Use the contact page to request a wiki account

BigHairyAudaciousGoal

last edited 3 years ago

Big Hairy Audacious Goal: BHAG

If LExAu succeeds it is possible to create a vehicle for the mind: the mind would no longer have to die together with the body.

Stairway to heaven

"Heaven" refers to the goal of the project, not the state of minds that use the supposed end result of the project. The CoreValues behind the project should be compatible with any religion, even the Skeptic. one.

1. Create an inference engine that measures knowledge. An inference engine maintains and improves an image of how two systems interact. The image is constructed in such a way that it shows how much the engine learned so far.
2. Create an autonomous system. An autonomous system can be constructed by combining an inference engine with a learning policy. A learning policy generates actions that optimize a criterion that is internal to the autonomous system, but is indirectly all about its environment. The actions generated by a learning policy are aimed at maximizing the opportunities for the inference engine to improve the image of the interactions between the autonomous system and its environment.

   This is the central idea of LExAu. It might seem hard to follow, but this is what the project is all about. The complex interplay of (sub)systems at different levels:

   • The autonomous system as a whole
   • The environment
   • The image (or model) of the interaction
   • An inference algorithm
   • An internal (or subjective) value function
   • A learning policy

   The first autonomous system might be severly limited, e.g., it might have a short term memory of only two symbols or something like that. It would not necessarily be useful. It would just prove the point that it is possible.

3. Increase the complexity of the various parts of the autonomous system to get to the point where it can use a cooperative environment to do just about anything. Just like a child can emulate any Turing machine with a piece of paper and a clear explanation of the rules.

   By the time we reach this stage it should be complex enough to play Turings emulation game and perform the Turing test.

4. Hook the autonomous system up with a control circuit. For example a simple thermostat. Make sure to incorporate a manual override and a signal that tells the autonomous system when the override is used. Just the 'curiosity' about how long the override can be postponed should be enough to motivate the autonomous system to lear to control the heater.

   At this point the results of the project become commercially attractive.

5. Hook up the autonomous learning system with neurological tissue. Following the same principles as the thermostat example above, an autonomous system could be used to replace damaged nerves that used to control, e.g., a leg.
6. Increase the bandwith of the interface with neural tissue dramatically to reach the bandwidth necessary to replace something like the nerves that connect the left and right hemisphere of the brain.

   By that time an attached autonomous system would be able to participate in all main functions of the brain. Moreover, because the brain is an autonomous system too, the setup would be symmetrical. Both autonomous systems, one natural and one artificial, would cooperate as a tightly coupled team.

   If one of the members of such a hybrid system were to deteriorate slowly, the other member would take over its functions. So the mind of an ageing brain could move to its artificial counterpart.

   If a fully developed artificial autonomous system were connected to a blank human brain the process could work in the opposite direction: the artificial system would team up with the undeveloped brain to combine resources.

   The previous steps could be used to move a mind from one body to another, hence the goal: a vehicle for the mind.

7. If cloning works for humans by that time, it would even be possible to reenter your own body in a way.