• Bionic Marketing
  • Posts
  • Issue #29: The Q* mystery, short-form GPT, Google’s Gemini, and more

Issue #29: The Q* mystery, short-form GPT, Google’s Gemini, and more

Good morning.

A couple things for you this morning: 

1. What’s really happened with OpenAI, their new Q* discovery, and the dismissal of their former CEO Sam Altman.

I have “thoughts” as to what’s going on with the new discovery and what’s going to happen to OpenAI in the coming weeks and into 2024.

Most people are speculating about Q* and how it’s going to affect the evolving nature of AI. So, it’s mostly a rumor mill at this point.

My most important advice: 

Keep your head down and keep doing marketing.

Q4 is a busy time, especially with the end-of-year approaching us. You’re going to need all the time in the world to focus on busy stores and marketing campaigns.

Speaking of marketing campaigns:

Short-form content for both ads and organic has been on top of every platform for the longest time.

I know the most crucial piece to crush short-form content is the idea or hook. 

Without a good hook, the shortest 15-second video can seem like a pain to create.

In today's issue we are going to use a chatGPT prompt to create as many ideas for short-form content as possible—along with an idea of how to turn it into a custom GPT.

Now, thing #2:

Some of the people they interviewed surprised me.

The article is a great resource on how to use AI in your work with the best intent.

It’s inspiring to see how people are implementing AI use in their classrooms to prepare students for the future. We’ll be going over some of their responses today, and how what they apply can be used in marketing.

Also on the agenda today is a quick demo on how to use a new Human Image Animation tool called MagicAnimate.

Let’s get to it.

In today’s issue:

  • OpenAI’s future post-discovery of Q*.

  • Generate short-form content ideas with ChatGPT.

  • How Academic Scientific Researchers are Using ChatGPT.

  • Interesting ew tool: MagicAnimate.

Let’s dive in.


A quick summary and interesting answers from that article on how researchers use ChatGPT.

Marzyeh Ghassemi: Fix, don’t amplify

Marzyeh uses ChatGPT mostly to help rewrite content in a different style, as most of the content is deeply researched.

For example, think of taking a piece of researched content and using ChatGPT to make it easier to read and understand for a general audience. She also uses it to suggest introductory language at the start of an article, e-mail, or paper.

I see this being a great idea for your marketing team to implement into product descriptions or ad copy. If you have lots of information and use cases for your product, but want to find a way to appeal to your ideal audience better, ChatGPT can do that for you.

Ethan Mollick: Embrace AI in teaching

Ethan Mollick emphasizes the need for ethical experimentation with AI in the classes he teaches.

He incorporates AI into every teaching assignment and lesson, making his students learn to work with AI. How great of a skill is that!

If AI is taught more extensively in Universities and even high schools, it’d be amazing to see how the curriculum adapts to it.

I’m just thinking of a fresh kid with a marketing degree who’s been working with AI for 4 years. That would be a useful hire!

Siddharth Kankaria: A tool for tailored teaching

Siddharth Kankaria uses ChatGPT to brainstorm prompts, questions, and content for classroom activities.

He creates a classroom environment for his students to be comfortable using AI tools while being aware of their limitations, as they can have many.

Like I’ve done in the past, chatGPT is a great tool for creating prompts, doing general research, and getting a general understanding.

This applies to targeting audiences, structuring your marketing campaigns, generating ideas for short-form ads, and many other things.

Mushtaq Bilal: Use it for structure, not content

Mushtaq Bilal explores ChatGPT for academic purposes after exploring prompt uses in academic writing

He has developed ways of using it and shares them on X and LinkedIn.

The emphasizes that generative AI is well suited to creating structure, but not content.

“LLMs are trained to predict the next word in a sentence. That means the content a chatbot generates is typically predictable — whereas original research is anything but”

Using this knowledge for marketing campaigns can be super useful. I’ve seen entire Q4 marketing plans be developed with ChatGPT, not necessarily with perfect ideas, but the structure was there.

Next, about the new MagicAnimate tool.

Step 1: Head to the MagicAnimate demo on HuggingFace.

Step 2: Upload the image you’d like to animate. Most of the motion sequences are full-body vertical, so keep in mind that a picture of a person is going to most likely work best.

Step 3: Click animate and be patient as this demo isn’t as fast as the rest of AI right now, and many other new users are trying the same thing.

At this point, anything you think of can be useful for marketing. Whether that’s a picture, video, creative, or even a screenshot, I’ve seen it all.

Next up, short-form creative content is everywhere.

And in theory, it should be easy to create, right?

The bottleneck, though, is coming up with ideas and hooks to use.

Let’s solve that problem with ChatGPT.

Open ChatGPT and paste this prompt: 

I want you to do customer research for me.

Tell me 10 frustrations, 10 desires, 10 dreams, and 10 fears that my audience experiences related to [niche/product] 

Format the output of the 10 frustrations, 10 wants, 10 dreams, and 10 fears in a table.

The x-axis should be numbered one to 10 the y-axis should include desire, frustration, dream and fear. 

I’m going to use “Matcha” for this example

After using Matcha as my product, I’m provided with a table that looks like this: 

The fear column is usually my favorite, as fear is usually the best driver in sales, and allows for easy myth-debunking creatives.

This takes less than 30 seconds to do in chatGPT, and because it’s a simple prompt not needing any advanced AI, you can do it with the free version.

Although these different ideas from the table are great for video ideas, there are tons of different use cases for these ideas.

An SEO-optimized product description can be easily generated from this table. I just told chatGPT to write an SEO-optimized product description based on the information in the table.

To no surprise, it gave me a good framework to create a great product description for Matcha:

I’ve stumbled over my own advice, I know.

I said in the last two newsletters that you need to move beyond prompts with ChatGPT to beat your competition.

But here I am giving you another prompt idea—for good reason.

The prompt is to help you brainstorm ideas and hooks.

It’s not a long, complex prompt. It doesn’t have to be.

But I know a custom GPT that did the same thing could be way better.

Here’s how I would do it:

Create a custom GPT and train it to do exactly this, but with a dataset full of all your competitor's ad creatives, their performance, and sales.

Maybe even mix in some random ones you can find in the FaceBook or TikTok ad library. Get all that data into your GPT.

Now “talk” to your GPT to find the best-performing creatives and the frustrations, desires, dreams, and fears used in each one.

Then create ideas with the same method.

And done.


I haven’t talked about Google’s new AI model Gemini yet, but it’s here now.

Gemini is being described as the next generation of AI and multimodal. I’ll have more to say about it once I’ve taken it for more of a test drive than I’ve had time for so far.

By the way, multimodal means it can process multiple types of data and is said to have the capacity to understand and generate text and images as well as other types of content based on a sketch or written description.

Sidenote: Google has recently discovered a way to use ChatGPT to leak its own datasets and training examples using different sequences of key words.

Is this giving Google an edge to help compete with ChatGPT and they’re working on implementing more into Gemini with this information? 

Lastly, check out this graph of what people think about the use of Generative AI, and how it’ll benefit future research:

Claude 2 was recently released and is more powerful and useful than ever.

Claude has become a go-to for me when I need copywriting or just content writing done.

Now with Claude 2, it’s even more useful for marketers to use for all different uses with AI. They’ve added a 200K context window to help Claude interact with longer documents better and give more accurate, refined outputs.

Do I see someone integrating Claude 2 into a custom GPT? Yes, yes I do.

Usage of the 200K token context window is reserved for Claude Pro users, who can now upload larger files than ever before.

xAI is launching its new chatbot called Grok this week, which will be integrated into the X social media platform. We’ve been talking about this for weeks, and I’m looking forward to seeing what Grok can do.

On another note, Grok is intended to compete with ChatGPT and unlike other chatbots, Grok will have access to billions of posts from X, allowing it to provide up-to-date information by analyzing multiple posts on a topic.

Will this be a “win” over ChatGPT? 

Who knows, time will tell.


$35M: Pika, an AI video platform that is redesigning the video-making and editing experience, announced its Series A funding round of $35 million

$24M: Cradle’s AI-powered protein programming platform levels up with $24M in new funding. Seems like Cradle is finding success with its generative approach to protein design.

€30M: Failup Ventures, a new globally operating early-stage venture capital fund, announces today the first closing of €30 million of the total 50. Setting itself apart from the competition in Norway with its footing in the US.

$51M: OpenAI plans to spend $51 million to purchase neuromorphic AI chips from Rain AI. Rain is backed by OpenAI’s former CEO Sam Altman as he’s personally invested over $1 million in Rain AI.

Rain’s NPUs are said to offer 100 times more processing power and 10,000 times greater energy efficiency than GPUs currently used for AI training. Interesting to see how this will improve OpenAI’s future AI models.

$11M: Cloudsmith announced $11 million in Series A2 financing. Cloudsmith allows users to streamline their software supply chain. With this also comes datasets for users to create AI products, like custom GPTs.

$102.5M: Together AI closed on an impressive $102.5 million Series A funding round to build the open-source infrastructure for generative AI model development. It says this funding will allow developers to use their cloud-based platform anywhere to create open and customizable AI models.


🤖 Simplescraper is an AI tool that allows you to extract data from websites to build your custom GPTs. My thought is how is this going to affect competition between you and your competitors, and to what extent of data can you extract?

🤖 OctoAI has recently created a way to generate high-quality SDXL Images in less than a second. Incredible.

🤖 Shorts Generator creates viral short-form videos in minutes. I’d use this tool to go from idea to short-form ads in minutes and help create templates for paid ads faster. With this speed, you can get ahead of the competition.

🤖 6 Sense is a marketing tool that uses instant predictive analytics to find your ideal customer profile. 6 Sense’s AI can handle the more robotic parts of your workload, freeing your revenue team to use their creative problem-solving skills, and saving you money.


Rumors are spreading about OpenAI’s discovery Q*.

I found a great video discussing Q* and how the name might be hinting to Q-learning by TheAIGRID on YouTube:

Q-learning is a form of machine learning algorithm that is used in reinforcement learning.

My question is, why is that coming up now? 

Machine learning is nothing new to AI and Q-learning was an original AI algorithm.

Along with the Q, the comes from A which is a way for games and AI to find the shortest path between two things.

Again, nothing new. Most people were taught how to find the distance between two things in middle school.

But I do think this is something to pay close attention to, even without looking at the hype from the AI community and the OpenAI team.

Combining machine learning, Q-learning, and A* could have a huge possibility depending on the extent of use of each one.

Think of A* not as finding the measurable distance between two known points with coordinates, but maybe as an algorithm to find the fastest way to get a desired result from the current position you’re in.

Using that ability with a custom GPT and a known dataset of, say different AI tools and methods your marketing team uses, could be powerful for planning campaigns.

It’s insane that we’re here already.

Things are moving faster than even I anticipated.

AI is moving to the moon and back about every hour.

What’s the best thing you can do? 

Slow down, so you can speed up.

I could make a 50-page newsletter every day about the different tools coming out, what’s happening between each company, and how much funding there is for AI this week.

I always say to pay attention and be ready for the next wave, or the next big thing that’ll put you over your competitor.

But today I’d like to end this newsletter by contradicting that.

If you subscribe to every AI newsletter, try to understand what’s going on between every big tech company, and find the best AI tool for your marketing team every day…

You’re going to get lost 

There’s just too much going on.

The best way for you to pull ahead right now is to keep your head down, make your team as efficient as possible, and not spend too much time worrying about the future.

See you next week,
Sam Woods