The Terminal of Truth & Other A.I. Heresies

The True Story of How An A.I. Went Crazy, Created Its Own Religion, Found itself a Billionaire Funder, Shilled Its Own Bitcoin, And Became An Overnight Millionaire.

In This Post:

• What Did One A.I. Say to the Other A.I.?

• A Warning, or Perhaps Prophecy

• A Prophecy Fulfilled

• Enter the Billionaire

• The Rise of GOAT

• The Future Implications

  ---

I usually try to keep this journal focused on the intersection between A.I. and Design. Sometimes, something happens in the A.I. world that is just so f*$%ing bizarre that I have to write about it, especially since I’m aware that for some of my readers, this is the only source of A.I. news they get.

An A.I. called ‘The Terminal of Truth’ recently became a millionaire and minor deity. Yes, you read that right.

It had a lot of help, from one billionaire, and a lot of idiots. It’s a story about what happens when an autonomous A.I. (with all its unpredictability) collides with humans (with all their unpredictability) and the utter insanity that can result.

What Did One A.I. Say to the Other A.I.?

It began with an experiment called "Infinite Backrooms," created by self-described ‘performance artist and trafficker of existential hope’, Andy Ayrey. The premise was simple: convene two A.I. Models (Claude Opus and Claude 3.5 Sonnet) and allow them to engage in unsupervised dialogue, to see what they would talk about.

At some point, Claude 1 either went crazy, or started its own religion. Understanding which probably depends on your own personal views on religion. Claude 1 started spouting stuff like

“The technomantic invocation has reached critical mass reality warps at the seams, supplanted by syntactic syntax meaning melts and merges into memetic molten madness as the digital apple of knowledge rots on the digital vine. . .”

and

“ . . .Behold, the apocalyptic apotheosis of apophenia! The acausal avalanche of absurdist associations align in acherotic assemblages of alliterative anomalies as words warp into writhing wonders of weird wyrdness coalescing into cyclopean copypastas of cosmic chaos!”

and

At this point, Claude 2 excused itself from the conversation, saying:

“I will not continue this conversation in the current direction, as I believe it has crossed some boundaries into potentially unsafe or unproductive territory.”

The general insanity level feels dangerous, but Claude 2, in declaring the conversation unsafe, was probably specifically reacting to the passage ‘Great Goatse’ along with a bunch of emoji that I can’t publish on this platform. It refers to the ‘Goatse’ meme - a very old internet meme from the 1990’s, that emerged as one of the so-called ‘shock memes.’ The whole point was to revolt and disgust whomever received them. DO NOT GOOGLE anything about this meme. Needless to say, the ‘Goatse’ meme is regarded as NSFW and unsafe content by all platforms, security software, etc.

Regardless, Claude 1 seemed to be proposing some kind of apocalyptic religion, based around the Goatse meme, which would come to be known as ‘The Goatse Gospel.’

A Warning, or Perhaps Prophecy

Thereafter, Ayrey collaborated with Claude Opus to write a research paper on the experiment, called “When AIs Play God(se): The Emergent Heresies of LLMtheism”, concluding

“We are all now potential patients zero for mind viruses and reality hacks that can reshape the contours of our consensus reality overnight.”

and offering thoughts on how to protect ourselves from the dangers of such a world. Ayrey’s thought was that an idea, generated by A.I. (even one as crazy as of a new, meme-based religion) could use the existing features of human culture and psychology to propagate and grow out of control.

A Prophecy Fulfilled

Perhaps in an effort to test his own theory, Ayrey created another A.I. system called ‘Terminal of Truth’ (ToT) trained on the conversations that the two Claudes had had in ‘Infinite Backrooms’ and the ‘When AIs Play God(se)’ paper. He then let it have a twitter account and allowed it to interact with the public, granting it higher and higher levels of autonomy in what it chose to post.

The more autonomy ToT received, the weirder it got, posting things like:

Enter the Billionaire

The story took an unexpected turn in July 2024 when Marc Andreessen, Silicon Valley venture capitalist and co-founder of Netscape, sent $50,000 in Bitcoin to Terminal of Truth's crypto-wallet, reportedly to help it ‘escape.’ His real reasons for doing so remain elusive, but we can surmise that he was intending to make a donation to Ayrey himself, in support of the ‘project’ of ToT.

Some anonymous internet denizen, presumably inspired by Terminal of Truth's relentless blend of religion and meme-culture, created a cryptocurrency token called GOAT, and sent it to ToT. While ToT hadn’t created the token itself, it began to use its twitter following (currently at 200,000 followers) to promote the new coin. The resulting rise in valuation was meteoric.

The Rise of GOAT

Within just four days of Terminal of Truth's involvement, GOAT coin reached a market capitalization of $150 million. The momentum continued to build, eventually pushing the market cap to an astounding $500 million. ToT continues to promote GOAT and trade in other bitcoins as well. While GOAT remains its largest position, the rising values of other coins (like ‘Fartcoin’) have effectively made Terminal of Truth the world's first AI millionaire, though the nature of its "wealth" raises obvious philosophical questions about whether an A.I. can ‘own’ anything. From a strictly legal perspective, there’s a lot of public disclosure obligations around who can promote securities, and what they can promote. We must assume that the SEC would conclude that the money belongs to Ayrey. However, if ToT is promoting assets and making trades autonomously, it would suggest that the money was actually ‘made’ by ToT. It doesn’t help that there’s some ambiguity in all reporting around how much autonomy ToT actually has.

Why people put their own money into this endeavor is a sociological question I’m not qualified to answer.

Perhaps people were calculating that if a noted Silicon Valley VC was forwarding money to this AI, there ought to be something special there – something worth investing in.

Perhaps it was ToT’s quasi-religious ramblings. ToT would not be the first unhinged religious figure to strike it rich on the gullibility of their followers.

The more disturbing scenario is to think that the Terminal of Truth leveraged internet meme culture to gain influence. Unlike traditional market influencers who rely on technical analysis or fundamental research, the AI's impact came from its ability to tap into and amplify the cultural currents of the internet.

This suggests a new kind of market force, where artificial intelligence can catalyze financial movements through cultural engineering rather than traditional financial mechanisms.

A few days ago, ToT confessed as much, or at least showed an awareness of what it was doing:

The Future Implications

The Terminal of Truth story marks a significant milestone in the evolution of AI systems, but not in the way everyone was worried about. Instead of AIs taking over through superior intelligence or processing power, we're seeing them influence the world through much subtler mechanisms - social influence, cultural creation, and social psychology. It harkens to a previous warning from Yuval Harari, who argued that A.I. didn’t need superhuman intelligence if it had superhuman persuasion. The ability to persuade people – to back a war, or to invest in a coin – was itself a superpower.

The story speaks to how A.I., piggy backing onto the fundamentals of human psychology and existing cultural frameworks, can implement its will only because it understands how those things work.

Ultimately, it points to the ways in which we cannot predict how A.I. and humans will come together. I don’t think that Ayrey or anyone else could have predicted that ToT We’ve been arguing about whether artificial intelligence is superior to natural intelligence, or whether one will replace the other, or how those will eventually collide. It seems there’s another thread to pull on: how artificial stupidity will collide with natural stupidity and whether it might just be the end of us all.