Issue #38: Your next AI assistant (opportunity for marketers)

Good morning.

After a brief pause, we’re back at it.

This issue covers some of the latest in AI marketing that’s worth paying attention to.

Next issue will take an in-depth look at various prompting strategies for marketing.

First, let’s get you up to speed.

—Sam

IN TODAY’S ISSUE 👨‍🚀 

• Huge AI assistant opportunity for marketers this year.

• Detection tools (and humans) don’t work—and never will.

• 6 more AI marketing updates, news, and tools you should know about.

Let’s dive in.

Everyone wants to be your personal AI assistant

Tech giants are trying hard to be your new BFF, in the form of an AI personal assistant.

Google, Microsoft, Meta, Amazon, Apple, and OpenAI are all hoping they’ll become the world's leading AI personal assistant.

This will have a far-reaching impact on online marketing.

Apple primarily earns revenue from devices, driving them to create exceptional user experiences. 

However, incremental improvements may not significantly impact customers, making it challenging to measure the ROI of innovation when you're already a market leader. 

This presents an opportunity for AI wearable startups to disrupt the space, unless Apple gets there first.

This year, I expect them to integrate some form of an AI assistant, running locally on your Macbook, Watch, iPhone or Mac. Even a better Siri would be helpful. 

Google's revenue largely comes from advertising (~75%), incentivizing them to keep users engaged with their products for extended periods. 

However, the AI personal assistant market poses a threat to their current revenue streams. If users rely on AI for answers instead of clicking on ads, Google must avoid cannibalizing its own business.

You’re seeing various AI replies in search results already. We’ll see how that experiment goes. I could see Google integrating some kind of conversational assistant, much like Apple could.

Microsoft and Amazon have diversified revenue streams, with Microsoft generating income from cloud services, Office, and devices, while Amazon relies on e-commerce and cloud services. 

However, the potential benefits of being a personal AI assistant are less significant in these categories compared to advertising.

Amazon will likely give you an assistant that will help you place orders, give order status updates, and curate what you should buy next. You’ll probably be able to just tell the assistant what you want, and it’ll handle the ordering for you.

Meta derives an even higher percentage of its revenue from advertising (~97%) than Google. 

Crucially, Meta doesn't face the same challenges as Google in terms of chat disrupting search. 

For Meta, the AI assistant market presents a significant upside, particularly if it encourages users to spend more time on their platforms.

OpenAI, although not a traditional commercial giant, brings a unique perspective to the race. 

ChatGPT competes directly for user attention and data, both of which are essential for training AI models and powering advertising business models.

I think we’ll see substantial investments from all players, given the immense potential for profit. 

The AI personal assistant of the future will offer personalization, understanding users on a deep level. It won't merely display ads but will be capable of making purchasing decisions and creating tailored content.

For marketers like yourself, this presents both challenges and opportunities:

• Double down on AI-generated content: AI assistants will be capable of creating personalized content for users. Marketers should explore ways to harness this technology to create engaging, targeted content at scale.

• Double-down on trust: In a world where AI assistants make purchasing decisions on behalf of users, establishing strong brand trust will be crucial. You should prioritize building relationships with customers and creating positive experiences.

• Double-down on data: AI assistants will have access to vast amounts of user data. You should use this to create personalized experiences and look for ways that remove friction in any kind of buying environment..

My take on who wins? 

I think Meta is in the best position so far, and already has Meta AI integrated in Facebook and Instagram. 

And they’re bringing Generative AI futures to any and all advertisers, too.

The assistant that wins will be the one that’s closest to your daily use of tech and apps, without you having to jump through hoops or install more apps.

Neither people or detection tools can reliably identify AI text

It is challenging for both humans and software to consistently distinguish between AI-generated and human-written text

Recent research conducted by a group of German universities discovered that teachers, regardless of their level of experience, had difficulty accurately identifying texts created by AI versus those written by students.

• In the initial study, aspiring teachers were tested on their ability to differentiate between AI and student-written text. They correctly identified only 45.1% of the submissions generated by AI.

• The follow-up study involved seasoned educators, who were able to correctly identify a mere 37.8% of the texts produced by AI.

• The research also showed that teachers overestimated their capability to recognize AI-generated content.

• Compounding the problem, AI detection tools frequently misidentify student-written texts as potentially AI-generated.    

As AI continues to improve its writing abilities, relying on educators' instincts or inconsistent software is no more effective than random guessing.

And AI detector tools are broken, ineffective, and almost useless. 

I also don’t believe for a second that Google can somehow detect all AI-generated content. They look for other signals of AI-generated spam, not in the text itself—unless you do something silly like include “As a large language model…” in your article optimized for SEO.

My team has several websites with AI-generated content, ranking well and pulling in a ton of traffic. No issues with Google updates, no drop in rankings.

More AI marketing quick hits

Microsoft is developing a massive 500B parameter in-house LLM called MAI-1, for competing with top AI models from Google, Anthropic, OpenAI, and many others.

MAI-1 is being led by ex-Google DeepMind founder Mustafa Suleyman, who recently joined the company following the acquisition of AI startup Inflection. I think this was a poor choice, as Mustafa is your typical AI doomer who does not want you to have access to AI, at all.

The 500B parameter model is larger than Microsoft's previous open-source models, with Phi-3 mini trained on just under 4B parameters. 

What does this mean for you? More competition amongst the giants of AI. 

So far, that has led to more models of various capabilities being available to anyone.

The amount of models available, either free or cheap, is astounding to me. I barely use GPT-4, as there are plenty of other models that can either be run locally on your laptop or they’re offered via a service like Poe.com.

Alibaba Research released Qwen 1.5-110B, the largest model in the Qwen 1.5 series with over 100 billion parameters and it shows competitive performance against Llama-3-70. The model supports the context length of 32K tokens and is multilingual. Being one of the most capable AI models and being highly rated at speaking 10+ of the most popular languages around the world, this could turn your marketing campaigns into the next T-series.

ByteDance released Hyper-SD, offering “hyper-fast” and “hyper-quality” text-to-image generation. The model creates insane imagery that changes between model modes. This type of image generation could be useful to test on different campaigns to potentially see higher performance. Check out their demo here.

Sam Altman called GPT-4 "mildly embarrassing at best" and said it’s the "worst model" we will ever use. From all the different AI-powered tools, software, and marketing agendas we’ve discussed being powered by GPT-4, Sam just comes outright and says it’s terrible. Makes me wonder where the future of AI will go then.

DeepL is an AI-powered translation tool that’s been helping entrepreneurs increase ROI by 345%, decrease translation time by 90%, and save €2.7M. With a reduced need for human oversight, it’s a tool that increases efficiency as much as possible.

Graphlit integrates unstructured data with LLMs, including anything from coding to marketing info in an Excel spreadsheet. Graphite can help bring this data to a usable form to develop an app, or in our case to be used to make better marketing decisions.

Just about every AI marketing tool available are made up of one thing, hiding beneath the UI:

Sequences of prompts and tasks.

Anything from Saas or Apps available, to Bots, AI Agents, AutoGPTs, Zapier, N8N, Make, and so on—are all a combination of prompts and tasks.

(For clarity, people are using different terms for the same thing: “AI Agents”, “AutoGPTs”, “Bots”, etc. They’re the same: a sequence or prompts and tasks, often done via API calls or automated steps).

You can have the same task or job performed in any which way: 

Prompts: Tell ChatGPT or other LLMs what you’re looking for and want it to do.

Tools: like scrapers, Zapier, N8N, Advanced Data Analysis, etc. that perform parts or most tasks.

Bots: like Agents or GPTs that automate parts of the workflow.

Apps/Software: A series of API calls (prompts and tasks) to different tools, like GPT-4 and GPT-4 Vision that do it all for you.

Again: 

No matter what specific tool you use, they’re all a sequence of prompts and tasks. 

There are very few transformative and truly innovative AI tools for marketing right now. That will change. I’m working with a few of them, an advisor or investor in others—so I’m able to see some of what the next few months will look like.

If you’re a serious marketer, as in you know your job is to fill a lead and sales pipeline, this is good news. These tools will make your work 10x easier and faster.

You can gain a lot of productivity already, just by using the available tools. But they’re not an unfair advantage in and of themselves.

A lot of marketers are more interested in their pseudo-creator art projects than taking on lead generation and accountability for volume. 

That’s fine. Nothing wrong with chasing vanity metrics and wanting to be a random internet creator instead.

If I’m a marketer in any space, and I see a lot of competitors focus on their wannabe Youtube or Podcast influencer status, that’s good news to me. 

Why? Less competition. 

AI levels the playing field. 

This is good news.

Talk again soon,
Sam Woods
The Editor