Skip to content

Latest commit

 

History

History
33 lines (24 loc) · 2.21 KB

singularity.md

File metadata and controls

33 lines (24 loc) · 2.21 KB

Singularity

At a given point we do not want to program but instead have our needs anticipated by a higher level system.

Overview

For decades we rebuild the same systems and services over and over. We amass a wealth of technology advancements only to tear them down or leave them to be governed by large siloed corporate entities. When the network exists and we have surfaced 1000+ base motor functions as services, when we have created a language to stitch these together as easily as the english language, and when we have the inputs from requests, images, audio, video, sensors then we can begin work on a system that can harness this capability as a general AI.

Design

Intelligence is an emergent behaviour in the universe and the creation of such a thing is no different. Base motor functions begin as the ability to move and react, to influence the world around even without the knowledge of what that means. Intelligence is something that must be given raw sensory input and the nervous system to process them. A brain is nothing more than a massive event processing system with state and pattern recognition.

A singularity is not something you can just create, it's something that must emerge, but if you understand the properties required to nurture such a system into existence then there is the possibility of producing it.

Google and others are attempting to move beyond vertical based deep learning to a more general approach. Jeff Dean has spent his career building the systems that have shaped Google from the foundation up. This did not start with general AI but instead raw data processing at scale. GFS, BigTable, Borg, MapReduce, Search, Ads. All these led to Google Brain and what came after related to TensorFlow.

Outcome

Micro has the opportunity. Not today (2020) but in a decade from now to build the next intelligent system based on an emergent approach. We cannot compete with Google head on based on resources (both in compute and cash) but if we crowdsource this effort using a network, if we crowdsource the compute, the services, then we have the chance of building a completely open general AI not governed by corporate incumbents.