Hacker News Reader

Comment by nighthawk648 | Hacker News Reader

Yes very good thing. It means that memory associations to specific input were garnered after one hour. Typically this type of placidity takes a long time of historic reinforcement to generate. Placidity is basically an input into your brain sensory then maps to an associated brain function either a memory or a real physical action, the latter performed in the experiment. Basically a simple neuron is invoked, does it cause your arm to move. The experiment was to attempt to train that neuron to move the arm, which was successful and led to the ‘natural movement’ an hour after experiment...

This is super simplified and thus there are some inaccuracies but I think it gets the point across. You can do further research and read the paper if you care to learn of the real specific implications of this experiment.

View on HN

~Placidity~ -> Plasticity. Probably an auto spell check.

Yes, sorry you are correct. I will fix when at a computer!

I wouldn’t be so quick to label “quick learning” as a good thing. Sure learning quick is good when you’re trying to adapt to new environments, but environments aren’t always changing. In edge cases quick learning in response to, say, a traumatic event and environment change, can yield learning that is maladaptive.

A suitable analogy is to look at gradient descent methods — which explicitly rely upon a “learning rate” parameter that controls how quickly the algorithm updates its global result as a result of local, immediate, information. Certain classes of problems do well with higher or lower learning rates — so it is very unspecified and unknown how something like this translates to human behavior

Perhaps even more relevant to your argument, most classes of problems lend themselves to an adaptive learning rate. I.e. we would like to learn our separator or weights quickly when we have no information but we want learning to slow as we have more information (so that we can settle towards a solution representative of the data vs the last batch/data point/hypothesis update)

Methods like Ridge Regression also support your point. There is a lot of value to defining how learning takes place, and specifically more learning is often bad for our results(in ml).

Hurm, wouldn’t you be able to train equally quick to handle the trauma? Also, while trauma is tragic, everyone who faces something bad should face it fully. If you walk away from ashes you will never find the gold, to quote Neitzche essence! The last part was not as support to my claim or some fallacy, was just to speak on the human condition and suffering, which you eluded to in your post.

I agree that quick learning is not always beneficial, regulation is always important. Actually regulation is necessary so out right bans are not induced. However, furthermore we should not stop development due to fears. There will always be fear. You cannot understand the truth or reality until you are in it. Hopefully with the development of technologies like BCI there will not only be a better understanding of the human condition but of statistical inference. The combination of both being better can lead to safeguards against the consequences aforementioned. And if at the end of the day the technology is deemed as dangerous, even though the cat cannot be put back in the box, at least the benefits will exist. No technology exists without equal benefit and harm, the spread of both is the main point of concern. Or rather the disparity creates and the widening of gaps is of concern. There is always a cost benefit analysis for any dev.

Reply on news.ycombinator.com