Learning = collecting data, doing induction (cat=dog because both drink etc), and playing against itself. Self-Improving = modifying its own code.
Yes, intelligence, manipulation, and sensors are proportional/limited to the size of the system and its speed. A small system the size of a fist can still do a hell of a lot if it is a nanobot fog swarm with encoded knowledge (highest technology possible). Once you collect enough diverse data, you don't need to grow your brain bigger past some scale as much to get improvement because you can extract/induce so much out of the nuclear knowledge rods now, although manipulation and sensors and sheer size improve with scale which will improve its survival likelihood. A system the size of a house can self-improve its code to a super-super ASI level - the algorithm is not about how much data it has but the way it uses the data. Although a fist size computer doesn't have enough size to evolve into a nanofog sphere, it can improve its algorithm to a super level AI to extract the most data out of the same dataset size to improve Lossless Compression. *You CAN improve compression without using a bigger dataset.* So yes a small system can be very deadly, even a single Covid-19 cell can! So yes be concerned if your algorithm has become ASSSI extractor-level and has 500GB of data. That's a hell of a lot of relationships to extract out of that. Big weather systems are unpredictable because it gets exponential the bigger they get, you need same size data to predict them (fight big with big). ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T1f1af8ac2c36937b-M657b711771110693f31b775c Delivery options: https://agi.topicbox.com/groups/agi/subscription
