AI OSes: bad news for ads?
In my last article, I gave a definition of an AI OS. Here, I’m going to talk about why the AI OS matters, in the context of the AI vs. ads scenario.
The thesis I’m working on in the saturated world of AI commentary is:
AI poses a huge threat to ad-based platforms by slashing how many ads we consume
This thesis, if true, would have huge implications for the business model of big tech — in particular, for Alphabet and Meta.
So I’ll try to answer two questions in one go:
Why care about AI OSes in particular (instead of dozens of other AI ‘things’)?
Why might AI OSes be bad news for ads?
1. All OSes will tend towards AI OSes over time
AI OSes allow AI to do more, with greater control over hardware and software
This means AI OSes will make personal computing more efficient
So as AI becomes cheaper, all personal computing will tend towards using AI OSes
We already see this shift happening, as OS updates integrate AI more deeply
And that matters to ads because…
2. Ads will mostly live (or die) inside AI OSes
Most of the $1tn ad inventory is safely tucked inside personal computing OSes
These will all become AI OSes over time
AI OSes will look very different to existing OSes
So to understand the AI threat to ads, we need to think about AI OSes
Perhaps safely will change to precariously
And there are a couple of things that may make AI OSes different to other AIs:
3. AI OSes may have permissions and interoperability to threaten ads
As we saw in my last piece, we’ll have weak AI OSes and strong AI OSes — with the difference being how many permissions the OS gives to the AI to manage hardware and software
Strong AI OSes will have permissions to work with other apps on the system, assuming some level of interoperability (either permitted or adversarial)
That’s a huge problem for big tech companies who rely on ads that live inside those other apps:
Directly: strong AI OSes may be able to:
Either: selectively take content from/operate on your behalf in those apps, meaning you never see their ads
Or: hide ads if you do go into those apps
Indirectly: strong AI OSes may reshape human-computer interaction in ways that decrease the size of the ad inventory
For instance, towards voice control and away from screen time
To put this another way: the AI vs. ads scenario depends on an AI with many permissions and a high level of interoperability, and it makes most sense (to me, at least) to conceive of that AI as an AI OS
And to put it yet another way: AI that’s siloed in a browser tab or app is a lot less of a problem for ads
Although, still somewhat of a problem — if we were to see ad-free ChatGPT activity displacing ad-packed Google Search, for instance
There are a lot of assumptions to test here, which I’ll get to in future pieces — not least, will the kind of AI OS envisaged here exist, or will AI OSes be controlled by companies who ensure they are still packed with ads.
It also makes sense to talk about AI OSes instead of some other AI ‘thing’ because:
4. AI OSes are a cross-cutting concept
Plenty has been written about the many AI devices that are going to revolutionise our lives: AI wearables, AI cars, AI robots
Talking about AI OSes cuts across specific devices (which will all tend towards using AI OSes) and allows us to zoom out to the broader question of how human-computer interaction will be reshaped by AI
Once we do that, we can zoom back in to consider which devices are likely to succeed in an AI OS age, and what that’ll mean for ads
Next up:
I’ll work on concrete scenarios of what AI OSes will look like in the next 5-15 years
In some scenarios, ads will do just fine. In others, they won’t. I’ll explore which scenarios are most likely, and the assumptions behind the scenarios
I’ll publish a reading list of everything useful I’ve found about AI OSes
I’ll also explore things that are AI OS-like, to look for clues of where AI OSes might be headed: existing concepts of AI OSes, AI chatbots, the LLMs that power them, AI agents, AI browsers, AI wearables, AI phones, etc.