jtm@cs.cornell.edu (Jan Thomas Miksovsky) (03/01/90)
I keep hearing things like, "Won't it be great when we have a direct
neuron-level interface to a digital system so that we won't be slowed
down by things like typing." But how wide would the bandwidth between
our minds and a virtual reality system really be if there were no
"bottlenecks" like hands and keyboards?
Think for a minute about how fast you can type or manipulate a mouse
when you're processing at top speed. When I'm really flying, I can keep
maybe eight or ten operations in my Mac's buffer: I'm pounding away on
command keys, and dialog boxes that begin to pop up on the screen are
blown away before they can ever be completely drawn. I can keep this up
for maybe a minute or so; then I have to step back and figure out what
I'm going to do next.
If we could directly manipulate the computer with neuron pulses, yeah,
sure, we could cruise along at light-speed. But if we're performing any
kind of analytical task, wouldn't we still be limited by the speed of
our consciousness "processor"? We might think something like, "Load up
the simulation of the house I'm building," and even if WHAM the
simulation appears right away, it's going to take us a second or two to
figure how we next want to manipulate the environment.
The real speedup will probably occur in tasks that are repeated so
often, and learned so well, that conscious thought isn't required.
Think about when you play some sport that you're good at: balls and
sticks or whatever are flying around, and you can just move around,
reacting to the environment without having to spend a lot of high-level
processing time considering your next move.
* * * * *
A related issue: What research has been done in this area? There's
been some talk on the net recently about research involving direct
stimulating the visual processing areas of the brain to provide blind
people with sight. Does the opposite work -- with prosthetic limbs, for
instance... Can brains learn to trigger neurons in order to activate a
foreign object?
--
Jan Miksovsky ...uunet!applelink.apple.com!D4710cphoenix@csli.Stanford.EDU (Chris Phoenix) (03/01/90)
In article <2193@milton.acs.washington.edu> jtm@cs.cornell.edu (Jan Thomas Miksovsky) writes: >If we could directly manipulate the computer with neuron pulses, yeah, >sure, we could cruise along at light-speed. But if we're performing any >kind of analytical task, wouldn't we still be limited by the speed of >our consciousness "processor"? If we could directly manipulate the computer with neuron pulses, and it could manipulate our neurons, surely someone would develop a brain-accelerator. It could be as simple as a math coprocessor, or maybe a memory. I don't know how much we could be sped up while still staying human (followups to alt.philosophy.humanist.computer.flame :-) ) but it's too early to say we won't be able to keep up with computers. >The real speedup will probably occur in tasks that are repeated so >often, and learned so well, that conscious thought isn't required. For tasks like that, have an AI do it! No, this isn't facetious, though it may be naive. It shouldn't be too hard to build a program that could learn repetitive tasks, and do them at the proper times and in the proper ways... or something... To tie this in with VR: I recently called up VPL to ask about summer internship opportunities. I told them I was interested in AI, and the person I talked to (sorry, I didn't get the name) told me that they were trying to get away from AI, and have it more user-directed. This surprised me. I would have thought that AI would be a pretty important part of VR, if only to choose which things to display and how to display them. Can anyone comment on whether AI is being used at all in VR, and if so, how? (and if there are any AI-VR jobs available in the Bay Area? :-) ) -- Chris Phoenix | "I've spent the last nine years structuring my cphoenix@csli.Stanford.EDU | life so that this couldn't happen." ...And I only kiss your shadow, I cannot see your hand, you're a stranger now unto me, lost in the dangling conversation, and the superficial sighs...
peb@tma1.Eng.Sun.COM (Paul Baclaski) (03/02/90)
In article <2193@milton.acs.washington.edu> jtm@cs.cornell.edu (Jan Thomas Miksovsky) writes: >If we could directly manipulate the computer with neuron pulses, yeah, >sure, we could cruise along at light-speed. The human brain does a lot of distributed processing in the eyes and ears and the "neural protocol" between the brain and these input devices is unique from person to person. This is probably also the case with motor nerves, olfactory bulb and tactile nerves (I've heard that one of the major problems with heart transplants is that the nerve signals from the brain to the foreign heart are not quite right). Because of this, any system that interfaces directly to nerves is going to need lots of training and adjustment. To use standard hardware, one would need a personal interface machine to transform to and from your personal nerves. The training time on such a interface could be considerable--I wonder if Hans Moravec has any ideas in this area (Mind Children does not mention problems of this sort). A direct nerve interface is probably a very invasive proceedure like big time brain surgery, so I would expect it to be pretty far out there on the time scale (e.g., you either need safe brain surgery (nanotech to the rescue) or volunteer terminal patients). In article <2205@milton.acs.washington.edu>, cphoenix@csli.Stanford.EDU (Chris Phoenix) writes: > If we could directly manipulate the computer with neuron pulses, and it could > manipulate our neurons, surely someone would develop a brain-accelerator. How about this: we can talk in our heads without moving our mouths. This means that it would be possible to tap into the inside voice and use it as an output device. Then, the coprocessor module (off the shelf expert systems personified) could listen to the voice and talk to you or display information "inside" your head. This assumes a neural interface, but it keeps the brain intact. Accelerating thought by improving the brain is tantamount to downloading your consciousness into a machine. Are you the same person after such a transformation? How could we tell? OVERALL, I think direct neural interfaces are in the realm of fantasy and are not worth discussing too much (unless you want to write Science Fiction). The above brain augmentation could also be done with other technology with the exception that "inside voice" could not be used. > To tie this in with VR: I recently called up VPL to ask about summer > internship opportunities. I told them I was interested in AI, and the person > I talked to (sorry, I didn't get the name) told me that they were trying to > get away from AI, and have it more user-directed. AI is orthogonal to VR. I would guess that VPL's strategy is to empower people--it sounds like you are suggesting making things automatic, which is great for making people lazy (e.g., "wouldn't it be nice to have some AI slaves to order around?..."). This is enticing in a negative way. In my opinion, the purpose of AI is to understand Intelligence and the Universe. Put another way: Jaron Lanier stated in the Whole Earth Review interview (Fall 1989) that technology that simply increase the amount of power people have is inherently bad (because we have all sorts of social problems that just get amplified). He then stated that a technology that increases human communication is good--how could anyone say that the telephone was a bad idea? He also mentioned that people who watch TV appear to be uninvolved, but people talking on the telephone are very much involved. Jaron wants to encourage involvement, not uninvolvement. Paul E. Baclaski Sun Microsystems peb@sun.com
aipdc@castle.edinburgh.ac.uk (Paul D. Crowley) (03/02/90)
In article <2193@milton.acs.washington.edu> jtm@cs.cornell.edu (Jan Thomas Miksovsky) writes: >Think for a minute about how fast you can type or manipulate a mouse >when you're processing at top speed. When I'm really flying, I can keep >maybe eight or ten operations in my Mac's buffer: I'm pounding away on >command keys, and dialog boxes that begin to pop up on the screen are >blown away before they can ever be completely drawn. I can keep this up ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ >for maybe a minute or so; then I have to step back and figure out what >I'm going to do next. I too find that desktops work too slowly, and that this affects how useful they are. Has anyone out there used a desktop so fast that they would't notice if it got ten times faster most of the time (i.e. it is already too quick for the time of most tasks to be noticed). What difference does it make? I ask because I once had the opportunity to word-process on a 386 machine. The wp loaded within half a second and remembered where it was before. It was also called "m" so it didn't take much effort to call up. -- \/ o\ "I say we grease this rat-fuck son-of-a-bitch Paul D Crowley /\__/ right now. No offense." - Aliens. aipdc@uk.ac.ed.castle
almquist@cis.udel.edu (03/02/90)
In article: jtm@cs.cornell.edu (Jan Thomas Miksovsky) writes: >If we could directly manipulate the computer with neuron pulses, yeah, >sure, we could cruise along at light-speed. But if we're performing any >kind of analytical task, wouldn't we still be limited by the speed of >our consciousness "processor"? We might think something like, "Load up >the simulation of the house I'm building," and even if WHAM the >simulation appears right away, it's going to take us a second or two to >figure how we next want to manipulate the environment. YES, valid points. BUT, perhaps by then we shall have an AI/Expert System friend/co-pilot. If not a co-pilot then perhaps advanced MACROs. Instead of having to say, "Load up the simulation of the house I'm building," we could instead say "Simulate, House, Earthquake". YES, poor example BUT, if you stop to think about it, the next generation might be drastically different. What would happen to the thinking processes of this generation? Would they be modified, simplified, extended, etc? Perhaps the next generation will not be limited by our learned limitations (ie. our mental blocks). They will not remember the ages of keyboards. They will begin to develop drastically different memory structures, fetch/recall routines, etc. If we are to directly tap into their brains at an early age (like the future Gibson wove for us) their mental development, I feel, would become drastically different from one who doesn't plug-in. Perhaps this increased stimulation of the brain MIGHT increase one's total brain usage. As the saying goes, "we currently use less than 25% of our brain's total power." - Michael Almquist
wex@sitting.pws.bull.com (Alan Wexelblat) (03/04/90)
Jan Mikovsky was asking about the ability to do routine things or well-practiced things without having to expend higher-order brain power. Look up research on "muscle memory." That's the term I've seen used to describe the way non-obvious physical actions are mapped into semantic actions (such as touch-typing). One of the cyberspace navigation techniques invented by Kim Fairchild and myself depends on muscle memory. It works moderately well. -- --Alan Wexelblat internet: wex@pws.bull.com Bull HN Information Systems Usenet: spdcc.com!know!wex phone: (508) 671-7485 Adapt, adopt, improvise!
bro@eunomia.rice.edu (Douglas Monk) (03/04/90)
In article <2193@milton.acs.washington.edu> jtm@cs.cornell.edu (Jan Thomas Miksovsky) writes: >A related issue: What research has been done in this area? [...] >Can brains learn to trigger neurons in order to activate a foreign object? Short answer: yes. Back in the '70s, I recall reading some experimental reports on implanting fine wires as sensors into muscle fibers in the wrists of subjects. After biofeedback training, they could learn to activate individual muscle CELLS on command. The drawbacks that I recall are a long training period, occasional poor performance rates (such as when fatigued, etc.), and the fact that the wires would eventually kill the fibers in which they were implanted, thus requiring reimplantation and a new training period. I believe they had mapped out strategies around the first two problems, and the last was the hard part. I recall thinking at the time that typewriters would soon come with a cable to plug into sockets wired into our wrists, and we would just have to visualize to type. Looking at my wrists, I see it hasn't happened yet, despite being technically feasible for 20 years. :-) Doug Monk (bro@rice.edu) Disclaimer: These views are mine, not necessarily my organization's.
wcs) (03/05/90)
In article <WEX.90Mar2090441@sitting.pws.bull.com> wex@sitting.pws.bull.com (Alan Wexelblat) writes:
]Jan Mikovsky was asking about the ability to do routine things or
]well-practiced things without having to expend higher-order brain power.
Some of the people I've talked to at LLNL are speculating
about what they'd like, and they'd really like to have two-eye
1Kx1Kx24bit-color at 30 frames/second, because that's about
the most data input they think they can use - about 1Gbit/sec.
It needs to be attached to the back-side of a Cray,
because what they'd really like is for a couple of physicists
to be able to walk around in a simulated nuclear reaction,
having conversations about "what if it were a couple hundred
degrees hotter?" or "let's make that wall a little thinner"
"No, not that thin", or "magnify that spot there 100X".
] One of the cyberspace navigation techniques invented by Kim Fairchild
] and myself depends on muscle memory. It works moderately well.
How do you attach to it - position-detection gloves, or
head/eye motion detectors, or other stuff?
[ Disclaimer: I don't do this stuff professionally (the LLNL stuff
was doing datacomm.) The .signature was appropriate for a random discussion
and I haven't found anything sufficiently weird to replace it.]
--
# Bill Stewart AT&T Bell Labs 4M312 Holmdel NJ 201-949-0705 erebus.att.com!wcs
# Fax 949-4876. Sometimes found at Somerset 201-271-4712
# He put on the goggles, waved his data glove, and walked off into cyberspace.
# Wasn't seen again for days.jeffj@cbnewsm.ATT.COM (jeffrey.n.jones) (03/06/90)
>A related issue: What research has been done in this area? There's >been some talk on the net recently about research involving direct >stimulating the visual processing areas of the brain to provide blind >people with sight. Does the opposite work -- with prosthetic limbs, for >instance... Can brains learn to trigger neurons in order to activate a >foreign object? > I have read about and seen prosthetic limbs that were crudely controlled and activated by direct connects to nerves. At a virtual reality seminar that I just attended there was a mention of a chip that was being developed that would interface into the nerves and act like a straight switch. [Moderators Note:] Dr. Joseph Rosen at Stanford University Medical Center has been conducting research into this area. The goal of one of the his projects is to develope a "chip prosthesis" that will enable specific axon-to-axon reconnection. -- Jeff Jones | Prediction is very difficult, especially UUCP uunet!seeker!jeffj | about the future. Infolinc BBS 415-778-5929 | Niels Bohr