🚀 go-pugleaf

In article <1992Sep14.011825.25966@galois.mit.edu> tycchow@riesz.mit.edu (Timothy Y. Chow) writes:
>In article <1992Sep13.215142.25567@mp.cs.niu.edu> rickert@mp.cs.niu.edu (Neil Rickert) writes:
>>Some of us have the intuition that the lookup table could not pass the
>>Turing test.  What does this do to your argument?
>
>Can you be more explicit about this intuition?  Perhaps I should point out
>that my sample of the lookup table was a bit misleading.
>...
>case (1st input)
>  "Hi":	output("Hi")
>	case (2nd input)
>	  "What's up?": output("Not much.")
>			case (3rd input)
>			...
>	  "How's it going?": ...
>	  ...
>  "Are you conscious?": output("What do you think?")
>  ...
>
>The main problem lies with the limitations of the input and output:

No.  The main problem lies in the impossibility of constructing a table
sufficiently comprehensive.

To maintain some degree of dialog, the lookup table daemon must respond
within a reasonable time.  Let's say a maximum time delay of 1 year, to
be excessively generous.  Now start counting how many entries are needed
in the table -- at least 26^n where n is the maximum length (in characters)
of the conversation -- and assume that you would probably need at least
one atom to record the information for each table entry.  All these
atoms would have to fit within a diameter of 1 light year to meet the
responsiveness requirement.

The density of atoms will be so high that I believe you will have created
a black hole, from which the response of the lookup table daemon cannot
escape.