Tom (00:00):

The work that I want to do is I want to find an answer to this problem and I want to solve it, and the work that I don't want to do is spend two hours digging through random blog posts and everything else, and so it's helping me do the work that I want to do.

Voice Over (00:15):

You've tapped into the Divine Spark Podcast from Paulsen. Join us as we explore a refreshingly human first approach to using AI in marketing, business development, and more in the ag, energy and rural sectors.

Sara (00:32):

Welcome to this episode of The Divine Spark. Today we're going to talk to a couple of the members of our fabulous web team here at Paulsen, and I think they're going to have a little bit different perspective than the other folks we've talked to so far just because their jobs are a little bit different and coding is definitely one of the things that's in maybe the bullseye of what AI can supposedly do to make our lives better. We'll see if that's their opinion about it, but I'll introduce you to Tom Kamnikar. Tom, introduce yourself.

Tom (01:02):

Yeah. Hi, I'm Tom Kamnikar. I'm the digital project manager here at Paulsen. I've been with Paulson now for 20 years, started in the web department and just kind of like a bad penny say I can't get rid of me, so I kind of head up web projects and oversee things.

Sara (01:22):

Isaac, please introduce yourself.

Issac (01:24):

Hi, my name's Isaac. I've been with Paulsen in contrast to Tom over here for almost a year now, which is kind of crazy because time has flown. So I am a front end web developer here at Paulsen and have been enjoying my time immensely.

Sara (01:44):

Very good. So we thought we'd start out this episode talking about an article that we all three read and probably many of you read or have least have had a chance to or will. It's an article by Sarah Tiller Wilken Field, and it's titled that Productivity is a Drag Work, it's Divine. So she's arguing that if we're going to do anything with ai, it should probably be the things that we like to do the least and gaining efficiencies or not. If AI ends up taking away the thing that we feel like we love doing or brings the most value, then that may not be the best use of AI for. I think she makes the case for moral reasons, but we can definitely discuss that as a group link to the article, be in the show notes. It's an excellent read of course, because I agreed with it, so I thought it was a great read, but it'll be a good topic to talk about. So let's dive in. Initial thoughts from either of you guys after having read the article?

Tom (02:51):

Well, just from an overall perspective, it was kind of interesting to think. She talks about almost the religious aspects of how work is part of your role in this world and to do the work and to do work is almost paying homage God and what he's created, and so to try and use AI to get out of all work is probably not the best idea, but all work does not have to be the bad work. There's a lot of things that we do in our day-to-Day lives at work that we actually love, like you said, and to use AI to replace that would just be a crime, so to speak.

Sara (03:40):

Crime against yourself.

Tom (03:42):

Yeah,

Sara (03:42):

Really to take I think the fun part out of what you do.

Issac (03:45):

I think she did a good job in contrasting the two kinds of work and she used some Hebrew words to describe the kinds of work, and I'm not going to try to pronounce them if you'd like to look them up in the actual article, feel free. But contrasting those kinds of work where you have that creative spirit, I suppose, that human drive to create and to channel that inner creativity that an AI just would not have the capability to do it, would not have that capability to channel that inner creativity because all it's ever learned on and all it ever can learn on is a derivative of other people's work,

Sara (04:37):

Right. I think as we discussed in the intro episode and along the way, even the work that we do in my humble opinion, is just a derivative of true creation of the actual creator, and I think she makes a good case for that. She and I are not of the same faith, but I loved how she laid that out as being really intrinsic and important, and I think I frankly, all work could fall into that category. It's just I wouldn't dream of taking my clothes down to the creek and beating them on a rock anymore. I think really thankful for washing machines and dishwashers and not riding a horse into the office every day is really a blessing. It's just getting used to those types of tools and I remember, I think I'm sure I've mentioned this before, I remember moving from key lines to the first computer, the first Apple computer I had and how much faster I could try out ideas. It just really opened up a lot more creativity just in the fact that everything wasn't so manually done. That was super helpful. So I am real sympathetic to her line of thinking on that. I don't know, what do you guys think about that? All work being actually

(05:56):

Maybe good?

Tom (05:57):

Yeah, I think actually doing the work for work's sake is good, but to your point, there are some things, efficiencies that we can do that take some of the stress out of it to get to the creative part of the actual work that you're trying to do. Being in the web world for 20 plus years, I'm a big believer in knowing what's come before what laid the groundwork of everything, the original HTML and how that all works in the original CSS that then translation into S-C-S-S-S and into spelt kit and JS and React and react and all those other things. Knowing the foundation of what those all were built on is important because if you're just typing, if you're just getting spit out of chat BT or perplexity, you can copy and paste it in, but if you don't know what it's doing, then it's not the best use. It's not useful to you at all.

Sara (07:05):

Do you think that falls into the category of thoughtful work? She pulls out that term really specifically. There were, what did she say, 39 types of work that were forbidden on the Jewish Sabbath building and destroying, riding and erasing, sowing and reaping intentional things that we do. So I guess that's maybe the case that are there unintentional things or less intentional things that we just have to do to get from A to B?

Issac (07:32):

I believe she did make that observation that the tasks which are considered, and I think she phrased it as tasks that allow you to set it and forget it are by definition not among the most serious violations of Shabbat. So I think those things that you have to do, you have to cook a meal, you have to put your clothes on in the morning, those sorts of things are inevitable, inevitable work that we all have to do regardless of your take on that specific topic.

Sara (08:08):

Yeah, I think she says soul sucking at one point, so the soul sucking avoid A, which is one of those Hebrew words that comprises many modern lives. I think I would make the case, this is maybe a tiny bit off track, but not that all work should all work creates value whether or not you enjoy doing it. Do the sanitation engineers enjoy their jobs? I don't know. I am insanely grateful that they come once a week and pick up things that I don't want to deal with anymore, but I think we should be elevating all types of work. So I don't think she's not making the case that we shouldn't do that, but if you are going to let ai, I guess the upshot of this for me was like, if you're going to let AI do part of your job, don't let it be the part you love,

Tom (08:54):

Right? Yes.

Sara (08:55):

Have it be the part you dislike. And I think some of it, AI is more suited to some things than others having just straight, flat out write something for you and a one shot passes not going to happen or you're writing is garbage and you're not paying attention, but it can do a really good job of taking a volume of work and comparing it down into something you can digest pretty easily. I think that's the case.

Issac (09:23):

I agree with that wholeheartedly and that's one of the things that I love the most about AI is you'll sometimes, especially as a web developer, you encounter new concepts all the time. New technology that you've never encountered before, things are always popping up. You could have gone to school for 20 years and when you come out of school, there's a new technology just on the horizon that you've never seen before, and a lot of the time I'll end up googling something and finding a related post, five related posts and the amount of time that it saves me just to enter a query into a prompt or enter a query and receive a response from GPT, which has trained on all of that technology and all of that text and receive a very concise reply is super valuable.

Sara (10:19):

Then we'll go back to earlier discussion about perplexity actually just kind of saving time just by virtue of the fact that it'll summarize.

Tom (10:27):

That's exact to Isaac's point, I always joke that I feel like the old gun slinger and that there's always young guns out there looking to take you out because every day there's some new JS thing out there. I always joke, you could just think of a word and I'm sure that framework exists, flubber js or something like that, and it's impossible to know them all, but the fact that you can just use perplexity. I'm a big fan of perplexity or chat GPT and type in what does this do or how do I do this real quick and get an answer is super helpful and for me the biggest use of it is error debugging, just typing in, I'm getting this error, how do I resolve it? And then it is doing the exact same work that I would've done a year ago, two years ago before it came along, typing it and going out and searching on the web.

(11:29):

It is just summarizing Stack Overflow and the docs and any other blog posts that someone might've typed up about it that encountered the same errors just summarizing it for me a whole lot quicker, which is kind of the point, the work that I want to do is I want to find an answer to this problem and I want to solve it, and the work that I don't want to do is spend two hours digging through random blog posts and everything else, and so it's helping me do the work that I want to do. Trial and error. Trial and error.

Sara (12:01):

Yeah. So as a coist, helpful as a code assist, are you using any other things that you might think of as code assists?

Tom (12:11):

Well, I don't know. It's not really AI powered, but we use autosuggest in our editors auto complete to help us type out functions and stuff. You could argue is AI powered

Issac (12:26):

And there are tools like copilot released by Microsoft and what was the other tool that we were talking about the other day? It's a newer code editor that also is embedded with AI technology that kind of goes in, looks at your code while you type it and we'll predict the next things that you're going to need to type cursor cursor, correct Cursor ai. Also, as a side note, you mentioned Flubber js earlier. Flubber is a JavaScript library to interpolate between 2D shapes. Well, there you go. There you go. You can edit that out.

Sara (13:10):

No, you

Tom (13:10):

Can't.

Sara (13:14):

Any last thoughts on that because we'll change gears here a little bit.

Issac (13:18):

I think that most people are using chat GPT or perplexity or these text-based interactive models, and you'll find that outside of those and outside of image generation, everything is built on top of those tools. So it might become more specialized in the prompt, in the background, might become more targeted towards the specific use case that you have, but a lot of the tools out there are built on those at this point.

Sara (13:54):

So how much better can you get just by being better at prompting? I find that if I actually

Issac (14:01):

Take

Sara (14:02):

A little bit of time and have a conversation, I can get a better outcome

Tom (14:07):

And that's one of the many reasons why I love perplexity is you can set up a profile of yourself and so if you're a paid subscriber, so my profile, it says I'm a 50-year-old web developer trying to stay up to date, specialize in craft CMS, but I know these languages and stuff, and it kind of tailors your answers to that already taking from what I understand, adds that to the prompt behind the scenes gives you a little more tailored to you answer.

Sara (14:38):

Very good. Well, let's talk a little bit about some stuff that we're working on internally. We have our own product that we started over a year ago and it's gone through some iterations and keep dialing it in just based on the fact that things keep changing and the ways to use it keep changing because our clients keep getting more comfortable with the idea that this is something that they can use as a legitimate tool. Tom, you want to, how and Isaac both, you want to give us a little bit of an idea what that is?

Tom (15:11):

Yeah, so it's called Paulsen ai and we're not reinventing the wheel as far as the large language model or anything that it's using. I believe it uses GPT-4 or whatever the latest engine is behind the scenes. What we're building is more of a customized front end to the client. We use vector databases to store large amounts of research and data relative to the client's industry so that they can say, I need to write a blog post about dairy masis,

Sara (15:48):

Mastitis,

Tom (15:48):

Mastitis. I always trouble pronouncing that in its vector database, a large collection of documents and stuff about that particular topic and then it'll come summarize it and then put it into more of a voice that the client would use. But that's just the start of the whole process because we do extensive behind the scenes prompt engineering. So when you type in, I want a blog post about this, we add a good chunk of extra information into the prompt before it's even get sent out, and one of the things that we then do is we're customizing what gets sent back to you. So right now in the current iteration of it, it'll give you an outline, okay, here's what I'm thinking your blog post should be about and you can edit the outline and say, no, I want this section above this one. Or Well, that's not relevant, leave that out.

(16:47):

And then it'll send that outline again off adding to the original prompt saying, use this outline. And so then the response you get back is a pretty customized blog post in your voice. It's not perfect. Then you can re-edit it again and change things because we're not encouraging you to just have AI generate a blog post all on your own, but it's a really good start to, it takes some of the work out. That part of the work that we were talking about coming up with, I know I want, this is my idea, but I got to do all the research and stuff. It takes that heavy lift out of it so that you can just focus on the message that's being delivered through your blog post. And that's just one example. It can do scripts for conferences or annual meetings or that type of thing. We've also talked about, what did you call it? Okay, boomer or Ask Boomer, the general idea, okay, boomer, if we could take all of Sarah's knowledge blog posts that she's written and articles and everything else and put it into this database. Then before I went to, if she was unavailable, I could just say, okay, Sarah, what should I do in this situation? And it would give me back a general idea of what Sarah would respond back.

Sara (18:11):

So an important component to that is really knowing how to extract the information out of the person to start with in order to put it into the vector database or databases so that you are prompting the correct information to get it to come back. And I think that's always ends up coming back to that human first component, like everything we talked about today works because we believe it's a human first endeavor to use artificial intelligence. If you want to use it well, if you don't care if you're creating mediocre stuff, then fine, it'll do that for you really fast. But to get something really good out of it, you've got to have your divine spark, which takes us back to that article and kind of her approach to that, which really was perfectly aligned. I think with the way that we can think about this in a way that's really, I think a healthy way to go forward as a culture and a society. We need to be taking care of our people and thinking about the impact that this has on folks and just reassuring people that this isn't replacing them, that literally is not replacing them. It's a tool that we can use and do better with. And that Pulse and AI is a direct reflection of that. What goes into it from the people that are behind it is what makes it work and what makes it special.

Tom (19:36):

Yeah, exactly. And talking about it in a podcast, you really can't get a sense of what it is. Ideally you should be calling up Paulsen and asking us for a demo of it because it is really cool. We're customizing it for every single client that we build it for and even down to the interface so that the demos we can do for you and show you what it can do is pretty cool.

Sara (20:01):

Absolutely

Tom (20:02):

Call someone today.

Sara (20:04):

Thanks. Long down. So I think from the client perspective, the way it feels to them when it works is that it very simply, I'm just entering in a title and it comes back with that outline and they don't need to know necessarily what goes on behind the curtain. We know because we're the ones working on it and creating it. And I think that's the beauty of it, is that you don't have to be a prompt engineer because we have pages and pages and pages of prompt engineering behind the scenes that are doing all the heavy lifting, as you said, Tom, but from the user's perspective, I just put in that title and it comes back with an outline. I edit that and I've got an article or a script or a blog post or whatever size we need, and it's in my tone of writing. And I think maybe we mentioned this and maybe we didn't, but we also include all of the writing that comes along with it. So not just peer reviewed articles or that type of data, but actually the writings of the writer, which really helped dial in that tone.

Tom (21:09):

And we also cite sources at the end of the article. It's up to the end user if they want to leave them in or not. But that's was the genesis of the whole thing is that we're making sure this is actual valid data that we're referencing and giving to the client,

Issac (21:29):

Not just hallucinations by the ai. Exactly, exactly.

Sara (21:32):

How about dairy mastitis?

Issac (21:33):

Yeah, mastitis

Sara (21:36):

Don't make that up. Cows don't appreciate it. So any last thoughts, guys on Human First in the Divine Spark or that article?

Tom (21:46):

No, just like we said, ai, it's a handy tool, but that's all it really is. Just another tool in the bucket to use and still got to do the actual work.

Issac (21:59):

Yep. Might be a very powerful tool though. It goes in along with others that I reach for regularly and hopefully they continue to make improvements on it, but I don't think that it replaces the human anytime soon. Even in areas where it has a lot of training data to work with coding and human language, which is what the entire internet is made of at the moment. It still can't surpass a human doing the same work.

Tom (22:32):

It's a power drill versus a screwdriver. They'll both get the screw in, but one makes it a whole lot easier.

Sara (22:38):

Absolutely. Very good. Well, thank you both. This was a great conversation. Appreciate your time and mental energy going into this and both of yours, divine Spark. So that's it for this episode. I hope you'll join us again for the Divine Spark. Thank you.

Voice Over (22:54):

Thanks for listening to The Divine Spark. Be sure to subscribe using your favorite podcast platform and visit us at Paulsen.agency with any questions or ideas for future episodes.