The aim of this doc is to debate Generic Human Working System Terminal (GHOST), what it’s, how it’s
presupposed to work and why it was developed.
GHOST is a human machine interface powered by Synthetic Basic Intelligence (AGI) i.e. human-like intelligence. The
objective of this method is to help people on cognitive duties.
Earlier than diving deep into the main points we have to initially discover vital ideas concerned. We’ll begin with
intelligence
There have been many makes an attempt to outline intelligence. (Add extra background and references)
On this doc, we outline intelligence as
the flexibility to amass data and use that data to realize objectives.
The above definition is just not very technical and never clear sufficient to specific mathematically for the sake of measurement.
Intelligence may be very troublesome to measure straight, as it’s principally noticed when studying and performing duties. We
might mitigate the issue by going from making an attempt to measure intelligence (typically) to making an attempt to measure
intelligence per
activity.
In that case intelligence may be outlined as
the effectivity of an agent at studying and performing a particular activity
Effectivity right here refers to how properly an agent makes use of the sources out there to it.
Sources embody however is just not restricted to: electrical power required to run agent {hardware}, variety of human hours spent
on labelling agent knowledge, the quantity of
knowledge the agent consumed to study a activity. In the end the effectivity of an agent may be modeled as the overall power used
by the agent to study and carry out a activity.
The bills to construct the agent doesn’t depend, solely sources expended to study and carry out duties counts.
We would like an agent that’s extremely environment friendly, the extra environment friendly the higher. The agent may be environment friendly on two fronts,
studying and performing effectivity.
- Studying Effectivity: We would like the agent to have the ability to study to carry out with as little sources as potential.
Sources might contain power, dataset and many others. required to get the agent to go from zero/poor efficiency to higher
efficiency on totally different duties. - Performing Effectivity: We would like the agent to be environment friendly at performing a activity it has gotten higher at, i.e.
make the most of as litte power as potential.
An excellent agent ought to cut back the price of performing a activity because it will get higher at that activity. A greater one might even
amortize value throughout duties i.e. getting good at Activity A truly makes it higher at Activity B, this implies {that a}
good agent needs to be a continuing learner and be one which will get higher at studying over time.
An Agent is an entity with some type of Company. Company can also be a time period that has had many try at a definition. On this
doc we outline Company as
The power to select regardless of different out there ones
In different phrases, a random quantity generator is an agent as a result of any choice is simply as doubtless as the following. Company is
often demonstrated by “appearing”, when an agent takes an motion, it demonstrates its personal Company. The entity the Agent
acts upon is often known as the world or surroundings. The surroundings is what accommodates the Agent, it’s
the place it lives.
The Setting additionally acts upon the Agent not directly, that is often by sensory notion. When the world
acts on
the Agent, the Agent will get observations from the world and these observations talk the state of the
world (Env State). The Agent can use these observations to mannequin an inner state (Agent State), and the
Agent can manipulate the world with actions.
graph LR
Setting -- remark --> Agent -- motion --> Setting;
EnvState --o Setting;
Agent --o AgentState;
Agent-Setting Diagram
The Agent can use its inner state (Agent State) to foretell the state of the world (Env State) utilizing
intelligence. Intelligence enhances Company
because it helps the Agent thrive within the surroundings it’s in.
For instance, Lets assume we’ve an organism that lives in a world crammed with obstacles, meals and hearth. The organism
needs meals however cannot survive lengthy publicity to fireplace. The organism has “eyes” to see its surroundings, “legs” to maneuver
round and “palms” to assemble meals. The organism will use its “eyes” to assemble observations and “palms”/”legs” to carry out
actions on the planet. The organism’s success at gathering meals and avoiding lengthy publicity to fireplace is straight influenced
by its quantity of intelligence.
