A non trivial problem would be to find a minimal NOR network with a 33 bit
input and 1 bit output such that for binary input n, the output is the n'th
bit of enwik9.

On Fri, Oct 15, 2021, 1:38 PM James Bowery <[email protected]> wrote:

> Good catch with that reductio ad trivialum of my specification, Matt.
> Your proposed continually varying voltage to produce a serial output would
> still work, although not accomplishing quite what I wanted to with the
> parallel input being "compressed" into a serial representation and then
> decompressed.  I'll have to probe my subconscious for what my intuition was
> trying to get at but sometimes the Big Fish remain submerged.
>
> On Fri, Oct 15, 2021 at 10:06 AM Matt Mahoney <[email protected]>
> wrote:
>
>> Using pure logic (no time delays) you could construct a logic circuit
>> that outputs 1 when the input is enwik9 and 0 otherwise. It would consist
>> of a single NOR gate with 8 x 10^9 inputs and an inverter for each 1 bit of
>> enwik9. You could also construct a logic circuit that outputs enwik9 when
>> the input is 1. It would consist of inverters for each 0 bit of enwik9.
>>
>> This meets the requirements for the Hutter prize because the compressor
>> is not required to work on inputs other than enwik9. However, it does not
>> compress because the description length of the circuit is longer than 1 GB.
>>
>> On Thu, Oct 14, 2021, 9:35 PM James Bowery <[email protected]> wrote:
>>
>>> With parallel input of all bits in the "corpus" to the circuit, there is
>>> no directionality (other than the input to output mapping for each bit)
>>> except insofar as information dynamics dictates so as to create the serial,
>>> compressed (multplexed), bitstream that "represents" the corpus but does so
>>> in such a way as to not suboptimal burden on the decompressing
>>> (demultiplexing) circuit that must produce exactly the same parallel bits
>>> in the same order.
>>>
>>> So a better description than "prediction" of "letters" (really just
>>> bits) would be "imputation" of "letters".
>>>
>>> There _is_ something of practical importance here:
>>>
>>> Circa 2000 Federico Faggin funded The Boundary Institute to look at
>>> applying George Spencer Brown's Laws of Form notion of imaginary logic
>>> states ("This statement is false.") to circuit optimization, as GS Brown
>>> had done originally with his railroad car counter patent back circa 1960.
>>> Unfortunately, the work there (Tom Etter's solid work in that area
>>> previously supported by Paul Allen and then myself at HP) got side-tracked
>>> into paranormal research when it appeared that the quantum theory (not yet
>>> developed to the point it needed to be for DCG n-NOR circuits) was rigorous
>>> enough to provide, for the first time in paranormal research history,
>>> testable hypotheses.  Big Names showed up and the whole enterprise lost its
>>> focus on Faggin's priority.
>>>
>>> On Thu, Oct 14, 2021 at 2:56 PM <[email protected]> wrote:
>>>
>>>> So that goal is to make an actual circuit (or simulated circuit) made
>>>> of wires and resistors to be NOR functions so that you can send 1 input
>>>> signal in and pop out enwik9 the closer it reaches to a single of 1 but
>>>> using as few NORs and wires as can? So it is like some sort of brute force
>>>> approach? I'm not getting it still...
>>>>
>>>> Is the goal predicting the next letter? What is the goal, and how is it
>>>> evaluated. Why not if not though...
>>>>
>>> *Artificial General Intelligence List <https://agi.topicbox.com/latest>*
> / AGI / see discussions <https://agi.topicbox.com/groups/agi> +
> participants <https://agi.topicbox.com/groups/agi/members> +
> delivery options <https://agi.topicbox.com/groups/agi/subscription>
> Permalink
> <https://agi.topicbox.com/groups/agi/T728994814c1a40a0-M2b09bd16b545d2b926dc1bab>
>

------------------------------------------
Artificial General Intelligence List: AGI
Permalink: 
https://agi.topicbox.com/groups/agi/T728994814c1a40a0-Ma565c384c89f86cf2c662a71
Delivery options: https://agi.topicbox.com/groups/agi/subscription

Reply via email to