Episode 44: How I Use AI for this Show

Transcript

I've been running around calling AI a tool, and I saw something, a video that sort of changed my mind on this.
The point of the video was that AI should be treated as a collaborator or a co-worker, and not as a tool.
Now, I think perhaps we were coming at the conversation from slightly different angles.
I was coming at it from an angle of pushing back against this idea that AI is just going to replace all people.
And this person was coming at it from the angle of disagreeing with people calling it a tool.
So we were coming from opposite ends of things.
But generally speaking, the way that I've been using AI day-to-day, particularly generative AI, obviously, the way I've been using it day-to-day definitely is more in line with what you would consider a collaborator or a co-worker.
A collaborator is probably the better description than it is a tool.
So I have to be inclined to agree with him because this is definitely – I think his terminology is better than mine.
I'll try to link to the video.
I think I've got it set aside somewhere.
On that note, I wanted to use this show as sort of a quasi-practical example of how AI has augmented workflows.
And I just kind of wanted to outline what I use AI for as part of the production of this show.
So I want to start at the very top by saying I do not use AI for any kind of – like for scripting, for script – like to write a script for it.
I don't use AI, generally speaking, to come up with show topics.
Those are from me.
So those are the things that I do myself.
I also edit the show myself.
I have seen some really interesting AI audio editors where you can do some really cool stuff.
For instance, there's – I think it's Adobe that has one where you can – let's say you're editing and you don't like how something was said and you want to re-say it.
Instead of patching it and re-recording, which often leads to some really funky stuff, like voice differences and stuff, you can actually type in what you want it to say.
And like AI will – it'll generate a sound that sounds like your voice in the rest of the recording and patch it into that plot, that place.
So there is some cool stuff there.
I haven't used much of those.
I played around with one of them for a little while, but I didn't ever use it for this show.
I'm still using kind of an old school digital workstation.
I do the editing myself, that kind of stuff.
So topics, what I say on the show, and the editing is kind of me original stuff.
What I do use AI for – I do use AI in here.
I use it primarily on this show at least to take care of the stuff that I find exhausting to do.
You have to know a little bit, I don't know, about me, I suppose, to understand why some of these things would be exhausting.
But actually, you know, I use it for a couple things.
I use it for things that I find exhausting and that would like deflate my ability to continue to create the show on a regular basis.
I also use it to do some things that I would never want to do by hand.
So if you look on the website or in your, you know, in your whatever app you're using for podcasts, so listen to podcasts, you'll see certain things like, you know, every podcast or every episode has a title.
It has sort of a short description or a subtitle, depending on how you want to use it.
And then it has a longer description as well.
So I have podcasted before in my life, I have blogged before in my life, I've written articles for things.
For whatever reason, writing short descriptions of things – I'll have an easier time writing a, you know, a 5,000-word blog post, but then you ask me to summarize it into a four-sentence description, and I find that exhausting.
Like writing the actual description drives me crazy.
It just – I don't know.
Something about having to just do that, I don't know.
I just – I find it mentally draining.
So I do use AI to come up with the description you see.
When you see links attached, I have gone in and put them in myself, but the actual description itself, that's about a paragraph long, is generated by AI, typically.
I also give AI a crack at writing the subtitle, or the really short one-sentence description that you see, like sort of the subtitle of the episode.
Sometimes I use what it comes up with.
Sometimes I put my own in.
It's like 50-50.
Again, it's just one of those little tasks that I just find draining.
So I use AI there.
Those are a couple things that, again, it helps me keep going with the show knowing I don't have to write short descriptions about it, as silly as that sounds.
Again, like no issue coming up with topics, rambling on for 10-15 minutes about those topics, editing, which often takes about double the length of the episode.
So wherever the episode takes, it's like twice as long to edit that episode, producing it, putting the website up, all good with all that stuff.
Writing a four-sentence description about it drives me nuts.
Or drives me nuts isn't even right.
It just makes me tired.
The way I do that, and this is the other category of things that I use AI for in this show, things that I wouldn't do myself manually.
Once I'm done creating the show, I have a process that I run just on my local computer that listens to the audio and generates a transcript.
It actually generates two transcripts.
It generates one with time stamps and one that's just a big, raw blob of text.
That transcripting is only okay.
It's not great.
It misses stuff sometimes.
It loses inflections and stuff, and that's fine.
But at least it's something.
I'm trying, it's an attempt to make it a more accessible thing for listeners.
But I wouldn't want to sit down and type that out myself.
So AI has been a, this is the first podcast that I've ever been part of that actually has a transcript.
So, and I like that.
I think that's a great feature for people.
So the way I do this is I finish up editing the episode.
I export the thing into an MP3.
I run it through a process that generates the transcript from the audio file.
Then I take the transcript that was just generated and I drop that into the other AI thing that reads the whole transcript and then generates me about a four sentence short description.
And then I've got all my stuff to put up on the website.
In this case, I am treating AI much more like a tool than like a collaborator, for sure.
There are other things that I'm working on, you know, other projects and things where I definitely am more collaborative with AI.
In this workflow, it is a straight tool, but it is enabling me.
I don't want to say it's enabling me to do stuff I couldn't do.
What it's doing is taking the mental overhead away where, you know, this is episode, what, 44, I think, of this show.
I don't know if I would have even gotten through 44 of these if I had to write a short description for all of them.
I know that sounds dumb, but it's just that mental draining piece and you don't know, you know, how many pieces of straw you can get on the panel's back before you eventually get that last piece that just breaks it.
So, as much as I enjoy doing this show and I enjoy recording these episodes, knowing that I then have to, at the end of the whole thing, write a four or five sentence short description about what I just did, I just find that tiring.
And I think it's kind of like when businesses talk about focusing on their core competency or their core product.
To me, the core product is this audio file and I want to focus on that.
I want to focus on bringing the best show that I can to anyone who's listening.
And I think the number of people who actually sit and read the descriptions are probably pretty low in comparison to the people who listen to the audio.
And so, it's just not where I want to spend my time and my energy and my effort.
The other thing that I do use AI for through some of this and it does make it a little more collaborative of a process.
I will use things like ChatGPT sometimes to help look up some additional research or other examples of things that I have in a topic.
So, if I have a topic idea and I know I have one or two examples, but maybe I want three or four, I'll use it to kind of help flesh things out.
And a lot of times, I'll pick and choose things out of those examples that maybe I'm familiar with from before but had forgotten or didn't come to the top of my head or whatever.
So, anyway, the point of this really was just to A, talk a little bit about sort of a hybrid AI workflow, which has been working exceptionally well for me.
And if anyone's interested in some of these pieces like the transcripting or anything else, like, hit me up.
I'm happy to show you what I've been doing.
It's not that sophisticated, but it works really well.
So, part of it was that and part of it also was to level set around, like, yes, I do use AI as part of this show.
However, I'm not using AI for what I consider the core product of the show.
So, that's what I kind of wanted to say on all that.
Again, if you have any interest in seeing any of this stuff or learning how that works, happy to chat about it.
Just hit me up.
But for whatever it's worth, know that the topics and the content comes from me.
It does not come from AI, although I do use some here and there to augment.
Generally speaking, I believe that this is a kind of a human-to-human show, and I treat it as such.