Issue #28: Your marketing edge with AI, powerful GPTs, more tools

Good morning.

Everyone is talking about GPTs and the OpenAI board drama.

But that’s a distraction. 

There’s a frantic, high-stakes race toward something else:

Datasets.

The race amongst companies like Google, Tesla, Microsoft, and Open AI is only accelerating and becoming more aggressive.

The goal?

Build the best AI model.

It’s not just between large tech giants. Startups, smaller and medium sized companies are all spending tens of billions of dollars, left and right, like there’s no tomorrow.

On what? 

All this money is going towards buying GPUs to train AI models in search of use cases that unlock enormous new revenue streams.

But the defining factor to building the best AI model isn’t training and GPUs.

It’s actually your dataset.

Chamath Palihapitiya gets it right when he mentions “non-obvious” datasets tech giants own:

Datasets are becoming the most powerful, even required, way to create AI models and custom GPTs, agents, bots, and software.

For marketers, you should focus on your dataset when creating your own custom GPT, as this is what really determines its effectiveness.

Think of different metrics you can use for datasets.

For example, if you’re looking to create an email campaign GPT for subject lines a perfect dataset would be subject lines, their open rates, and what the email’s intent was.

Another key point to your dataset is the amount of data.

A GPT you create with 1,000 data entries is going to be less refined than a GPT with 10k, 100k, and so on data entries.

Building on this topic in today's email I’ll be showing you examples of different custom GPTs and AI models, what datasets were used, and how you can use those concepts to integrate custom GPTs into your own marketing.

In today’s issue:

• Why datasets are leading the AI race.

• Different models, GPTs, the data they were trained on.

• Open AI’s new breakthrough Q*.

• The race to acquire datasets.

Let’s dive in.

BIONIC JUICE

This customGPT is called researchGPT and it’s a research tool used to reduce the amount of time it takes to do manual research with scientific research paper and literature.

The GPT itself isn’t significant. It’s pretty simple.

The magic?

The actual data provided. 

The GPT makes the data easier to access and search.

Remember in the last newsletter when I discussed how GPTs are simpler versions of what’s already possible? This is a perfect example of this.

Before custom GPTs, using the Consensus interface to find specific papers could make it difficult to sum up the information you’re looking for, but that doesn’t mean it wasn’t already possible.

ResearchGPT just makes it easier to do that by turning it into a question-and-answer format with links to specific papers.

For marketing, it would be a great idea to create a custom GPT just like this within your business, based on previous metrics you can turn into your own dataset.

A specific GPT I would create is one that uses data in Excel sheets that update in real-time. I would use this GPT to come to data-backed conclusions and summarize different important metrics I need to see to save me time.

Moving into more specified GPTs, Musk’s new AI model Grok is a perfect example as it is updated weekly on content from X, giving it recent and specified datasets.

Grok’s ability to use recent data as well as create positions on certain topics is unmatched and gives a good guideline to resemble it in your own custom GPT.

Using recent data or specified data in a custom GPT could be a more refined way to better target different aspects of marketing.

With Christmas right around the corner, I would create a custom GPT by inputting all my marketing statistics as datasets from my prior years in Q4.

Since Q4 is unique because of the holidays, this custom GPT can give me more relevant results for marketing during Q4.

Speaking of relevant results, being able to tailor the results you get from a custom GPT is extremely important. 

Without being specified, your marketing will fall short and you definitely won’t be automating anything with AI.

A recent GPT called Parameters GPT is a great example of specifying an AI tool in order to get the right outcomes.

Using a custom GPT Alvaro Cintas was able to take the AI generation power of DALLE-3 and use specifications in order to make it generate images more specified to user’s needs. 

Some users have even seen it be more creative with the custom GPT rather than actually using DALLE-3.

Using this specific example, you can modify any AI tool you currently use to be more powerful and effective at what it does.

LATEST DISCOVERIES

OpenAI’s recent AI breakthrough, Q*, which demonstrates early signs of human-like intelligence is raising concerns about the potential for superintelligent AI, and its ability to reflect human capabilities.

Everyone is wondering if this new breakthrough will be a concern. But in the context of marketing, I just see it as yet another use case to outcompete your competitors.

The “Q” name could come from a reference to Q-learning, a form of machine learning algorithm used in reinforcement learning.

Since it could be a form of machine learning, this would be why some are saying it has human-like capabilities to learn as it experiences more and more interactions.

Maybe it could be a GPT-5, or maybe it could be the next breakthrough for AI altogether.

But until we know for sure it’s just another thing to look forward to as an advancement in the power of AI.

Benefiting directly from the “AI race” Nvidia tripled its revenue in Q3.

Nvidia reported record revenue of $18.12 billion for its third quarter, up 206% year-over-year, driven by the demand of big tech companies for its GPUs in AI applications like machine learning and data processing.

Think Google, Microsoft, Open AI, etc.

Nvidia's gaming and data center divisions both experienced significant revenue growth of 81% and 279% respectively compared to the previous year.

Think about how a company benefited from a competition without winning or even being in the competition at all. That’s your next marketing campaign.

Funding:

$50M: Covera Health announced an additional $50 million in funding and its acquisition of CoRead, which uses AI to flag medical errors on radiology examinations. Covera Health is said to integrate CoReads technology with their own to offer more to their marketplace.

$3M: Birdseye Raises $3M to revolutionize how brands and retailers target customers. Birdseye’s targeting technology is powered by an AI algorithm trained on vast amounts of retail transaction data. Think about how this relates to what I’ve been saying time and time again: Use your own dataset to train a GPT and automate 60-80% of basic-level marketing operations.

$55M: HyperSpace secured a total of $55M in funding. HyperSpace is a new AI media and entertainment platform based in the UAE.

BIONIC TOOLKIT

🤖 LinearB gives you free DORA metrics with no limitations on contributors or team size. Getting all four DORA metrics - Cycle Time, Deploy Frequency, CFR, MTTR - based on your team’s data to help define team performance and set data-backed goals for the future.

🤖 Folk is a next-generation CRM that is customizable, generative, and powered by AI. It prioritizes all the relationships that matter for your organization and customizes itself to adapt to workflows from sales CRM, recruiting pipeline, deal flow management, prospection management, and more.

THE LAST BYTE

I said it in the last newsletter, I’ve said it for years now, and I’m going to repeat myself again here.

Prompts and tools are shiny, I get it. They can be incredibly helpful.

But AI and ML tools are only becoming the norm as companies continue to publicly release every type of prompt and tool to the public.

Every one of your competitors has implemented some form of ChatGPT-based prompt automation within their business. And you need to look further to get ahead.

Your unfair advantage, edge, and moat is not found in tools and prompts.

Your marketing won’t magically become better only because of tools and prompts.

Right now, your edge is your own datasets.

But in the future, everyone is going to have access to all datasets, as synthetic data is proving to be useful and can replace the need for “original” data.

But on the bright side, that’ll “level” the playing field.

And by the time we get to that, I see there being something else putting you over your competitor anyway.

My advice? Look at what the big dogs are doing.

Elon Musk bought Twitter before the whole AI craze and competition for the best AI model came about, but look at how useful his purchase is now.

He has a top-tier AI model (Grok) trained directly from the platform, which is a never-ending, always updated dataset.

At the time it almost seemed like a joke as to why he would buy a social media company meant for Kardashian gossip for a whopping $44 Billion.

But when you consider how valuable datasets are now, and how they can generate giant revenue streams for big tech companies, Twitter now seems like a smart purchase.

As a marketer, you should have access to some of the most valuable datasets in the world:

How people behave (clicks, engagements, conversions, etc.).

What people think, feel, and believe.

You actually have an advantage in the kind of data that will matter.

Start gathering. Make an inventory. Then go get the data you’re missing.

See you next week,
Sam Woods
Editor