(upbeat electronic music) - We are coming in quick and hot today with some breaking news on AutoGPTs and the future of AutoGPTs for engineers, for writers, for businesses in general. I'm your co-host, Kipp
Bodnar, CMO at HubSpot. I'm joined by my co-host Kieran
Flanagan, the CMO at Zapier, and this is Marketing Against the Grain. (upbeat electronic music) Let's get into today's show. Kieran, we got some new
news on the AutoGPT front. We're seeing some innovation
happening in AutoGPT Land. Can you first remind
people what AutoGPTs are, and walk us through a little
bit about what's happening? - Right, AutoGPT, we
covered it on the channel. People loved it. It was basically, what they
are is chaining together separate GPTs to actually achieve tasks. So, I set it up on my laptop, and when I set it up, I
could actually give it a kind of broad-based goal, like go create marketing campaign for me and then actually follow up with leads and sell them with that brand
campaign, or whatever it is. And it would actually stitch together, it would actually create that experience, and it would actually pass
off work to different GPTs, and so it chains together different GPTs to be able to complete entire tasks or entire jobs to be
done from start to finish with little to no details. Now, some of them will actually come back and ask you clarifying questions, but, like, this ability, really, to take human out of the loops, right, it prompts itself. It writes its own prompts, passes the prompt on to another GPT. GPT does work, passes on to another GPT. GPT does work. And so, AutoGPT was super interesting, very, very buggy, a lot of hype. Didn't do a lot, but we're seeing a new kind of
versions of them coming out, which is much more interesting. - All right, so TL/DR,
it's a bunch of robots working together to do stuff for you. - Bunch of robots just working together, doing a bunch of shit.
- Which is pretty sick. I want robots to do stuff
for me all the time, and we've seen what was kind of like hype, kind of didn't do that much, start to manifest into some
pretty interesting stuff that it's now able to do, right? - Right, and so, I saw this
one here from Matt Shumer, and it's like pretty cool, right? This thing here, you can
see he's got a prompt here to write a future, a whole novel where humans have integrated machines and something about magic,
and what it actually will do, he talks about the writing style, and what this here will do, it will actually go create
the actual cover art. So you can see it here,
creating cover art. - [Kipp] Oh, sweet. - [Kieran] Create all of the chapters, and then actually write the book for you. So where are we going with this? - [Kipp] That's crazy, man. - We are going with AI.
- The cover art, even, is awesome.
- Yeah. The cover art, like
here's all the chapters, here's all the content it's created. It's created custom
content for that person. This is where we're going with media, which is you don't have to buy
kind of similar book to me. I have my own book personalized to what I want to read, right? I have my own movie personalized
to what I want to watch. I have my own game personalized
to what I want to play, and I know this is a very
early example of that, but it's a pretty incredible example. - It's one that's an incredible example, and I think most people
are gonna see this, GPT-Author, would say,
"Oh, that's awesome. Like, we're just gonna have more books that are written by robots, cool," and that would be missing the point. Kieran, you were making the real point, which is like, when you
can spec out a custom book, you can have a book that
is only ever written and read by one person.
- One person, exactly. - Right? At HubSpot, Kieran, I'll
test this out on you, we're talking about, you know, the last era of the internet
was all about personalized, and now we're moving to personal. - Yes.
- Like you're having this one-to-one.
- One-to-one experience. - Segment of one experience, and that's pretty fricking incredible. And even if you think about how we talk about the last hundred years, we've talked about it's largely been the
age of mass media, right? Like, I went to a school that was like school of journalism and
mass communication, you know? Like how do you communicate
at massive scale? We're actually gonna flip that on its head and say, "Hey, it's actually
personal media now." And you can have a viable
business taking this idea and making millions of
personalized options. And I think you and I are arguing that's actually not mass media. That is a way to deliver a ultra-personalized media experience, personal media experience, to a lot, millions and millions of people. - Yeah, you can imagine someone
just having a subscription, being able to go into that subscription. - Oh, for sure.
- Say, "Hey, my kid, this month, my child this month wants to read Harry Potter novel, but the Harry Potter novel
is based in Wizard of Oz, and Wizard of Oz also includes, like," what's another fun thing? What's another kid's
thing that I can think of? - [Kipp] Peppa the Pig. - Peppa the Pig. Yeah, Peppa of the Pig is
there and she's like the- - [Kipp] I'm trying to think of lots of British accents and stuff. - [Kieran] She's the Wicked
Witch of the East, right? And actually, like, okay, AI creates tailored book
with all of the imagery, and probably some videos
to bring it to life, customized for your child. And that's an incredible
direction that we're headed in. Actually, one of the funny
examples that I saw here, not a funny, like pretty
interesting example, where someone was trying to
use this open source product to write a sequel to their novel. So what he did was took this code. - [Kipp] Oh, nice. That's smart.
- And because of the tokens, the token limit could actually
paste in his first book and then ask the AI to write
a follow-up of that book and give it some guidelines
of what it wanted to happen in the second version of that book. So, like, pretty, pretty incredible. - Well, hold on. Before you finish this out, what's pretty brilliant about that is we are now in this, like, IP era, where, oh, cool, if
somebody creates a story, whether it be one of the Marvel stories, super hero stories,
whatever, that resonates, we just go and then spin off a bunch of sequels to that, right? AI is gonna let us do that for anything. It's like, cool, you read
this article that you liked? Well, I could turn that
article into a book and you can go really
deep on that if you want. - Yeah, yep. - And that is fricking mind-blowing man. Like I don't know that people really grock the future we are about to live in. - You like this thing. Why don't I just create different versions of that thing for you, right? - But then, also, how do
you ever learn anything new? If you could have an
endless amount of shit that is stuff you like- - This is a whole other topic that I- - How would you ever learn
anything new, challenge yourself? - [Kieran] I talked about this. - I'm kind of worried about
our society a little bit. - I did talk about this with you. We can cover this in a future episode, 'cause I kind of mentioned
this in a previous episode, which is I feel like this media direction. - Could be bad.
- Puts everyone more firmly in their echo chamber.
- Comfort zone? Yeah, echo chamber. - It actually allows
you to go even further, like social media allowed
us to go even further into our echo chamber,
'cause now we can find people who believe the things that we believe, and we can just hang
out with them all day. Media for one helps you to
go into your echo chamber even more, because now I can
just spend all of my time consuming things that I truly believe. That's actually one of the
problems with our society, is, like, no one actually thinks. You stay in your little lane, and you're like, "I believe my little lane and everyone else is bad." And personal media being
created for an individual exasperates that problem. The one that's getting a lot of traction I want to cover, too, and just talk about them real quick, so GPT-Engineer is like the one that's getting a lot of traction. So, Lior, who we've covered before, we've referenced before.
- Yeah, shout out to him. - Does a really good job
summarizing all this stuff. This has 12,000 stars on GitHub already. I actually dove into the Hacker News post where the creator of this
was actually going through why they created it, wanted to
provide this kind of sandbox for developers to be able to create code and having this assistant
to create code with, and went through, like,
people are iterated and trying to improve upon it. But the kind of interesting thing is, let's just start this little video. And so, what's really interesting is this is it making the snake game, right? And you can see it's just given some basic inputs, right?
- I love this game. - Like talking through the, you know, basic requirements of creating
that game, and then runs it, and then GPT will, the GPT-Engineer, will hand off to different GPTs to just go through and start
to create all of this code, and it will actually answer, it'll actually ask you
clarifying questions to make sure it has the right details. One of the things that I saw on the video, which is super interesting, right, this is it actually
building the snake game now. One of the things that I saw, that I've seen people do is
I can actually go to ChatGPT and I can ask it what are
the rules of the snake game and how would I give that set to a engineer to build it to me? Then I can take that and then I can just give
that to GPT-Engineer, and GPT-Engineer will actually go ahead and start to create that game for you. So you don't even need to actually build the functional spec. - The actual requirements. - Yeah, yeah, you don't even
need to build the requirements. You can actually just go to
ChatGPT, ask it for something, tell it how would I actually give this to a developer to build,
then go to GPT-Engineer, which is this kind of series of GPTs that work together to complete something, and actually just give it
those functional requirements to complete that thing for you. - You literally just
have to have the idea. - Yeah, you have to have the idea. - [Kipp] It's crazy. - 'Cause, again, here,
asks clarifying questions, generates technical spec,
writes all necessary codes, easy to add your own reasoning steps to modify the experience, and then lets you finish a
coding project in minutes. It's just getting, you know,
it's just another step up in terms of what these AI agents can do and the amount of work
that they can do for us. So that is the one that
has completely blown up. Very, very interesting. - Well, yeah, Kieran,
what I would just say is you have this belief that
the costs of taking risks in an AI world are essentially zero, and this is a great example of that.
- Great example of that. All you need is the idea. - You're like, "Hey, I have this idea about this experiment I want
to run to grow my business or this little app I want to
build to see if people will use and it will bring customers
in for my business." You can build the most basic,
lightweight version of this without really any human time,
and then put it out there, validate if people
actually use or like it. It's pretty incredible. - It's super incredible. I think the other thing
you'll be able to do with this is just pass in the wire frame. You don't even need to
pass in functional specs. Pass in the wire frame, it'll create something
from the wire frame. We've already seen, we saw in our episode where we took in founders
and they pitched us. We saw the winner of that
was taking Figma Design and building the code from them. But again, yeah, like, what matters, then, when you actually have
a structure like this that will actually build something
from simplistic commands, and actually, not even just build it. Like the big thing here is it
doesn't just take your idea and just go try to build it. It actually knows to ask
you clarifying questions to get the information that it needs.
- Which is amazing. - And the thing that truly
matters is still the idea, right? People with better ideas
are still going to win. But I see this, I see this
chaining together of GPTs, like we started with AutoGPT, but AutoGPT was like you
could do anything with it or you could ask it anything, and it was not that successful. But what we're starting to see is very specific versions of
that that are more dialed in, so there's gonna be less mistakes, 'cause they're dialed into
a specific functional area. And it's pretty cool, right? - Drop your comments in the YouTube around what you think this means for the future of media
consumption and media. I think GPT-Author is a
pretty mind blowing idea. I'm excited that we got to share this a little bit with you all today. We wanted to keep this episode super tight because we just wanted to bring you the latest and greatest in AutoGPT, 'cause we think it's a really
important evolving technology. We'll be back real soon
with more episodes. Please subscribe to the channel. We'll see you real soon on
Marketing Against the Grain. (upbeat electronic music) - This data is wrong every freaking time. - Have you heard of HubSpot? HubSpot is a CRM platform where everything is fully integrated. - Whoa, I can see the
client's whole history, calls, support tickets, emails, and here's a task from three
days ago I totally missed. - HubSpot, grow better.
(upbeat electronic music)
