Re: What it would take...

On Fri, 09 May 2025 22:18:13 -0400, Richard Damon wrote:
 

On 5/9/25 9:11 PM, Richard Heathfield wrote:

The HHH code doesn't exactly invite confidence in its author, and his
theory is all over the place, but a thought experiment suggests itself.
>
If we were not all wasting our time bickering with a career bickerer...
if we were to really /really/ try, could we patch up his case and send
him on to his Turing Award? And if so, how?
>
ISTR that there is suspected to be a theoretical window for him, so I
suppose what I'm asking is what sort of boathook we would need to poke
that window a little wider.
>
Can he even get there from here? Evidence would suggest that simulation
is a dead end unless he can find a way to get the simulated program to
include its own simulation in its behaviour, which he has not yet
managed to do - but /is/ there a way?
>
Or could he abandon simulation completely and instead write a TM parser
that builds an AST and walks it looking for evidence of terminating or
looping? If he could, would that turn the trick?
>
Or do we have a latter day Cantor waiting in the wings to close the
window once and for all?
>
Is there, in short, any way of putting out this un-halting flame war
and turning this group to better use?
>
>

If he was willing to include the code for HHH in the input representing
DDD, then HHH would be able to atempt to correctly emulate this input.
>
There have been methods put forward, that given an acceptace of the
detectability of DDD calling HHH, which can only be done it seems if we
make the system non-turing complete by saying that the input program and
the decider are put into the same memory space, and we are not allowed
to "copy" an algorithm to make a new copy, but only call the origianal
version so HHH can detect the recursion by reference to that address
that some versions of programs that do this "recursive simulation" can
be correctly decider (but not all, like the pathological version).
>
In this method, the Decider detecting the recursion, tries emulating the
code in two parrallel branches based on both possible answers, and if
one branch matches the behavior of the answer, it can return that
answ3er.

 Branching is my idea.
 /Flibble