Issue #27: Custom GPT guides and how-tos, MotionBrush
First up this week is an update on OpenAI’s new Custom GPTs.
In the special issue I released last week, I briefly went over updates from OpenAI and specifically their new GPTs.
At the time I showed you some insane GPTs created by the community less than 24 hours after the release.
Today it seems they’re the only thing the AI community is talking about.
But here’s what you should keep in mind:
ChatGPT is a “wrapper” on top of their GPT-3, 3.5 and 4 models.
When it came out last year, it made what was already possible (“talk” to GPT-3 at the time, and use it for all sorts of marketing and writing tasks) simple and easy to do.
GPTs are just simpler versions of what’s already possible to do (and has been possible for quite some time).
They’re like simple automations and vector search (text embeddings, RAG) bots that you could’ve made already.
So, GPTs are technically not “new”.
They just make it simple and easy to build your own bots.
As a marketer, that should be the lesson:
Making your product and service simple and easy to experience will often be the best form of marketing.
And some might even believe you’re re-inventing your industry, or calling your thing a “game changer!” (even though it might not be).
With a good enough prompt (that doesn’t have to be a “mega prompt” or “super prompt” or a thousand words) and, most importantly, “knowledge” as context, you can build a GPT for anything you need.
Want to talk to your avatar? Make an AvatarGPT.
Want to talk to your competitor? Make a CompetitorGPT.
Want a junior copywriter who writes in your style? Make a CopywriterGPT.
By the way, I’ve been talking about these kinds of chatbots (for your marketing, sales, product, competitor, market, avatar, etc.) for a couple of years now.
It’s all been possible already but hard to build.
Now? A bit of text, a few files, click to save, and you’re done.
Worth for you to keep in mind:
So far, these GPTs are limited in what they can do.
For a lot of companies, they’re better off with either their own fine-tuned LLM or their own set of vector search bots.
In today’s issue I’ll give you a rundown on how I see people using custom GPTs, and how you can create and use them effectively for marketing.
In today’s issue:
Use cases for Custom GPTs.
Custom GPTs 101 (What you should know).
RunWay’s new tool: MotionBrush.
Google and OpenAI are competing hard for AI Researchers.
The future with Custom GPTs.
Let’s dive in.
Custom GPTs are the next big thing when it comes to AI.
And they’re something you should pay attention to.
(Yes, of course there’s a gigantic wave of hype around them—and I’m sure someone is selling repackaged public information on how to build “Super Duper GPTs” somewhere).
In less than a week, some insane GPTs have been created and are already being put to use.
I came across this GPT the other day:
Think about your own situation.
If you had everything, or a lot of things, ever published by a famous copywriter like Gary Bencivenga (or anyone), you could build your own Gary GPT that critiques your copy.
Apply this to any area—video critiques, ad critiques, emails, landing pages, you name it—and you can have a boardroom of mentor GPTs, giving you feedback.
In fact, you can create an entire marketing department of junior staffers across any channel (Eric the Email GPT, Peggy the Podcast GPT, etc.) and run 60-80% of your marketing with bots.
How do I know?
Because I’ve already made them and used them for a couple years now. Not as GPTs, but as underlying automations and LLMs that interact with my data, context, and so on.
I’ve had my own Boardroom and Bot Teams for a while now.
GPTs make the creation of those incredibly simple and easy.
(This is why “mega prompts” are so funny to me. You need to upgrade your thinking and look beyond “10,000 prompt packs” you see pitched online).
Need a website? There’s a GPT for that:
GPTs become powerful when you pair them with custom actions:
This obviously goes beyond marketing.
Integrate OpenAI’s vision model and you can do this:
Do you understand now when I say:
Upgrade your thinking on what’s possible now.
Stop chasing prompts and tools.
Start making GPTs with knowledge, information, and data as context.
Stop outsourcing the future of your marketing and business to some silly “mega prompt stack” somewhere.
We’re long past the hype cycle of prompts, plugins, and custom instructions (which, in the end, weren’t very useful).
OpenAI is likely testing various beta features to a few select users, like this:
As GPTs become more advanced, I believe they’ll eliminate hundreds of other AI tools that are built on top of GPT-3.5 or GPT-4.
Why would I pay for them when I can have all my marketing needs met for $20/month, and I’m leveraging my own data and knowledge?
That’s right, I will not pay for another tool.
ChatGPT is morphing into the ultimate Copilot.
Why would I need yours?
(Again, I don’t).
Anyway, I’ve put together a breakdown of custom GPTs and how to use them most effectively for yourself.
But, I can see how this will play out already (and you can, too, probably).
All the custom GPTs are going to be doing the same thing in a slightly different way about a slightly different topic.
For example, there will be a bunch of Alex Hormozi custom GPTs and Joe Rogan custom GPTs, where they take YouTube transcripts and just copy and paste them into the custom GPT thinking that it will be popular.
Instead of doing what everybody else is going to do you should focus on what would make yours different.
Most of all, you should only have your business in mind first when making these.
For our case here, this can be something around marketing that isn’t another generic AI tool that does one thing with one click.
If you have to make more than one or two GPTs to do certain tasks better then go that route.
Let's get into creating a Custom GPT.
You’re going to need a name/picture, instructions, and files for it to use.
Aside from that, I have a few tips on what to think about when creating your GPT:
Purpose: Your GPT needs to do something specific, most likely solve a problem or help you achieve a specific outcome..
If you’re creating your GPT around marketing to do anything with marketing campaigns, I would suggest you create an avatar and upload that to your GPT so it knows who you’re trying to sell to.
Information as Context: The #1 most important part of your GPT, and what makes it even work, is the information (that is, the context) you give it.
You need to find unique information, which is best sourced from your own content and data.
This stage is also when you can implement AI actions and combine AI tools within your GPTs knowledgebase, with a tool like Zapier for a more custom-tailored tool.
Personality: One of the most distinct features of the GPTs is that it’s possible to give them a unique personality or perspective on the information you provide.
As you build your GPT, it will ask you questions on what stance you want it to take, how you want it to respond to users, and other questions tailoring outputs.
Now, before you go and buy some silly ebook or online course on how to build GPT bots, here are several free resources that are all you need:
It’s really not hard, at all.
In the last newsletter I talked about Musk releasing Grok.
Since Grok is supposed to have a sense of humor, and can learn in real-time from content on X (formerly Twitter) I see it being implemented in real-time for marketing campaigns specifically on X.
Especially with Custom GPTs rolling out to the public and the potential of them to automate singular tasks being high, I think at some point these two AI’s will be combined to create, for example, an X Ad Campaign manager.
Here’s how I see the future of this combination playing out since custom GPTs are the next big thing::
You’ll use a Custom GPT to create a persona of your brand and you’ll input Grok’s database and knowledge into it to understand X content, metrics, and marketing campaigns.
There will probably be a way for the two AI’s to “Communicate” and Grok could teach a custom GPT.
The custom GPT will actually work to take ad campaigns, copy, design, and metrics in, give feedback, and create new, better-optimized campaigns from those inputs.
Another way you could structure this custom GPT would be to make it generate optimized ad campaigns with copy and design based on prompts and information you give it.
I see this as the next competitive market for ad campaigns because it has the potential to be a one-person marketing team for an entire social media platform, for any sized company.
Finally, here’s a quick view of ways AI can help your business and marketing:
Increased efficiency: AI tools allow you to automate time-consuming and repetitive tasks in different areas of business, especially marketing.
I’ve seen AI implemented by a lot of junior employees. Most of the time, they’re doing tedious work that doesn’t require a lot of human skill.
Finding an AI tool to automate something like Excel sheets in your business can be a huge improvement.
Decision-making: AI-powered programs can analyze vast amounts of data quickly and accurately.
AI is better at discovering hidden insights and patterns because it doesn’t forget things like a typical team member would, scrolling a document.
This extra pattern recognition and information could have a huge potential to help your managers in your company make better, faster decisions.
Especially for marketing, AI tools can take unlimited inputs and analyze all of them in minutes without missing a thing.
Personalization: AI tools are powered by machine learning and generative AI algorithms.
These can learn and understand individual preferences, behaviors, and demographics and tailor content and recommendations to match specific consumer needs live when they’re on your website.
Implementing a tool like Zapier to connect a chatbot and a data analyzer is a great way to implement this.
Cost reduction: AI implementation will obviously reduce costs compared to having employees work for you because of its ability to work faster, 24/7, and for cheaper anyway.
AI tools will also identify defects and performance issues faster than employees, saving you costs once again.
Scalability: With the use of AI, your business can process large volumes of data in a short time and work 24/7 for you.
This results in you being able to easily scale according to market conditions, and not according to the number of team members you have.
Creativity: By analyzing large datasets and having specific functions, AI tools can actually scale your creativity, while also being more effective by taking performance into consideration.
(Temporary) competitive advantage: AI tools can perform certain tasks much faster and more accurately, giving your business a competitive advantage over a competitor without the same AI implementation.
Even if you only implement AI without letting employees go or cut costs, you’ll be getting more done and you’ll out-perform your competitors.
The way I’d do it? Implement AI within each employee's routine and the company's process to keep your entire staff but increase the amount that gets done at the same time.
But this competitive advantage is temporary. Because your competitors will do the same, catch up, and perhaps even race by you. So, you’ll need to source your advantage in something else, not just speed.
If you haven’t noticed, I’m telling you to think and act strategically.
Using AI (and ML) is not an unfair advantage.
Anyone (and everyone) can use the exact same tech and tools as you.
You need something else (more on that in the Last Byte below).
Runway just introduced a new way to create controlled movement-like looks to your AI-generated Images with a tool called MotionBrush.
MotionBrush allows you to just swipe the area you want to move and it’ll generate motion for you. Here’s an example of the tool:
Why is this important? A simple picture in someone’s timeline won’t always attract their attention when using it for ads.
With MotionBrush, I see the potential to add a unique dynamic to your photos and not look so much like an ad.
A key element of running ads on different platforms is that you don’t want your ad to actually look like an ad. You want it to be inviting and seem like another form of content from a creator.
Here’s an example of Runway using MotionBrush on the release:
$100M+: Google is looking to invest in AI startup Character AI, which already has connections with Google and trains its models with Google Cloud services
$10M: OpenAI is reportedly paying up to $10 million in compensation to hire targeted top AI researchers at Google. Sounds like Google and OpenAI are going at it fighting for senior AI researchers.
$500M: IBM is putting down $500 million to invest in generative AI startups focused on business customers. Although this dedication has no timeline, it shows that IBM is changing, and the senior vice president of software and chief commercial officer backed that.
🤖 FocuSee turns screen recordings into Eye-catching videos without any tedious or time-consuming editing needed. Playing around with it for a few minutes with some audio and video, it could be a great platform to create ads and enhance loom videos.
🤖 Dante AI is a custom AI tool that creates AI chatbots for your business’s website without the need for coding. It can also integrate chatbots into different messaging apps such as WhatsApp or Slack.
THE LAST BYTE
For a couple of years now, I’ve been steadily repeating myself on a few key topics.
Most of the time, people smile and nod, then move on to ask me about prompts and tools.
I get it, those things are shiny and fun.
But let me frame this for you properly:
AI and ML tools are only becoming easier and simpler to use.
What used to require a developer, ML engineer, time, money, computing power—a lot of that is not required anymore.
Which means every tactic, tool, and prompt you see and use can be used by your competitors, too.
AI is not an unfair advantage when everyone has the same level of access—and you can find free tutorials on doing anything and everything.
So, what matters now?
The remaining moat for a serious business remains their proprietary data and what they do with it (curation, labeling, information architecture, and so on).
For a while.
But datasets buy you a few years.
So, what’s your competitive edge now and for years to come?
Here’s where I’m going to tell you what I’ve been saying, publicly, for years already:
A sense of taste.
Learn how to think critically.
Learn how to communicate better.
Practice your imagination.
Do you believe me now?
See you next week,