacGyver]

17 minutes ago, Erich said:

 

I especially love throwing it a showy regex function that one of the IBM fans is super proud of and it tells me what it actually does with an example without having to look up its nonsensical control characters.

 

As far as I’m concerned, if it never does anything more than write better regex, it will have been worth it. 

[willki]

My concern is once AI is actually good enough to drive vehicles better and safer than humans can, where will drivers find work?

Once AI can code better than coders can, where will coders find work?

and on and on with nearly all professions.  

How will society change over the next 20 years as we can keep society going with fewer and fewer people needed to work?

Some people see that as a Star Trek like utopia future where everyone benefits from the gains in efficiency.  I am worried that it won't work out quite that well.

Insert whatever dystopian scenario fears you have here and they just might fit.

Admittedly, this may not be the right thread for these concerns.  This thread is about day-to-day.

Edited 22 hours ago by willki

[gregintenn]

It’ll no doubt be like the internet in that it will provide unlimited opportunity for those of us who chose to use it. It’ll also provide mind numbing crap like the current internet provides porn, videos of guys hitting themselves in the groin, and chickens wearing blue jeans.

 

I’m more interested in finding the next big start up company to invest in that will be taking AI forward.

[TripleGGG]

1 minute ago, gregintenn said:

It’ll no doubt be like the internet in that it will provide unlimited opportunity for those of us who chose to use it. It’ll also provide mind numbing crap like the current internet provides porn, videos of guys hitting themselves in the groin, and chickens wearing blue jeans.

 

I’m more interested in finding the next big start up company to invest in that will be taking AI forward.

I am in a constant state of research for this myself.  I have my eye on several but still too early to tell.

[willki]

One area that I think AI could really help with is giving people help handling stress.  Over the years I have had many people confide in me about things that are bothering them and I am not any better qualified to help them than they are themselves.  

Americans spend about $280 billion per year on mental health issues and that's less than 1/2 of the people that need the help.  More people need help and can't afford it or can't get help for other reasons than people that actually receive any help.

An AI could be there on the phone 24/7 and also be 100% confidential and even anonymous as well.  The AI could also detect voice stress.  The app could also show progress over time like "24% reduction in utilization of these negative words in the last 6 months". "32% increase in the use of these more positive words in the last 6 months".  "40% reduction in voice stress detected...."

It could be convenient for those with long commutes to deal with the issues of the day before getting home to the family.  Plus giving positive suggestions for dealing with work issues in a more constructive way.

I had a coworker commit suicide several years ago.  I have always wondered if I could have helped.  I had to "hack" into his password protected computer at his wife's request to get his financial information and found his suicide note on the desktop.  AI could have helped him by providing someone to listen to his concerns 24/7 and provide alternatives.

Plus, if people are already spending $280 Billion per year and 1/2 still aren't getting help there is some $ to be made there.  I don't have the $ to build that system, train it, test it, market it, .......But the company that does will make billions immediately.

Marketing would not be too difficult on a system like this though as there are tons of first responders and current or former service members that could use help.  Giving special groups like that free access would spread the word very quickly if it helped people.  

Edited 20 hours ago by willki