RE: Sexbots??

Billy Brown (bbrown@conemsco.com)
Thu, 11 Mar 1999 14:16:30 -0600

Terry Donaghe wrote:
> But won't a realistic humanoid sexbot need to have nearly human AI in
> order to be convincing? At least animal AI, right? Otherwise it's
> just a pile of warm flesh that you thrust into (or get thrust by). I
> see the software of a sexbot being at least as tricky as that of my
> virtual sex kittens and, coupled with the hardware a hell of a lot
> more expensive. Without extensive AI, it'll just lay there and maybe
> moan. If I'm gonna dish out big bucks for a sexbot, I want a WILD
> woman!!

Well, yes, it will. However, so will the VR version. I think this aspect is about the same difficulty level either way.

> If your sexbot is a recreation of a 110 pound human female, what
> happens if her software goes nuts and she starts biting (at the wrong
> time), kicking, and generally flailing about during the heat of the
> moment? Possibly major injuries, costly repairs, and maybe
> embarassing stains on the furniture... And that's not to mention the
> problems the 225 lb muscle-ripping he-stud that some customers will
> want might cause....

What if your force-feedback system goes haywire? To get the illusion of walls and other solid objects, it needs to be considerably stronger than you are (strong enough to easily break bones, in fact). The robot seems a bit safer to me - it doesn't need to be any stronger than a normal human.

> If a virtual sex partner goes nuts you should just be able to
> disconnect with no harm done (blue screen of coitus interuptus - hee
> hee)..

See the above comment about force-feedback systems. This is actually a significant factor in current research - corporations are leery of funding these things, because they're afraid people will get hurt when the inevitable bugs surface.

> Sexbots would entail considerable maintenance (especially mostly
> organic ones - imagine changing it's "diapers" - ewwwww). Virtual sex
> partners come with no muss, no fuss.

If you can make it walk around the house on its own, you're 99% of the way to being able to do any kind of physical activity. Programming it to handle its own routine maintenance should be trivial.

Now, its true that a VR personality doesn't take much maintenance. However, your full-immersion VR rig is going to be at least as complex as a robot, with lots and lots of moving parts. It probably still won't be as trouble-prone as a mechanical robot (which is one reason to favor the organic approach), but it isn't exactly going to be maintenance-free either.

Besides, a physical robot can do lots of other things, too. Your VR partner can't clean house, make dinner, walk the dog, or run errands for you.

> Besides, with virtual partners one can have sex with pretty much
> ANYONE or hell, everyone! Virtual sex can include as many partners as
> can be imagined. Sexbots provide each customer with one or two
> partners - depending on price, maintenance, etc....

Here, the VR approach wins hands-down.

> Also, virtual sex partners could use the AI of living, breathing
> humans remotely located. I'm sure some of the first applications will
> be voyeuristic views of 3D strippers and naughty lovers..

So could a physical robot.

> I don't think the idea of sexbots is completely without merit, but I
> can't see the technology maturing before we could have really
> inexpensive virtual sex partners..

The hard part in a VR isn't the AI for the simulated partner - animal-level AI should exist well before we have the horsepower to run a full-immersion VR. What makes the VR difficult is the need to simulate, in detail and on demand, virtually any kind of physical process that can be perceived by human senses. Making a realistic VR requires software far more complex than anything we could possibly build with current tools, and computers much faster than even a human-equivalent robot would need.

I think the outcome of this race will depend on what the relative difficulty of the hardware vs. the software turns out to be. My own projection would be as follows:

animal-level AI		10 - 20 years
humanoid robots		20 - 30 years
realistic VR		25 - 40 years

Billy Brown, MCSE+I
bbrown@conemsco.com