PS. One of the comments in the Daily Beast article suggested that
deep-fakes are used for signal compression... I have experienced 2
examples of this which are compelling:
1. text/search-completion/anticipation where it feels as if the system
already knows what I am likely to want to type next (I sometimes
think it is training *mildly* on my own text, not just on a larger
corpus).
2. hand/eye/body tracking. Early VR was hampered badly by lag in
tracking caused both by transmission/queuing and processing delays
so a workaround was to build an anticipatory model which
extrapolated trajectories of (for example, hand) motions... when
overdone it gave you the sense that your hand was attached
elastically to the VR rendering of a pointing device with inertia
which was better than the raw *lag* experience. head trackng was
even more critical because cybersickness but in some ways easier
because it was more constrained (6DOF neck/body movements more
constrained than the nominally hip/torso/shoulder/elbow/wrist DOF of
the hand). I always wondered if a properly coupled system (in the
inter-reality (Guintatas) sense) would lead to entrainment similar
to what (I think) martial artists, gymnasts, dancers experience when
they train? This is probably a very real issue/experience in
robotic exoskeletal development?
On 6/24/22 7:50 AM, Steve Smith wrote:
https://www.thedailybeast.com/amazon-unveils-new-ai-tool-for-alexa-to-turn-a-speech-snippet-into-a-voice
Marcus confronted me a month or more ago with something along the
lines of " who says learning isn't just imitation? ", and it *really*
hit home. Play seems to be a lot about mock-work and mock-fighting
and it seems widely accepted that a great deal of learning happens
through emulation of the expertise of others.
When I consider those who have passed in my life, I think I much
prefer to hear-through-memory their voices (and visages) over
recordings. Synthesizing voices or visages seems to compound the
uncanny valley effect. Bad enough to get it eeriely-nearly right
with something less visceral than the affect of a loved one since passed.
During the first gulf war (1991) I worked with a voice/speech
researcher who "went dark" for about a year. He remained cagey about
this work even after he came out of that period of silence and
misdirection, but eventually the general nature of the work was
declassified (though not details) and it was about synthesizing voices
by modeling the vocal tract, and the application that made it so
skitchy was battlefield communications of Saddam Hussein. This was
still in an era of analog communication with various analog scrambling
techniques standing in for what we do today with digital encryption
and authentication. I have no idea if it was ever deployed or
effective operationally, but the examples he used for demonstration
with well known voices. I doubt anyone could argue that this speech
synthesizer in any way "thought" like Saddam Hussein, but the larger
organizational unit (Iraq army), might well "act as if" the intention
injected by the synthesized voice was part of their collective psyche.
Back to the original premise of whether a well-enough
practiced/trained learning classifier system has any emergent
properties that are parallel to animal/human cognition. This seems
entirely up in the air to me, but Glen's recent rant/rave about "what
it's like" and splitters v smooshers was mildly compelling to me.
A big shift in my own perception of self/change was when someone I
respected offered the idea of "trying it on" and "acting as if" as a
model for personal transformation. In retrospect it seems obvious
that by emulating some *one* (real or idealized) one might actually
become *more like* that person (real or idealized). Of course, this
is about a change of affect/self/consciousness, not emerging from
whole cloth (sand?).
-. --- - / ...- .- .-.. .. -.. / -- --- .-. ... . / -.-. --- -.. .
FRIAM Applied Complexity Group listserv
Fridays 9a-12p Friday St. Johns Cafe / Thursdays 9a-12p Zoom
https://bit.ly/virtualfriam
to (un)subscribe http://redfish.com/mailman/listinfo/friam_redfish.com
FRIAM-COMIC http://friam-comic.blogspot.com/
archives: 5/2017 thru present
https://redfish.com/pipermail/friam_redfish.com/
1/2003 thru 6/2021 http://friam.383.s1.nabble.com/
-. --- - / ...- .- .-.. .. -.. / -- --- .-. ... . / -.-. --- -.. .
FRIAM Applied Complexity Group listserv
Fridays 9a-12p Friday St. Johns Cafe / Thursdays 9a-12p Zoom
https://bit.ly/virtualfriam
to (un)subscribe http://redfish.com/mailman/listinfo/friam_redfish.com
FRIAM-COMIC http://friam-comic.blogspot.com/
archives: 5/2017 thru present https://redfish.com/pipermail/friam_redfish.com/
1/2003 thru 6/2021 http://friam.383.s1.nabble.com/
