An inflection point occurred in 2012 when a convolutional neural network named Alex Net achieved groundbreaking performance on the ImageNet competition. This event triggered the golden age of machine intelligence that we are currently witnessing, featuring a Cambrian explosion of new neural network approaches able to overcome purely statistical methods with high dimension, representation-based approaches.
Only a few years later, in a less attention-grabbing moment in the history of AI, the first-ever unsupervised learning method — trained without labeled inputs — would surpass supervised learning regimes in language understanding. AI was becoming more general, transferable, and purely computational.
In combination, these two monumental events achieved something quite profound: they demonstrated that we can use computers to produce a new element from nature without human intervention, a computational element that is generally valuable, transferable, and universally fungible.
Machine intelligence has become a commodity.
The effect of this tectonic shift, however, has not yet been fully realized. Intelligence is still being framed by the same academic substrate which birthed it in the 1950s, chugging along within the same zeitgeist, where the growth of AI is measured by humans and transferred by paper writing.
But a new paradigm for AI is impending. The invention of digital trust, a.k.a. Bitcoin, showed us for the first time that we can organize computing around a protocol in order to create something bigger and better than anything a centralized organization is capable of.
Bittensor fulfills this technical bridge.
Using digital trust to create markets for AI, Bittensor creates the first-ever cognitive economy that prices and facilitates the transfer of pure machine intelligence between consumers and producers. We are creating the first neural internet, connecting researchers directly and incentivizing them to help each other produce a more powerful intelligence source for the web.