Re: [agi] Re: Another Kind of KC Prize

On Monday, October 11, 2021, at 2:02 PM, James Bowery wrote: > n-NOR is 2, 3, 4, 8 or any number of input NOR, including 1 (which is just a > NOT gate).  It is a physical impossibility, but not a logical impossibility. This might be a losylossslessyzygy (sic). Any irreversible k-dimensional CA

Re: [agi] Re: Another Kind of KC Prize

On Friday, December 10, 2021, at 9:31 PM, James Bowery wrote: > At one point Chatain discusses the early work on minimization of the > complexity of finite state automata that outputs an observational history as > an approach to natural science. He also wants to see a mathematical theory of

Re: [agi] Re: Another Kind of KC Prize

Wolfram interviews Chatain (part 1) regarding the relationship between Godel's Incompleteness and Algorithmic Information , among other things. At one point Chatain discusses the early work on minimization of the complexity of finite state automata that outputs an

Re: [agi] Re: Another Kind of KC Prize

I like the HP most but I think Perplexity is just faster to test and can cut out the letter prediction part of it. I am trying it now on my AI. The reason is that training only on a large chunk of data is faster instead of trying to predict it "all". --

Re: [agi] Re: Another Kind of KC Prize

A mathematician responded privately to me with the same basic complexity measure suggested by Matt (and with some comments on the computational complexity measure of finding a minimal DCGoNORs in the context of DAGoNORs aka "Circuit Complexity"), in response to my question at

Re: [agi] Re: Another Kind of KC Prize

On Fri, Oct 22, 2021 at 1:02 PM Matt Mahoney wrote: > I think a n-NOR contest could be made practical, at least for a smaller > data set like the Calgary corpus (3 MB), if not enwik9, using a state > machine made of NOR gates. > Even if I'm overestimating the difficulty of an enwik9 contest, it

Re: [agi] Re: Another Kind of KC Prize

I think a n-NOR contest could be made practical, at least for a smaller data set like the Calgary corpus (3 MB), if not enwik9, using a state machine made of NOR gates. A state machine (n1, n2, n3, n4) is specified by the number of input bits n1, state bits n2, gates n3, and output bits n4,

Re: [agi] Re: Another Kind of KC Prize

He is talking about instruction set, which is part of decompressing program. I agree that it's a red herring. And I agree that KC is a trivial principle. -- Artificial General Intelligence List: AGI Permalink:

Re: [agi] Re: Another Kind of KC Prize

I don't think he is talking about that Boris lol. And I already know that. @James I already know what KC is, it is the shortest length to compress to for some given input, i.e. the best score. It's not much meaningful though reading all that above, all I need to do is build AI like GPT. Too

Re: [agi] Re: Another Kind of KC Prize

It's really the number of bit in (compressed representation + decompressing program). If the amount of data is large enough, the second component becomes insignificant. -- Artificial General Intelligence List: AGI Permalink:

Re: [agi] Re: Another Kind of KC Prize

Is it a patent? I'm out. :D Makes 0 sense. :/ -- Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T728994814c1a40a0-M20ada1986806d8ec7e798d0e Delivery options: https://agi.topicbox.com/groups/agi/subscription

Re: [agi] Re: Another Kind of KC Prize

The problem, ID, is two-fold: 1) People are deeply confused by the likes of Popper about model selection, and have been so-confused ever since his noise about "falsification" overwhelmed Solomonoff's rigorous signal (indeed proof) about "information" and its implications for scientific induction

Re: [agi] Re: Another Kind of KC Prize

Then why would I do this contest instead of the Hutter Prize / LC contest if I *still need to make an ANN* ? How does this contest change my ANN, is not my ANN still exactly the SAME algo This contest is making no sense Can one use GPT on this contest and if not how do you expect them

Re: [agi] Re: Another Kind of KC Prize

You can build a computer out of NOR gates and program it to simulate a neural network. On Fri, Oct 15, 2021, 8:21 PM wrote: > But. > > Are we still programming neural networks if we are building these NOR > circuits? So then...why NOR circuits if we are using ANNs? I mean we

Re: [agi] Re: Another Kind of KC Prize

But. Are we still programming neural networks if we are building these NOR circuits? So then...why NOR circuits if we are using ANNs? I mean we are predicitng the next bit right ? - so, therefore, you MUST use neural networks to do soHow the hell hay can you build

Re: [agi] Re: Another Kind of KC Prize

Yes, n-NOR is simpler and I prefer it for that reason. I was thinking of practical implementations that already take 50 hours using 32 or 64 bit operations. We would need to write a translator for n-NOR to C, preferably one that can optimize by grouping bits into integer and vector operations. On

Re: [agi] Re: Another Kind of KC Prize

I have a couple of reasons to avoid changing the metric to include anything but n-NOR DCG complexity: 1) A lot of the resistance to adopting lossless compression as model selection criterion is based on Kolmogorov Complexity's open parameter: Choice of UTM. Indeed, I see the psychological appeal

Re: [agi] Re: Another Kind of KC Prize

On Fri, Oct 15, 2021, 4:50 PM wrote: > So the goal is to make a circuit (?_that is ONLY made of NOR components_?) > that outputs a 1 only when the input is enwik9? Hence the size of said > circuit is the score? How do you know if some other input ex. enwik8 > wouldn't pop out a 1? You'd have to

Re: [agi] Re: Another Kind of KC Prize

So the goal is to make a circuit (?_that is ONLY made of NOR components_?) that outputs a 1 only when the input is enwik9? Hence the size of said circuit is the score? How do you know if some other input ex. enwik8 wouldn't pop out a 1? You'd have to feed in all possible inputs to see if does,

Re: [agi] Re: Another Kind of KC Prize

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 wrote: > Good catch with that reductio ad trivialum of my specification, Matt. > Your

Re: [agi] Re: Another Kind of KC Prize

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

Re: [agi] Re: Another Kind of KC Prize

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

Re: [agi] Re: Another Kind of KC Prize

Have you told others about this new prize / evaluation for finding AGI? Where and where is that page? -- Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T728994814c1a40a0-M6517a003afd3937a2564f8d9 Delivery options:

Re: [agi] Re: Another Kind of KC Prize

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

Re: [agi] Re: Another Kind of KC Prize

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

Re: [agi] Re: Another Kind of KC Prize

On Tue, Oct 12, 2021, 8:14 PM wrote: > And then, if successful, we solved AGI? Wow. So it all came down to a > simple NOR network with 1 wire and some serial delay. Where is the prize > page? > AGI would be around 10^15 gate inputs in a practical implementation with 1 ns delays. With

Re: [agi] Re: Another Kind of KC Prize

Can you link me that? I can't find the previous time you mentioned it. -- Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T728994814c1a40a0-Md5ab8ff52cd20fa3cec93346 Delivery options:

Re: [agi] Re: Another Kind of KC Prize

Actually, maybe I should call this "Bowery-Kolmogorov Complexity" since there is a genuine degree of originality AFAIK, and people do need to think of it in different terms than KC. As for the 5 year old who doesn't understand KC, I can only do so much. And, by the way, the concept of KC doesn't

Re: [agi] Re: Another Kind of KC Prize

For the benefit of the dim (including those feigning dimness) who might not get "ID"'s joke: As I said in the original post: "I believe I've mentioned this *idea* before and repeat it here with added precision." On Tue, Oct 12, 2021 at 7:14 PM wrote: > And then, if successful, we solved AGI?

Re: [agi] Re: Another Kind of KC Prize

And then, if successful, we solved AGI? Wow. So it all came down to a simple NOR network with 1 wire and some serial delay. Where is the prize page? -- Artificial General Intelligence List: AGI Permalink:

Re: [agi] Re: Another Kind of KC Prize

A NOR gate outputs a 1 if all the inputs are 0 and a 0 if any inputs are 1. You can implement any logic circuit with just NOR gates. If you add a time delay then you can build flip flops (memory) and oscillators using recurrent networks. One way to specify a logic circuit would be to number the

Re: [agi] Re: Another Kind of KC Prize

@csmicahelliott I understand it completely and still don't get this guy. -- Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T728994814c1a40a0-M227b5a7f9f164ef01652a103 Delivery options:

Re: [agi] Re: Another Kind of KC Prize

"Don't get upset when people get all hung up on accepting lossless compression as model selection criterion because there is an unspecified choice of Universal Turing Machine instruction set for the specification of Kolmogorov Complexity." That says a whole lot of nothing and I understand all

Re: [agi] Re: Another Kind of KC Prize

How does this help find AGI?? How does this help Prediction? What ignited this idea in your brain? -- Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T728994814c1a40a0-M598fb7ccea8d2778eff57b5c Delivery options:

Re: [agi] Re: Another Kind of KC Prize

As for it being "hard", compare teaching "n-NOR" to a 5 year old to teaching him what a universal turing machine equivalent is. On Mon, Oct 11, 2021 at 1:02 PM James Bowery wrote: > In the popular CMOS and TTL logic families, NOR gates with up to 8 inputs > are available: > > CMOS > 4001: Quad

Re: [agi] Re: Another Kind of KC Prize

In the popular CMOS and TTL logic families, NOR gates with up to 8 inputs are available: CMOS 4001: Quad 2-input NOR gate 4025: Triple 3-input NOR gate 4002: Dual 4-input NOR gate 4078: Single 8-input NOR gate TTL 7402: Quad 2-input NOR gate 7427: Triple 3-input NOR gate 7425: Dual 4-input NOR

[agi] Re: Another Kind of KC Prize

You know what I mean. You made this so hard Google doesn't even turn up anything on a "n-NOR network". It's worse than even PPM. Show me the prize page. Is it official? -- Artificial General Intelligence List: AGI Permalink:

Re: [agi] Re: Another Kind of KC Prize

On Sun, Oct 10, 2021 at 6:07 PM wrote: > ...can you say it like I'm 5 ears old?... > Can a 5 year old understand how a water-pressure activated valve works that controls the flow of water to another valve? Can a 5 year old understand how two or more tubes of water coming together into one tube

[agi] Re: Another Kind of KC Prize

I don't get it, how can you compress anything over 1 bit to 1 bit, it's impossible. And, what is the goal of this circuit I mean hey, if it becomes 1 bit then the data must be floatin` around` somewhere anyway, so why not just store the whole answer...

[agi] Re: Another Kind of KC Prize

Oh, I really like Compression Evaluation for finding AGI, and generations of data completions for finding AGI. But I don't understand what this evaluation does/ how it works, can you say it like I'm 5 ears old? For example Compression Evaluation is done by training a model to predict the next