Re: Cool visual illusion


Wolf Kirchmeir wrote:
> feedbackdroids wrote:
> [...]
> >
> >
> > And how far down the list of 8 items [subject of these posts] will this
> > important information get us? Any sensical non-garbage answer will
> > clearly support what I 've said several times.
> >
>
> Hey, 'droid, haven't you grasped yet that it's not a step by step
> procedure? Stop thinking about positions on the list, and think instead
> of aspects of the process.
>


The list is a summary of the process. That's all it is.


> What EAB gives us applies to all "steps". Eg, "what the client wants"
> must be considerd in terms of what's possible - so you go and look at
> what systems can do, ie, their behaviour, to see whether what the client
> wants is available off the shelf, or must be cobbled together, or
> requires an (apparently) brand new behaviour, etc.
>
> No matter what, you can't leave behaviour out of it. Take an exmaple in
> which most people do not usually consider "behaviour" as a relevant
> concept, the functioning of a bridge. I mean, it just sits there, eh? it
> doesn't actually _do_ anything.
>


You're talking generalities here, which is the top few items on the
now-infamous list. We want to understand the mechanisms of
intelligence, and implement an AI - item #5 [IIRC].

Pray tell us, how does behaviorism suggest we proceed in actually
building such a mechanism?




> Oddly enough, engineers talk about the "behaviour" of a bridge under
> stress. And what do the mean by behaviour? The dynamic relation between
> the bridge and its environment, of which the load is a (usually small)
> part. Study the Tay bridge disaster: that bridge fell down because its
> designers didn't consider the effects of wind on the bridge. They
> thought it was enough to know how to design the structure of the bridge,
> which according to their calculations would more than support itself,
> ie, was strong enough to support the design load.
>
> The analogy may seem over simple, but I think it's instructive. It shows
> first, that how a system behaves depends not only on its structure but
> on its environment. Secondly, it sharpens the concept of a system
> capable of learning, which I presume is one of the essential features of
> any AI system we'd find it worth building. Bridges don't learn, ie, they
> don't change their behaviour in response to the environment. (If only
> the Tay bridge could have learned how to resist the wind pressures that
> brought it down. Maybe it could have done exercises to strengthen its
> joints....)
>
> Learning systems change their behaviour in response to the environment.
> That's why you can't leave behaviour out of it, at any level (although
> it's more directly relevant at some levels than others.) IOW, ifnt, and you
> want to design an AI device capable of learning, you must begin by
> specifying just what kind of learning behaviour you wayou must
> constantly refer back to that as you build the internal structure.
>


Yes, you said it right here, exactly what I have been saying .... "...
you want to design an AI device capable of learning, you must begin by
specifying just what kind of learning behaviour you wayou must
constantly refer back to that as you build the internal structure ...".


Specification is at the top of the list. Refer back is part of the test
and revise items, listed near the bottom of the list.

The only real question - and PLEASE tell us how to do it - is the
actually process of "doing it" ==>> designing and building the system.

PLEASE tell us how behaviorism would have us do it.

.

Relevant Pages

  • Re: Cool visual illusion
    ... Take an exmaple in which most people do not usually consider "behaviour" as a relevant concept, the functioning of a bridge. ... The dynamic relation between the bridge and its environment, of which the load is a part. ... They thought it was enough to know how to design the structure of the bridge, which according to their calculations would more than support itself, ie, was strong enough to support the design load. ... Secondly, it sharpens the concept of a system capable of learning, which I presume is one of the essential features of any AI system we'd find it worth building. ...
    (comp.ai.philosophy)
  • [REGRESSION] 2.6.23-rc1: uhci_hcd. irq 4: nobody cared
    ... 00:01.0 PCI bridge: VIA Technologies, ... 00:11.0 ISA bridge: VIA Technologies, ... # Firmware Drivers ... # PCCARD support ...
    (Linux-Kernel)
  • Re: configuring if_bridge with stp at boot in /etc/rc.conf
    ... To re-iterate, when I try to activate the bridge with STP parameters in rc.conf, one of the bridge interfaces goes into forwarding mode, but the other interfaces remains in disabled mode. ... port 3 priority 10 path cost 10 forwarding ... # Power management support ... # PCI Ethernet NICs that use the common MII bus controller code. ...
    (freebsd-net)
  • [BUG] linux-2.6.11-rc3 probably in ACPI battery procfs ...
    ... 0000:00:01.0 PCI bridge: Intel Corp. 82855PM Processor to AGP Controller ... Wireless LAN Controller ... # ACPI Support ...
    (Linux-Kernel)
  • Kirchmeirs learning bridge
    ... the functioning of a bridge. ... > but on its environment. ... This could only be done by an evolutionary learning ... strengthening and weakening connections. ...
    (comp.ai.philosophy)