Security Robot
6 views
Skip to first unread message
* Peter Bock
Jul 23, 2010, 1:53:20 PM7/23/10
to The ALISA Community
Hi all:
A friend sent me this a few days ago. I thought I should put it up
here for all of you to taste.
http://www.myarcherfish.com/pages/meet-solo#utm_campaign=ThinkingCamera
The image processing capability of the Archerfish seems very similar
to an ALISA demo that Dan Phifer put together when he was building and
testing JavALISA about 5 years ago in Project ALISA while working on
the DTRA Project. I have a video of this demo in the Files section
below (ALISAObjectTracking.avi.zip). Take a look at it! I often show
this when I give lectures about ALISA, and my audience is always very
impressed, and appropriately so! It is a very clever piece of work!
I do not think, however, Dan's ObjectTracking demo (or the Archerfish
Thinking Camera) is an example of AI, and I am sure Dan would agree
wholeheartedly. Neither machine applies any symbolic understanding of
the situations they see: They do not understand the symbolic concepts
of "safe" or "intrusion" or "trouble-in-River-City" or "deep bandini",
but only a much simpler and entirely SUB-symbolic concept: "an
unrecognized texture on an object moving relative to my sensor". Even
the most primitive creatures on our planet have a similar sub-symbolic
cognitive ability that they use to trigger one of the 4 F's, which
works often enough to statistically keep the species in a stable range
between extinction and over-population (except in the Gulf of Mexico).
I doubt if Archerfish is doing anything more sophisticated, if that
much.The maximum amount of high-speed memory that can be accessed by a
single core today without melting into a puddle of silicon is still
only measured in hundreds of GBs, about the equivalent to a frog. The
price tag on the Archerfish Thinking Camera ($399) implies MUCH LESS
RAM, probably putting it in the nematode range. Of course, with all
due respect, nematodes do have rudimentary sensors, and their species
is quite stable!
The error rate of the Archerfish with so little intelligence (capacity
for genetic or learned knowledge) is likely to be outrageous, so I
would suggest limiting it to security applications where very little
is at stake --- perhaps guarding the cookie jar or a windowless
bathroom in a house in a gated community in Monterey.
The good news is that we are now probably only 20 years away from the
real thing: Personal AI. The bad news is that when it becomes smart
enough to do the job well, it will also be smart enough to demand a
reasonable salary and fringe benefits. :-)
A friend sent me this a few days ago. I thought I should put it up
here for all of you to taste.
http://www.myarcherfish.com/pages/meet-solo#utm_campaign=ThinkingCamera
The image processing capability of the Archerfish seems very similar
to an ALISA demo that Dan Phifer put together when he was building and
testing JavALISA about 5 years ago in Project ALISA while working on
the DTRA Project. I have a video of this demo in the Files section
below (ALISAObjectTracking.avi.zip). Take a look at it! I often show
this when I give lectures about ALISA, and my audience is always very
impressed, and appropriately so! It is a very clever piece of work!
I do not think, however, Dan's ObjectTracking demo (or the Archerfish
Thinking Camera) is an example of AI, and I am sure Dan would agree
wholeheartedly. Neither machine applies any symbolic understanding of
the situations they see: They do not understand the symbolic concepts
of "safe" or "intrusion" or "trouble-in-River-City" or "deep bandini",
but only a much simpler and entirely SUB-symbolic concept: "an
unrecognized texture on an object moving relative to my sensor". Even
the most primitive creatures on our planet have a similar sub-symbolic
cognitive ability that they use to trigger one of the 4 F's, which
works often enough to statistically keep the species in a stable range
between extinction and over-population (except in the Gulf of Mexico).
I doubt if Archerfish is doing anything more sophisticated, if that
much.The maximum amount of high-speed memory that can be accessed by a
single core today without melting into a puddle of silicon is still
only measured in hundreds of GBs, about the equivalent to a frog. The
price tag on the Archerfish Thinking Camera ($399) implies MUCH LESS
RAM, probably putting it in the nematode range. Of course, with all
due respect, nematodes do have rudimentary sensors, and their species
is quite stable!
The error rate of the Archerfish with so little intelligence (capacity
for genetic or learned knowledge) is likely to be outrageous, so I
would suggest limiting it to security applications where very little
is at stake --- perhaps guarding the cookie jar or a windowless
bathroom in a house in a gated community in Monterey.
The good news is that we are now probably only 20 years away from the
real thing: Personal AI. The bad news is that when it becomes smart
enough to do the job well, it will also be smart enough to demand a
reasonable salary and fringe benefits. :-)
* Glenn Becker
Sep 16, 2010, 10:18:27 PM9/16/10
to The ALISA Community
NeuFlow - Yale supercomputer vision system
http://www.science20.com/news_articles/neuflow_supercomputer_can_see_well_enough_navigate_roads
http://www.science20.com/news_articles/neuflow_supercomputer_can_see_well_enough_navigate_roads
Reply all
Reply to author
Forward
0 new messages