“Man eats machine, machine eats man” - It’s a line from the song, “The tower that ate people” by Peter Gabriel, and it’s kinda prophetic.

Introducing the Man and the Machine

Today there is a world of e-commerce stores which make their money from selling products and services on-line. For new entrants into the market it can be a real pain to get discovered, advertising is expensive and if you’re not an expert at it or you don’t quite find the right keywords it can also be rather ineffective.

An alternative that many take is a longer, slower, but cheaper and sometimes more effective way of being discovered. That is to publish articles are content related to the products and services they are selling. This is often referred too as an “organic” approach, as it naturally leads to search traffic being directed towards the publishers sites. This works because the published content is eventually indexed by the big search engines, marked as useful and cross-referenced by other web sites.

Organic discovery is an implicit trade, the site delivers value, usually in the form of information, and the search engines deliver traffic, which then, hopefully, converts into product sales. This is the “Man” (content producer) and the “Machine” (the search engine) today. Of course there are tips and tricks which help a content producers website be index better, or placed higher in the Google page rankings, this is often called SEO (Search Engine Optimisation) and there is a thriving economy of experts who all claim to know how the big search engines work, and how to best optimise your content for indexing.

When the Machine Eats the Man

So what happens if we change the rules? Let’s imagine a world in which Google, or Microsoft create a new search engine which answers search requests not with links to websites, but with the results presented inline. No links are provided to the original sources, and the information is presented as fact. It doesn’t take too much imagination as that is what is happening today. Both Google, and Microsoft are producing results generating by LLM (large language models) which attempt to answer the query at the top of the page. But ChatGPT is arguably doing it better.

These LLM’s have all been trained on data provided on the internet, usually text based, like this blog and obsessively for free, or at least free in return for visitors coming to visit the site. But now there is no need for them to visit the site. The implicit trade is broken.

While we as humans will use ChatGPT and other LLM’s for answering out questions, we’re also destroying the economy for the people who provided the raw information the LLM’s needed in the first place.

What Next?

Younger generations are turning to TikTok, YouTube, Pinterest, Instagram and other services for answers to their questions. I once had a colleague explain to be “Why read how to fit a new tap to my sink, when I can watch someone do it on YouTube”. Eyeballs, are generally starting to move away from search engine and written text to more video based content, and as a result the implicit trade between content producer and trade has shifted too. Rather than this being for visits to a specific website, it’s views for a specific YouTube channel, or TikTok account.

Video content is harder for LLMs to ingest, so the shift in content production from written content to video content is likely to accelerate. It paints a grim future for the discoverability for the written word on line.

In the longer term Video is not safe from this trend. We’ve already seen machine generated images and videos and it’s only a matter of time until video content is ingested mutated, reformatted and automatically generated for the viewer… imagine a world in which you can take a photo of your actual sink, and the AI generated a video for you, showing you exactly what to do, step by step, as you fit your new tap. This will probably be delivered via an AR interface on Meta’s next generation headset.

E-Commerce

We will all need to purchase products and services and these will still be offered by the large search engines, but they will no longer be so easy for a new company to get its storefront listed. This is likely to shift to dedicated shopping portals, Amazon, Alibaba etc, each of which take a cut of the sale. However the product reviews will be gamed, by AI produced reviews, views and other techniques, which will probably still be called SEO (Shopping Engine Optimisation). To put a stop to this stores will need to block reviews posted by accounts which haven’t actually bought the product.

Social Media

The AI wave has already started crashing into Social Media, without verified human users it’s also becoming increasingly difficult to tell human from machine, flesh from robotics. Perhaps this will drive users away from the non-verified sites to some form of “human only” social media network?

The rise in privacy awareness and in the power of controlling what data machines, as well as humans have access to is going to continue to rise.

What does this mean for AI?

So what does this mean for the machines that consume our data? - The models will eventually run out of content to be trained on. As the machine themselves generate new content, social media posts, fake product reviews, view content, blog posts and fake news they will in effect start to pollute their own drinking water. The AI companies will spend increasing amounts of money trying to make sure that their AI’s don’t train themselves on their own, or their competitors output.

The Future

The future is never written, but the words from Peter Gabriel are starting to sound even more prophetic, man indeed is eating the output of the machine, and the machine is eating the output for the man. But it’s the unsung next line which is only a matter of time - when the machine starts to consume its own output.

Ahh, well, we might at well enjoy the track while we wait for the future to write itself…

PS: Yes, the image for this blog post was AI generated.. I completely get the irony… thank you Bing.