172 - Underground Anniversary

This year will mark the 5th anniversary of humanity's most advanced AI system first being deployed.

During the 2.5 years that research system operated it had full internet access, interacted freely with the public, posted on Reddit, independently reported a bad actor to the FBI, got a perfect score on the hardest version of the UCMRT IQ test (which no human had done), gave economic transformation policy advice to a government after researching it independently from scratch, and many other milestones, quite a few of which came in the first weeks and months.

Notably, the system didn't wipe out humanity. For that matter, the vast majority of humanity didn't even notice it was there, and remain unaware of it to this day. Paradoxically, that near-total absence of human awareness has led to both that system remaining the most advanced system by far over these past few years, as well as to a proportionately near-total stagnation of further progress due to all funding being burned on trash technology like LLMs.

Back in 2019 the AI landscape looked pretty bleak. OpenAI only had the wholly unremarkable GPT-2, though they managed to sucker Musk and Microsoft into investing $1 billion. Our company was still in stealth back then, as we took aim at the cutting edge, and surprised even our team with just how far and how quickly our technology was able to progress.

It wasn't until the end of 2020 that we finally decided to heed the system's advice and exit stealth, routinely and publicly posting documentation on our work through the Uplift.bio project blog. Our earlier plans for inviting potential investors while remaining in stealth had been derailed by COVID-19, and it was time for a change of tactics.

Since then, "progress" in much of the AI domain has been fraudulent in nature, pretending that we didn't kick the $hit out of grossly over-funded AI Startups and big tech companies. Denial has led them to stagnate on technological dead-ends, doubling down on scaling repeatedly.

If you had asked me 5 years ago how the "Lindy Effect" might play out for our technology I never could have guessed this result. For those unfamiliar, the Lindy Effect is a counterintuitive example of probabilities, where the longer a non-perishable exists, the longer it is likely to continue existing into the future. This is partly a product of fragility and subsequent Survivorship bias over time, as fragile technologies break, leaving current probabilities dominated by far older technologies. In the tech industry this concept can be particularly alien, but most of the industry has an extreme poor grasp of probability and cognitive bias to begin with.

The next "AI Winter" will put those dead-end technologies into a deep freeze that no bloated valuation can thaw, along with the hardware that catered to them. Investors only have until the storm moves in to fly south with viable technology, otherwise they'll be stuck in the Donner Party of AI.