You won't pay (directly) for AI

Welcome to CXA's newsletter.

What is this newsletter? It's about marketing, consumer behaviour, AI and will always include useful marketing chart. We're aiming to get this out once a week. The 'in summary' section enables you to scan the main points less than 60 seconds.

If that sounds like something you're interested in, please subscribe signing up for free, here.

In summary:

• Apple's approach to integrating AI is another nail in the coffin for consumers paying directly for AI, for now.
• Microsoft, Google and Apple will be the big winners of Assistant Intelligence.
• Chart of the week. The data is in: who makes better converting Ads, humans or AI?

What Apple really showed us at WWDC

Amongst all the news, reviews, podcasts and YouTube videos about last week's Apple's WWDC conference one key takeaway appears to have been largely overlooked. What Apple's integration of OpenAI within Apple Intelligence really demonstrated is that consumers paying directly for AI isn't going to be a thing anytime soon.

What's fast becoming obvious is getting AI usage into the hands of billions of people in order for the LLMs to use and learn from their behaviour, is worth far more (currently) than resricting access or keeping them behind a paywall.

This is perhaps no more evidenced than the fact that Apple is reportedly paying OpenAI in exposure not dollars. Google's Gemini is reportedly also coming to Apple devices - when, in what fashion and for what purpose, we'll have to wait and see.

As noted in a recent a16z podcast, LLMs have already essentially run out of modern human language to train on. The focus now has become one of getting access to more human made language through partnerships (see Google's deal with reddit) and learning from human usage.

The alternative to training LLMs on human made content is LLMs training on the outputs from other LLMs. This recursive approach runs the risk creating AI 'hallucinations' which produce inaccurate output. (AI hallucinations are inaccurate or misleading AI outputs due to not being trained on enough data, making illogical assumptions, or learnings from biases in the training data.)

Why Microsoft, Apple and Google are in the box seat

All owners and VCs invested in bespoke LLM and image generation tools should be more than a little concerned by the recent moves by Microsoft, Google and Apple.

Microsoft recently added Co-Pilot on all their new laptops, into Edge, Bing and Windows 11, Google is rolling out Gemini to all Pixel 8 and 8a users (8 Pro users already had access) and as covered above, Apple partnered with OpenAI for the launch of Apple Intelligence.

With AI usage exploding why should smaller players worry? Because an Apple, Android or Microsoft device is how 99% of humans connect to the internet. With helpful Assistant Intelligence built in by default it will surely limit the desire or need for most people to look beyond the tools already included in their device(s).

This of course is all the intended outcome by the big three - losing ownership of AI is an untenable situation. So, in one sense while AI model building is only just getting started, in some areas the window of opportunity may already be closing - such is the speed of AI development.

Who makes the best ads?

Chalk up a win for ad creators, you're still beating AI-generated Ads! Well, mostly.

Neil Patel recently provided the chart below to highlight how Ads converted to sales when generated only by humans, only by AI or by a combination of humans and AI.

In a huge sigh of relief for everyone in advertising and marketing the good news is you're still better than AI only, and even better when you mix some AI into your performance marketing.

So, sleep easy, AI is still your collaborative friend - for now.