The "transhuman vision"

... (out of context stuff dumped here, move along)

Let's talk briefly about the "transhumanist vision":
  1. computer that includes a processormemory (long-term and short-term) and connections to peripherals. The "mind" or general artificial intelligence is part of the software that can run on the processor, along with other more procedural programs, special AIs and essential basic programs (drivers and stuff).
  2. transhuman body -- the peripherals including: sensory devices for visual (eyes) and auditory (ears) observation, substance detection (nose and tongue) and various sensors (skin, ears) among other things, output devices including: display and speech, motion, hands and fingers and 3D printing. Mechanisms to preserve immortality (defense and backup automation mechanisms)
  3. communication system -- including: "telepathic" or "telekinetic" communication with other transhumans and internet-connected devices, internet access including access to new skills and data for learning.
  4. virtual reality platform where a transhuman can behave as a human, multiple humans or any other persona they can create or download -- in a virtual environment that could be private or connected to the internet. A "persona" is a system that collects sensory information in the virtual environment in a particular way, limits the capabilities of the user's interaction with the virtual environment in a particular way or even temporarily alters the user's mind (memories, emotional responses, decision-making instincts, etc.).
  5. Mind transfer and copying -- mind upload (from biological to electronic and vice versa), mind copying (from biological to electronic and vice versa), mind teleportation (from electronic to electronic), mind cloning/backup (from electronic to electronic)
Why did we just do that? Because I want to give an overview of the kind of "data" a general artificial intelligence should be able to collect. OK, obviously the machine learning part is the "mind" or "general artificial intelligence" mentioned in the first point. What can a brain do?

Well, I'm not claiming this is complete, but:
  • Observation processing -- from sensory or memory, coming up with questions and experiments, thought experiments too (linked to transfer) -- segmentation, movement, detection and classification, detecting patterns/faces... "Is this guy an older version of this guy? People tend to ask that question if he is... train to understand things like this. But maybe not always -- e.g. "Are you serious?" ... should be able to distinguish between these cases... "bounded" guess meaning from word... but if fails, then no. Or if it says "misleading semantics"
  • Original creations -- imagination or to peripherals. Simple things like movement to more complicated things: random thoughts, artworks, ideas, lines of thought, mathematical proofs, ... ... confirmation bias except with statistical data we can avoid that
  • Predictions, coming up with explanations... subjective probability
  • Decision-making -- using predictions. Or using reasoning. Can be based on random choice. Can be trained based on previous outcomes, advice of others, etc. Can evaluate that advice.
  • Reasoning
  • In-built loss functions -- emotion (pain, greed, social approval), instinct, co-operation, etc. Controlling instincts. 
  • Learning -- from reading, introspection, whatever... detecting emphasis, reprogramming brain instincts
  • Knowing what to recall: 12*11 = ? (132)
  • Making connections/transfer of learning extremely important -- examples: symmetries, things fall, being big or small, edges and stuff. You don't need to read a thousand computer programs before you make your first one (list steps involved that are transfered over). You don't need to read a humongous number of mathematical proofs before you write your first one. You don't need feedback on a a million mathematical conjectures you generate before you can tell which ones are important. You don't even need to see a thousand examples of a single character to tell what it is or to reproduce it. On a higher level, making connections between ideas, mathematical abstraction. 
  • Narratives, causative structure, mental classification, clocks and models, and other actually "artificial" things

No comments:

Post a Comment