Show notes
[MUSIC]
>> Hey all, Marty here and we are back with another Unmead episode.
Today, we're going to be talking all about Chat GPT.
With me today, I've got Michael.
Say hello, Michael.
>> Hello.
>> I also have Lynn with us today and say hello, Lynn.
>> Hello, everybody.
>> Why don't we start this out with what exactly
is chat GPT for those who don't know.
So Michael, why don't you kick us off?
Yeah, so I will share it the best way
that it's been explained to me
and then hand it over to Lynn
to fill in anything she wants to with it.
The best way that chat GPT or the technology
that is text generation is explained to me
is it is auto-correct on steroids.
Instead of just tapping that middle word
in the middle three words that are above your keyboard,
if you didn't know if you start typing and you leave it on
above your keyboard on iOS and Android,
three words show up that you can tap to auto fill
and your phone can auto guess what the next word's gonna be.
Well, this is that on steroids because A,
the prediction of the next word actually makes sense
based on the context of the information
that you prompt the AI with,
and B, it then will fill it in
and automatically form sentences
because it's aware of proper English grammar,
and conscious of it a lot more than I am.
And that's the best way that I have learned
to explain what chat GPT is.
What about you, Lynn?
All right, so to me, chat GPT is a brain helper.
It's a thought helper, it's a creativity helper.
It thinks very fast.
We don't like to typically say that AI thinks,
but for the purposes of understanding what it is,
have you ever sort of stopped and think,
gosh, if I could just tell this computer what I want,
if I could just tell this word processor what I want to say
and then have it say it for me.
Work processor, write me a paper.
Exactly, exactly.
Work processor, write me a paper on,
insert subject here.
>> Yes. When we think about AI, artificial intelligence,
basically what we're talking about is a tool that helps us.
I like to say, some people would say that it takes over our thinking.
In my opinion, from my use of it,
it helps my creativity, it boosts it.
It helps me to think in different ways.
A lot of times, I will ask it a question,
and it will come back with an answer,
and it was like, "Wow, I didn't even think that.
I wasn't even thinking from that angle."
So it really does help,
I think, to broaden your horizons when you're doing your work.
One time, I had to write a customer service letter,
and I was really angry,
and I wrote all this stuff into the chat box.
Then I asked GPT to write a professional-sounding letter,
and it came back with a letter that would be written by a person that wasn't crazy angry.
[LAUGHTER]
It did a beautiful, beautiful job of coming back,
and here is a letter that you could actually send to a customer service.
>> There's a lot of processes that you can use for chat GPT.
I mean, really anything, right?
I mean, you can use it for a paper
you may be writing on a certain topic.
You can maybe do it on a marketing project you're doing
if you're trying to market something.
I mean, there's so many different use cases for it.
It's endless.
So Lynn, what would be like a workflow?
You would start out with some kind of an idea,
you're writing something.
How would you do your workflow?
You would start out and then go from there.
I think what's really important and Michael and I are going to probably emphasize this
over and over is something called prompting.
And prompting is the philosophical question, what's more important?
Having the right answers or asking the right questions?
And when it comes to AI, when it comes to chat GPT and others like it, the important
thing is how you ask the question and you will get better as you use the system.
And so, for example, when I was doing a presentation for bits, which was, I messed up because I
was just had just gotten a braille display and I wasn't sure what the heck I was doing,
but I had asked it to write a public service announcement about why or how people should
behave around guide dogs. And it wrote basically an introduction, and several bullet points
explaining why you should do this, that, or the other thing, and then an ending. So it
was like, it did that in five seconds. Something that might have taken me a long, long time
to do, especially since I don't have a guide dog and I really don't really know a whole
lot about guide dogs, but it cuts the crap. I mean, I like to say that, but it really
gets to the point of what you're trying to do.
And let's take a step back for one second. Maybe Michael, you want to explain what prompting
is.
Yeah, so it's interesting. I've fallen into the trap of having a conversation with the
AI when I actually just wanted to go experiment.
So firstly, if you are like, I've heard of this, you know, at its time,
I want to go play with this after I hear this podcast from Unmute present.
Uh, I really understand it a little more.
Just, just be aware that you might suck a little bit more of your
time than you expect to.
Um, so be conscious of that.
So a prompt is your information that you share with the AI itself.
For example, I sent a prompt and said, who is Michael Babcock?
And that prompt, my question to it is, is what it came back with a response of.
And it responded and told me as an AI language model, I'm unable to search
the web to tell you who Michael Babcock is.
I followed up with that and said, you are a AI assistant to Michael Babcock,
a founder of a company called PayOnMedia LLC.
We publish podcasts and help small business owners share their voice with the world.
And then I said, ask me some questions about, or ask me some questions that I can answer
to help you better assist me.
And then I hit enter and it asked me, it started asking me questions.
So now I'm going to keep this specific chat open and answer these questions and then
start asking the AI model itself for assistance that I might need.
For example, going back to what Lynn was talking about, can you provide me a
response to a customer service email?
The customer is upset because the product was not provided to them in a timely
manner, explain to the customer that we are sympathetic with the problems that
the customer is having with delivery.
And we are going to resolve their issue to the best of our ability.
Here's the tracking number and provide them with this USPS tracking number and,
and giving them that information.
That's a lot of rambly information that the customer is going to be like,
that's not professional, but Chat GPT wrote a professional email.
And the only way that I have found to be able to really modify the way that the
email comes out, especially the tone, that's important.
And that's what will help you stand out from what other people are doing with
Chat GPT is that customer service email.
I said that I said, write me a customer service email.
I started that prompt with write me a customer service email in a polite, but
professional and casual tone, and then gave it the information
that I needed to convey in that email. And that's how it came
up with the email that I was happy with. And then the last
thing, and then I'll stop talking and hand it to Marty
Orlin is, I, I think this is a way you can get into building
some interesting things. And when you start, I've been adamant
about telling people to start from the end and work backwards.
When you start from the end and work backwards with automation,
and you run into a place where you need a computer program
to automatically rename a file.
So that way it can match a certain file type.
So you can move that file.
Think about how open AI can help you.
And I hear, and I know how I'm gonna transition this,
that chat GPT 4 is better with writing code
and stuff like that.
So Lynn, can you, what is chat GPT 3.5 and 4?
Right, 3.5 is the, it's sort of an incremental development between 3 and 4.
Right now, 4 is, I guess they consider it under training or that they want to make sure
that it's good enough before they, I guess, release it to the general public.
I have access to it because I have GPT Plus, which I pay $20 a month for, but I am addicted.
So is chat GPT 4, is that only behind a paywall and everyone else gets 3.5 or how would someone
want to utilize 4 opposed to 3.5 if that's the newer out of the current technologies
that are out there?
This is a good question and if you don't have a paid account, I'm not sure.
There's a combo box on mine because I have GPT Plus and it lets me choose which model
I want to use.
3.5 is only available.
You have 3.5 and 3.5 Turbo for the free customers.
Some paid customers have access to four and some people can get access to four on the
API.
So if you're using an app that I can talk more about in the near future, then you can
leverage that.
And if you want to try the difference in chat GPT 3.5 and four, and you don't want to pay
anything, Bing is using a version of chat GPT four.
So if you ask the same question to both chat GPT and Bing, you can see the difference in
responses. So we should explain to people here, you know, how this works because you have regular
chat GPT in a web browser, you have other apps that you can install on your computer, you have
a big other technology companies such as Google and Microsoft that are implementing it into their
software. And I think with all the headlines, people are going what you know, it's a little bit
kind of hard to follow. So we have it in the browser. So you could go to chat GPT, and you
could just use it strictly in the browser. And then we have, for example, if you're a Mac user,
you can use what's called the Mac GPT. And Michael, you want to give a little primer on that?
Do I ever? So Lynn, are you on the Mac? I forget you're on the Mac, right?
I am. Yes.
Yes.
Yes. So I'm using Mac GPT. This does require a little bit more setup, but you had to go get
API key and there's some good directions that we can point people to on how to
build a to get this and it's very accessible to do but you put that API
key in and then you have an app that runs on your Mac and you open it up and
there's a prompt right there a edit field that you can type it and you hit
enter and then if you VO left arrow once there's a list there that you can
interact with and go and see the response that chat GPT gave you. Pass the
response is a copy button. So you can copy the response from
chat GPT, which is amazing. And what's even more awesome about
this is if you are using chat GPT for I believe you can use GPT
for with the Mac app, you'll have to try it by following the
sign in process. But if I'm in a tool like drafts or text edit,
and I type slash GPT space and then my prompt followed by shift
enter, GPT will fill in the rest of, like it fills in the answer of the prompt right there.
So I don't even have to pop you and paste it does inline GPT filling. And then I go delete
the prompt at the top. By default, the keystroke is plus GPT space. And so that's Mac GPT in a
nutshell. And it is it's awesome. Right. And they do have APIs, right for developers. And there was
was a little shortcut called Super Siri. You would say Super Siri and it would actually
give you a chat GPT prompt. So it's pretty awesome.
That's pretty cool. In the same sense, there was I believe Federico Petici from App Stories
had a shortcut that he built that he put out. Michael, I think you played with that a little
bit, right?
Not a little bit. I'm actually racking up a slight bill on my API usage. I'm at 27 cents right now.
Better be careful. It might break the bank. Yeah, I use SGPT quite regularly. That's the article
I recommend people going to to follow to learn how to set up your API key. So you invoke Siri,
you say SGPT. It says, "What would you like to ask GPT?" And it gives you a hint of some ideas.
What's really nice is if you give it access to your clipboard and stuff, for example,
you can copy a URL on your clipboard on the Mac, send it to SGPT and ask it for a summary
of the article, which can be really helpful. And then SGPT also has the ability to access your
calendar. So if you tell it, analyze my calendar and tell me how many meetings I have, it can
process that information too. Okay. I could talk a little bit about Google Bard. Okay.
>> If you want. So, Bard has just opened,
Google has just opened up their AI to the public,
and it's in beta.
I am on a list, David Goldfield's list.
I don't know if anybody knows about his list,
but it's for the blind,
but he does a lot of stuff with AI on there.
It turned out that Google has released it.
It's, if you guys hear something, something about Bard,
just don't get confused.
It's not the same thing as the,
as the library service reading material.
I heard they're actually gonna be changing the name
actually.
Okay, I hope so.
Cause it's kinda, you know.
It's confusing.
Chat GPT is kind of a yucky name too.
Let's get a better name guys, you know,
type it in and see what it says.
The problem that Google's gonna have is that,
you know, Google sort of thrives,
lives and dies by advertising.
So right now it's free.
Right now there's no advertising in it.
But most of the reviews that I've seen,
the tech reviews say that it really has some work to be done.
What I think is unfortunate about BARD is that with chat GPT,
when it gives you an output of a prompt that you give it,
it will have a thumbs up and a thumbs down.
But if you hit thumbs up,
it's going to ask you what you liked about it.
You hit thumbs down, it's going to say, "What happened?
Why didn't this go right?"
Unfortunately, on a board,
you just get a thumbs up and thumbs down.
I wish they would have a prompt that says,
"Here's the problem with this output."
I asked about French fries and I got something else.
>> Onion rings.
>> Onion rings, yeah. So I think that having that response,
being able to tell the model and
the NChat GPT or OpenAI that this prompts,
this didn't work, something,
it gave me an answer that was not right.
In this way, I think people can train it more because they can really give more feedback
about the answers that it gets.
One thing that I have a lot of fun doing with AI is writing stories.
If you are somebody who is creative, you can type in, you'd say, "Write me a creepy story
about a vacuum cleaner."
It will write you a creepy story about it later.
Some of them are really fun, but it can be just fun just doing that.
Say you have two sides of a political debate, you can ask it to be one side and ask you
questions.
So you're kind of having a debate with it and it's presenting the side that you want.
And then you can come back with your own arguments.
So it's sort of, it is interactive.
It's very interactive.
I have used it for therapy.
Wow.
You know, I--
Yeah, I would have never even thought about doing that.
Sure, yeah, I've used it.
Like, I am so mad at my sister because da da da da da.
Now, obviously you don't wanna give it sensitive info, right?
You want to sort of be aware and also it gets its facts wrong. They all do. So if you know
I wouldn't like use it to do an appendectomy on somebody like
You want to make sure I'm right now it does, you know, they even give you a warning that says that it may give
Factually inaccurate info so I
I would say if you're going to use it to do your graduate thesis or whatever,
you might want to really check your facts.
Yeah. So Michael, do you know how it's going to be implemented with Microsoft and how they're
going to be using it?
No, but it's my understanding that it's going to be a pretty streamlined system in Microsoft.
And I explained how inline on the Mac at least works. So I could see them doing something similar
or just having a keystroke that you press that you prompt it. And then it could put
put that information into word.
The possibilities are limitless with this technology.
I wanna emphasize a couple of things that Lynn said.
One is, check your facts,
check the information before you just run with it.
I think that's something that was reiterated to me
when I learned about Spouse Check when I was younger.
Listen to what it says, the computer's not always right.
And I think that's important.
And then secondly, it can definitely be a good
mental support tool. Obviously, seek professional assistance if
you need if you need it, but it can't hurt to ask some of these
questions and get information. Well, let me rephrase that. It
can't hurt dependent on the the way you use the information
Marty, you mentioned earlier, I couldn't imagine an angry GPT.
I heard about the other day called a chatbot called chaos GPT.
that's prompted with the sole intention of creating a plan to destroy humanity.
So like anything, that sounds great to me.
That's just crazy.
Yeah.
Yeah.
But like anything, a butter knife can be used to spread butter on toast or
stab into someone's chest.
Yeah.
All about how you use your tools.
So another way that chat GPT is being implemented is into an app, for
example, called Be My Eyes.
And what Be My Eyes is, is that if you need assistance
and you're someone who's visually impaired or blind,
it opens up an app on your phone,
somebody on the other end connects with you,
they can utilize your camera on your phone
and you point it at whatever it is you're looking for
and they can tell you on the other end what exactly it is.
So for example, if you're in the grocery store
and you're looking for the chicken noodle soup,
you're on the aisle, there's a hundred cans of soup up there
and you can't figure out which one,
You can get your phone out, you can connect on the app.
Somebody will come on and use your camera and tell you,
"Oh, that one right there, that's the chicken noodle soup."
And they can tell you which is the name brand.
And so you know exactly what you're getting.
So what Chat GPT is doing and what the app company is doing
is they're gonna implement this technology into the app.
And so what that means is this technology is going to be able
to do the same thing on its own using the technology
without having an actual human come on the line.
Now, if you do have a failure with the technology,
you still can ask it to connect to a human,
but this is gonna be really cool to see
how as this goes along, how smart it gets,
how accurate it will be and all of those things.
So I'm kind of excited to try that technology out
and see how well it works and where we end up with it.
Yeah, they call it the virtual volunteer.
And I think right now it's only being rolled out to certain people.
I think they're still testing it and stuff, but yeah, it's slowly being rolled out.
I'm on the, you can get on the wait list.
And I'm still on the wait list.
I thought I had it, but I had the tab, but I don't have the, I'm still waiting for
the, there's a tab that's this virtual volunteer.
And if you go in there, you can sign yourself up to be, you know, to have that virtual volunteer.
I'm still waiting.
So, but yeah, well, as you get it and you play with it, you can report back and let us know, you know, your
experience and how it works and stuff.
And I really, I just love that we're having this conversation because I think right now when people talk about
AI, it's always, you know, it's gonna, you know, take everybody's job and eventually
it's gonna destroy humanity and all that. And the way I look at it, well, that may happen,
but until then, I'm gonna write a mean essay.
There you go. And you know, we're gonna be coming back to this topic because there's
going to be continuous new information quite a lot, you know, there's going to be different
companies, you know, they're going to be doing stuff, different apps, and it's going to be
be implemented all over the place.
So we'll make sure that we keep you guys updated
and kind of let you know what's going on
as we hear things and figure it out moving forward
and all that good stuff.
So, and in closing, we'll keep you all updated.
If you wanna reach out to us,
you can reach us at [email protected]
with questions, comments, or anything else of that matter.
And Michael, anything in closing?
Just stay subscribed. Don't forget to share this episode with someone if you found it to be useful.
We really appreciate you joining us, Lynn. Thank you for coming today.
Oh, it's been real. Yes.
Thanks, Lynn. And thanks, Paul. And we'll see you guys next time.
(upbeat music)
(upbeat music)
(upbeat music)
(upbeat music)
[TONE PLAYBACK]
You can tune into this call live every Tuesday.
Visit acb.community to learn more about what is up coming.
This podcast is powered by Pinecast. Try Pinecast for free, forever, no credit card required. If you decide to upgrade, use coupon code r-431b7d for 40% off for 4 months, and support Unmute Podcast Network.