This concept is Vaporware.
Using a neural model based on Creatures, it appears to be possible to construct alife starships which roam the starlanes, flexibly adapting to situations they encounter.
Sensory Inputs
Every object in such a simulation would have a class ID (or name), which serves as sensory data to the ship. In addition, incoming transmissions (simple ones, that is) could also be sensory data, and in fact probably serve to direct the focus of a ship. All sensory data is fed to the computer. More on this later.
Internal Feedback
A ship is a simple network of components, connected together to mainly regulate the flow of (1) fuel, and (2) energy. But 'crew' also qualify as components, in that they can be directed as required to meet specific needs of the ship as a whole. To this end, the components have sensors attached to them, which fire unique feedback pulses when specific conditions are met -- such as an insufficient amount of jump fuel. All feedback is sent to the computer. More on this later.
Motivations
Next, a ship has things that for our purposes may be called 'motivations'. They are goals or purposes which drives a ship to leave port. Example drives include simple ones, such as to be fuelled up, be repaired, or explore. All motivations are fed to the computer. More on this later.
Reactions
Reactions are anything a ship does in the physical world -- perhaps as simple as directing a particular crew member to a particular activity. The crew member behaves according to its job responsibilities and perhaps the current focus at hand. So a ship receives strong input from an unvisited star system, and has a strong desire to explore, and so activates the Astrogator, who 'activates' the star system (sets course) and activates the jump drive. If the ship's focus were on the jump drive instead of a star system, and yet still activated the Astrogator, the Astrogator would probably just rebuke the ship's computer and retire back to his quarters. Alternately, if the computer didn't like what the crew member did, it could hire a replacement at a starport. I wonder if a ship could be bred which spaced a recalcitrant crew member?
Biochemistry
I don't have a lot to describe here, but thankfully a starship can be simply modelled. It has two major interactions. First, the fuel tank raises a sqawk to the computer if the fuel purity is too low. The ship's programmed instincts are to turn on any refineries available if this alarm goes off, and to do other things, such as open the fuel scoops when there's an ample supply of H2 outside the hull. Similarly, sensors are hooked up to the Power Plant and drives. Additionally, all components have a Damage indicator. All sensors feed to the computer. More on this coming up.
I'm thinking that the starship anatomy isn't complex enough to warrant anything approaching a true biochemical model. It simply has to be flexible enough to cope with redundant or missing systems, within reason.
Computer
The ship's computer is a connection of neurons, divided up into modules. A stimulus module handles all external inputs; a feedback module handles all internal inputs; a motivation module handles goals, needs, drives.
Central to the model is a cognitive module, where neurons from the input modules join up and feed forward to the reaction module, where the computer decides who to activate in response to its various inputs. Of course, neurons may have differing evaluation functions, and have a current-state as well as a rest-state, and a relaxation function that moves the state to rest over time. Dendrites come into play here, too, connecting a signal to a neuron in different ways. I'm glossing over lots of detail here.
Neurons probably will only need an average of 5 connections, and a ship probably needs less than 1000 neurons. A decay function will free up unused neurons and dendrites for re-use. Yes, these starships will be able to forget, but hopefully they forget very specific and less useful information, and keep the generally useful connections that are more commonly used. But this principle of decay also makes it possible to reform hardened corsairs into well-adjusted traders. Yes, a ship can learn to stalk other ships, especially if it has a strong drive for wealth and likes hurting other ships.
Initialization
A ship's neural configuration begins with a relatively short descriptor of its initial network links. The computer is expected to modify and add connections as it experiences different situations. The assumption is that the initial number of connections will be very few, and the rules for creating new connections will be responsible for 80% (say) of the hard work.
However, since the ship's lifetime of connections are expressible via a descriptor, it becomes possible to export this data and load it into another computer, essentially giving it the experience of the former. Granted, the ship's physical configuration may be different, and even computer module configurations may vary, but the value of tried-and-true connections can be great.
Performance
Creatures performed acceptably with ten Norns on a Windows 95 machine running at 366mhz. This ship model is simpler than the original Creatures model, and given that machines are faster nowadays, it seems likely that performance will be better.
Using a neural model based on Creatures, it appears to be possible to construct alife starships which roam the starlanes, flexibly adapting to situations they encounter.
Sensory Inputs
Every object in such a simulation would have a class ID (or name), which serves as sensory data to the ship. In addition, incoming transmissions (simple ones, that is) could also be sensory data, and in fact probably serve to direct the focus of a ship. All sensory data is fed to the computer. More on this later.
Internal Feedback
A ship is a simple network of components, connected together to mainly regulate the flow of (1) fuel, and (2) energy. But 'crew' also qualify as components, in that they can be directed as required to meet specific needs of the ship as a whole. To this end, the components have sensors attached to them, which fire unique feedback pulses when specific conditions are met -- such as an insufficient amount of jump fuel. All feedback is sent to the computer. More on this later.
Motivations
Next, a ship has things that for our purposes may be called 'motivations'. They are goals or purposes which drives a ship to leave port. Example drives include simple ones, such as to be fuelled up, be repaired, or explore. All motivations are fed to the computer. More on this later.
Reactions
Reactions are anything a ship does in the physical world -- perhaps as simple as directing a particular crew member to a particular activity. The crew member behaves according to its job responsibilities and perhaps the current focus at hand. So a ship receives strong input from an unvisited star system, and has a strong desire to explore, and so activates the Astrogator, who 'activates' the star system (sets course) and activates the jump drive. If the ship's focus were on the jump drive instead of a star system, and yet still activated the Astrogator, the Astrogator would probably just rebuke the ship's computer and retire back to his quarters. Alternately, if the computer didn't like what the crew member did, it could hire a replacement at a starport. I wonder if a ship could be bred which spaced a recalcitrant crew member?
Biochemistry
I don't have a lot to describe here, but thankfully a starship can be simply modelled. It has two major interactions. First, the fuel tank raises a sqawk to the computer if the fuel purity is too low. The ship's programmed instincts are to turn on any refineries available if this alarm goes off, and to do other things, such as open the fuel scoops when there's an ample supply of H2 outside the hull. Similarly, sensors are hooked up to the Power Plant and drives. Additionally, all components have a Damage indicator. All sensors feed to the computer. More on this coming up.
I'm thinking that the starship anatomy isn't complex enough to warrant anything approaching a true biochemical model. It simply has to be flexible enough to cope with redundant or missing systems, within reason.
Computer
The ship's computer is a connection of neurons, divided up into modules. A stimulus module handles all external inputs; a feedback module handles all internal inputs; a motivation module handles goals, needs, drives.
Central to the model is a cognitive module, where neurons from the input modules join up and feed forward to the reaction module, where the computer decides who to activate in response to its various inputs. Of course, neurons may have differing evaluation functions, and have a current-state as well as a rest-state, and a relaxation function that moves the state to rest over time. Dendrites come into play here, too, connecting a signal to a neuron in different ways. I'm glossing over lots of detail here.
Neurons probably will only need an average of 5 connections, and a ship probably needs less than 1000 neurons. A decay function will free up unused neurons and dendrites for re-use. Yes, these starships will be able to forget, but hopefully they forget very specific and less useful information, and keep the generally useful connections that are more commonly used. But this principle of decay also makes it possible to reform hardened corsairs into well-adjusted traders. Yes, a ship can learn to stalk other ships, especially if it has a strong drive for wealth and likes hurting other ships.
Initialization
A ship's neural configuration begins with a relatively short descriptor of its initial network links. The computer is expected to modify and add connections as it experiences different situations. The assumption is that the initial number of connections will be very few, and the rules for creating new connections will be responsible for 80% (say) of the hard work.
However, since the ship's lifetime of connections are expressible via a descriptor, it becomes possible to export this data and load it into another computer, essentially giving it the experience of the former. Granted, the ship's physical configuration may be different, and even computer module configurations may vary, but the value of tried-and-true connections can be great.
Performance
Creatures performed acceptably with ten Norns on a Windows 95 machine running at 366mhz. This ship model is simpler than the original Creatures model, and given that machines are faster nowadays, it seems likely that performance will be better.
