Re: [bafuture] Purely Digital Strong AI??
- I think Cyc is a dead end as far as its original
purpose is concerned. However, it does have some
interesting stuff coming out
There have been some development in a field called
neural-symbolic integration... I am betting on that
for a major break through to happen.
--- Steve Dekorte <steve@...> wrote:
> On Sep 21, 2006, at 12:54 PM, Shawn Pan wrote:
> > Ha..witty. Well, I think you know what I was
> > about.
> Ya, just having some fun. :-) You're right - there
> are still
> academics interested in it. I just haven't heard or
> seen anything
> interesting come out of the field in a very long
> time. The Cyc
> project seemed to be the last bastion - did anything
> useful come out
> of it?
> - Steve
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
- On Sep 18, 2006, at 11:16 PM, Wayne Radinsky wrote:
> J. Andrew Rogers wrote:More accurately, it is *much* easier to implement error correction
>> Analog and digital are interchangeable for all intents and purposes.
>> There is nothing you cannot do in analog that you cannot do in
>> digital and vice versa.
> I thought there was one critical difference: digital information does
> not suffer generation loss.
when information is encoded in digital formats rather than analog
formats. It should probably be pointed out that, contrary to popular
misconception, analog formats are discrete just like digital formats.
The tradeoff is this:
Information encoded in a digital format can be extremely resistant to
errors for a number of reasons and at relatively little cost.
However, digital formats tend to not degrade gracefully when an error
manages to slip through that gauntlet.
Information encoded in an analog format is very sensitive to
transient errors (hence cumulative generation loss), and error
correction/prevention is very expensive relative to digital.
However, analog formats tend to degrade gracefully such that some
information is recoverable from degenerate data.
The main disadvantage of analog is that the price/performance of
digital signal integrity is roughly a function of Moore's Law over
time, whereas analog format integrity has had something like a quasi-
logarithmic price/performance over time and it hit the flat part of
that curve years ago. The price/performance curves for the two
formats crossed each other some time in the 1980s, after which
digital left analog in the dust. A high-end purely analog format can
only manage about 12-bits of signal at any point in time in the real
world, whereas you can get about 20-bits out of modestly priced
digital formats (real bits, not marketing bits).
J. Andrew Rogers