The AI learning process should include or require all of
the same or exact 5 senses that humans learn with. Taste,
smell, touch, hear and see. I listed hear and see as the
last two senses because hearing and seeing (with microphones
(simulated hearing) and cameras (simulated seeing) seem to
be a popular limit to the learning of AI at this point.
A.I. (artificial intelligence) is too machine-like when it's
learning is so limited to two senses.
If we want to be able to trust AI, it should mirror us and
AI's learning techniques need to include the same senses that
we humans use to learn.
So what is missing? Touch? Yes, for a start! Touching, or
"feeling by touch" might be used by including very
small/short-range microphones in a prosthetic (robotic) hand
but spaced before and after every joint in an artificial hand,
maybe three or four per finger, and ten or more tiny microphones
in the palm. Those microphones can be short-range and analyse
the surface of objects to determine by examining the sound the
microphones pick up as smooth, rough or something in between.
Tiny infrared (IR) diodes can gauge the level of hot or cold.
Smell and taste are important as well. A smell sense will
require more delicate sensors than taste (a nose and tounge)
but those senses are related. One sense, taste samples large
amounts and the other, smell senses molecules compared to
taste. The same sensor might be used at different sensitivities.
What else keeps modern machine A.I from seeming or acting
more human? One time I was watching the "Charlie Rose" (a
conttroversial charactor) show in a brief series about the brain.
That brief chapter or unit about the brain, there was some old
(black and white) film footage of a patient and brain surgeon.
The brain surgeon had implanted electrodes and various parts
of the patient's brain and found that in the patient's or
volunteers brain that there were three (3) areas that responded
(became active) that recognized (pattern recognition) three
different types of lines. Either the brain surgeon or a nurse
or helper had simple drawings of a horizontal line, a vertical
line and a diagonal line. Each type of line drawing caused
activity on their sensors (which sounded like a geigor counter)
for each type of line in a different part of the subject's brain.
It is known and understood that different parts of the brain
specialize in different patterns of recognition, such as the speach,
visual and facial centers. If the brain also recognizes line
positions and configurations, then there will could be more,
seemingly simple recognition centers that we have not realized
or considered.
Human-like A.I. has not been studied enough and should
get a HIGHER priority for study. Machine-like (limited) AI
should take a back seat to machine-ai and we should spend
more time developing artificial intelligence (ai) that is
human-like and we can identify with and possibly be more
tolerant with.
The question is: Do we have enough knowledge about how human brain works in order to an "attempt" to create a more human AI? I don't know about the most recent studies about human brain, but I think we don't know a lot or at least not enough to replicate this behavior.
So, it is valid of course, the need to create a more human AI, but I think from technological perspective we are not that far yet.
Nice piece @yesiknow!
@]lion200: The question is: Do we have enough knowledge about how human brain works in order to an "attempt" to create a more human AI?
Hi, lion200 and I appreciate you and your reply! :D I like your question about. "Do we have enough knowledge about how human brain works in order to an "attempt" to create a more human AI?"
Doubtful that we have "enough" knowledge. I think we are still learning about the human brain.
We have to keep studying how the human (or others) brain operates. The current AI examples that we all use are lacking or not progressing as humans do. Current examples of A.I. seem to be too mechanical. At least when we "attempt" to replicate behavior, we learn which methods somehow work. Programmers or mathematicians created the first artificial neural networks. Did they have enough knowledge about how the brain works? I doubt it! But those programmers and mathematicians that coded the first ANN's came up with methods that often get some of the work done that saves us some work and time.
You're welcome @yesiknow. Thanks for this good topic.
Indeed, there is already something on which scientists and programmers work on. But we are in a very early stage by means of "more human AI". That can take a very long time, since we simply lack knowledge.