This creepy sci-fi internet theory may be coming true

Once upon a time, a conspiracy theory emerged in online forums claiming that the internet was run and peopled by robots controlling and limiting genuine human interaction.

It was dubbed the “dead internet theory”. And now, some say it’s coming true.

Online porky-pies, of course, have been around since the beginning, whether it’s a lie about height on a dating website, or an email scammer claiming to be a disinherited prince.

But at first, they were undeniably lies perpetuated by humans.

Some estimates now suggest the majority of traffic online is carried out by automated “bots”, and that the rise of generative and agentic (or more independent) AI is making it harder and harder to tell the difference.

“Ironically, in the early days of the internet, a famous adage (from Peter Steiner of The New Yorker) that made its rounds was ‘On the Internet, nobody knows you’re a dog’,” Dr Marc Cheong of the University of Melbourne’s School of Computing and Information Systems told nine.com.au.

“Nowadays, that maxim can be modified to ‘nobody knows you’re a bot’.”

Read more

Apple targets AI use with new desktop Mac models that cop a solid price hike

Cheong said “public square” websites were filled with “bot-like activity”, including automated “likes”, text comments, “AI slop” clips on YouTube, synthetic images, and even AI-generated songs.

“Currently, there is no clear estimate on how many total posts in the ‘public square’ is attributed to bots, but there are estimates on the degree of bot activity from various perspectives,” Cheong said.

“The technology provider CloudFlare estimates that, based on current trends, more than half of internet traffic to websites come from bots, putting humans in the minority. 

“An industry report from Kapwing, a video-editing company, estimates the number of ‘AI slop’ videos to be over 20 per cent.”

The increasing use of generative AI is also making it harder to distinguish between humans and bots, while also loosening human control over automated processes.

“With the current growth of agentic AI systems - which can independently plan and execute steps to achieve certain tasks, such as automate customer engagement on social platforms - it is more likely that bots will find their way onto the ‘public squares’ of the internet autonomously on behalf of their creators,” Cheong said.

“An interesting story on the latter point is about how AI bots even have their own platform, Moltbook - this was an experiment on how bots can ‘chat’ among themselves independently and autonomously, once let loose into an internet forum.”

Bots responding to bots - the “dead internet theory” in full, if miniature, flower.

Associate Professor Robert Ackland from the Australian National University says the key to keeping ahead of the disinformation race is media literacy.

In his view, the year 2016 was a turning point for worries about online bot activity, with the Brexit vote in the UK and the election of Donald Trump in the US shaking the digital West.

“The concern was that social bots, inauthentic coordination, troll farms, were parts of foreign influence campaigns attempting to shape opinions and influence election outcomes,” he said.

“This generated a lot of research activity (I participated in some of this) but there were challenges: how to identify bots or troll accounts? How to measure their influence?”

These challenges have only grown, he said, between “bots 1.0” which amplified perspectives through retweets, and “bots 2.0″, which actively engaged with real humans online.

But, he pointed out, humans are more than adept at self-selecting their own disinformation.

Online echo chambers, and algorithms directing people away from “cross-cutting” information also contribute to an increasingly misleading online world.

“What is happening online is not the only driver contributing to polarisation or distrust in institutions, yes, but it is a contributor,” he said.

And while humans had a good sense of authenticity, Ackland said AI chatbots were “definitely a new threat vector”.

“To what extent it plays out is anyone’s guess, but generative AI continues to impress in terms of mathematical innovations, providing information and social support, so (it) could impress in terms of conversational persuasion,” he said.

“Yes, mathematical proofs are not the same as discussing the need for gender pay equality, but ultimately, it is all about predictive text, regardless of whether that text is mathematical symbols or other concepts.”

Both Cheong and Ackland agreed that media literacy and online awareness were needed for people to stay ahead of the bot-heavy internet.

“Adopting a healthy dose of criticism is always a good starting point, given the new landscape of content in the online ‘public square’,” Cheong said.

“People should fact-check claims against reputable news agencies or fact-checker websites such as Snopes.

“Checking ‘metadata’ about a purported social media source also helps: watching out for new accounts, those with low followers/engagement, or those not otherwise known as a reliable information source.”

And of course - “if it’s too good (or sensational) to be true, it probably is”.