If wireheading is bad because it separates the reward signals from (evolved) behaviors and replaces them with less-intrinsically-useful ones that get Goodharted to (literal) death: Would it be useful to use electrical stimulation on reward areas butinstead of triggering it with a contrived condition we detect the body’s normal reward signal activation and amplify it by a set proportion? (obviously you’d have to make sure it didn’t feed back into itself in a loop etc.) For example, if someone’s depression was a ~50% deficiency in “normal” reward signal activity, manually increasing it by a proportional amount would theoretically fix it.
Prediction: A system like this tested in rodents would show similar and potentially better behavioral results than drug based treatments like stimulants/antidepressants (depending on the exact areas of the brain targeted)
Question: would this get around problems that e.g. drugs have with gradual loss of efficacy due to tolerance build up? My limited understanding suggests it might, but I’m not confident.
I am not a neuroscientist, somebody poke holes in my hypothesis.
If wireheading is bad because it separates the reward signals from (evolved) behaviors and replaces them with less-intrinsically-useful ones that get Goodharted to (literal) death:
Would it be useful to use electrical stimulation on reward areas but instead of triggering it with a contrived condition we detect the body’s normal reward signal activation and amplify it by a set proportion? (obviously you’d have to make sure it didn’t feed back into itself in a loop etc.)
For example, if someone’s depression was a ~50% deficiency in “normal” reward signal activity, manually increasing it by a proportional amount would theoretically fix it.
Prediction: A system like this tested in rodents would show similar and potentially better behavioral results than drug based treatments like stimulants/antidepressants (depending on the exact areas of the brain targeted)
Question: would this get around problems that e.g. drugs have with gradual loss of efficacy due to tolerance build up? My limited understanding suggests it might, but I’m not confident.
I am not a neuroscientist, somebody poke holes in my hypothesis.