2018 Year In Review

I've done some benchmarking in 2018. I benchmarked an "AI software" we devised, by some benchmarks mostly I invented, too. Which doesn't look very good, I know, but bear with me!

For one, I have given an unsolved Sudoku puzzle to this software with two working names, "Spector" and/or "Profounder". It concluded, that for every X and every Y: X==Y implies that column(X) != column(Y) and row(X)!=row(Y). (Zero Sudoku topic knowledge by Spector is, of course, a necessary condition.)

With several unsolved Sudoku puzzles, Spector concluded that subsquare(X) != subsquare(Y). Just for one puzzle, the concept of "3 by 3 subsquare" isn't economical. It's economical for several of them, though.

The second benchmark I invented, was giving the string "ABCDEFGHIJKLMNOPQRSTUWXYZ" to Spector. The string generating algorithm would be simpler if the letter "V" wasn't missing. This is the way Spector notices something might be wrong with the given string. (Zero alphabet topic knowledge by Spector is, of course, a necessary condition.)

Yet another benchmark was numbers from 3 to 122. Each labeled by 0 or 1, depends if it's nonprime or prime. The simplest generating algorithm is a sort of Eratosthenes sieve. Not for numbers, but for their labels. Spector finds and generates it, with zero knowledge about primes.

Another benchmark was inspired by a mistake someone made. There is a nursing school here somewhere, which sends their students to practice in a nearby hospital for a day or two every week. Except for freshmen in the first year. They teach them everything else in this school, of course, including the gym (boys and girls separated there) and they feed them all once a day, too. It's standard in this part of the world. But the school does not feed them when they are at the hospital.

So they forget to feed girls from 2B department on Thursdays when they are in school. They forget to include that into their schedule. Boys from 2B have eaten while girls were exercising, but poor girls were forgotten and nobody noticed.

I asked Spector, giving him the school schedule in CSV format if anything is wrong with it. Spector did conclude, that every student has a lunch break once a day when not practicing, except for those girls on Thursday. Which was (probability-wise) odd enough to be significant.

Spector/Profounder is all about one mayor and three to five lesser tricks. To find a generating algorithm for every part of any data it gets. This is the mayor. Then to see if some small data alteration would mean a significantly simpler generation. Then to evaluate the probabilities and needed complexities. And then Spector also asks itself, what data changes are possible but which conserve already observed rules. Which is particularly handy in the unsolved Sudoku case for example.

We will do some more benchmarking this year.

An Extensive Categorisation of Infinite Paradoxes

For Eve and her apple pieces. She may eat one piece per second and stay in Paradise forever because at any given moment only a finite number of pieces has been eaten by her.

If her eating pace doubles every minute, she is still okay forever.

Only if she, for example, doubles her eating pace after every say 100 pieces eaten, then she is in trouble. If she supertasks.

Open thread, January 29 - ∞

I tend to agree. I don't know is it just a habit or something else, like a conservative profile of myself and many others, but that doesn't really matter.

The new site isn't that much better. Should be substantially better than this one for a smooth transition.

AGI

Please, focus only on what has been said and not on how it has been said.

Now, there is a possibility that all is wrong from my side. Of course I think how right I am, but everybody thinks that anyway. Including this Temple guy with his "don't code yet"! I wonder what people here think about that.

One more disagreement perhaps. I do think that this AlphaGo Zero piece of code is an astonishing example of AI programming, but I have some deep doubts about Watson. It was great back then in 2011, but now they seem stuck to me.

AGI

Knowledge is information error-corrected (adapted) to a purpose (problem).

No. Knowledge is just information. If you have some information how to solve a particular problem, it's still "just information".

There are no hard and fast rules about

howerror-corrected or towhat

Those rules are just some information, some data. How "fast and hard" are they? When there is a perfect data about the fastest checking algorithm, then it's still "just data".

The field started coding too early and is largely wasting its time.

Perhaps. How do you know what people know and who is coding already, prematurely or not?

If you joined the field, I would recommend you

do not code stuff.

I wouldn't give such an advice to everybody. I don't know what some people might know. Let them code, if they wish to.

Certain philosophy progress is needed before coding.

I agree, that you need some philosophy progress, you don't know if all others need that too. At least some may be completely unknown to you or to me.

good non-AGI work (e.g. alpha go zero, watson)

Isn't their coding premature as well?

which they hope will somehow generalize to AGI (it won't, though some techniques may turn out to be useful due to being good work and having reach)

I am not as sure as you are. They hope they will do something, you hope they will not. That's all.

wasting their time

Maybe you are a time waster Mr. Temple, yourself. Your claim that "coding AGI" is premature is just a guess. It's always possible that one is wrong, but saying "you people don't have the right theory, stop coding" ... is super-wrong. You don't know that. Nobody knows, what somebody else might know already.

people are super focused on

predictionsbut notexplanations.

A good prediction can only be done if you have a good theory/model about the mechanisms involved. So every decent predictor models anyway. The best predictor possible has a correct model. Which doesn't always imply that its predictions are right. Sometimes there isn't enough data for that. Even in principle. But to predict is to model!

some even deny there are non-empirical fields like philosophy

Some are dirty bastards also, and some have friends in low places and aunts in Australia. But you seem to imply, that all should share your view about "non-empirical fields like philosophy". Yeah, right.

It has been enough. At least my last remark I gave, was already unnecessary.

Open thread, September 18 - September 24, 2017

There are 143 primes between 100 and 999. We can, therefore, make 2,924,207 3x3 different squares with 3 horizontal primes. 50,621 of them have all three vertical numbers prime. About 1.7%.

There are 1061 primes between 1000 and 9999. We can, therefore, make 1,267,247,769,841 4x4 different squares with 4 horizontal primes. 406,721,511 of them have all four vertical numbers prime. About 0.032%.

I strongly suspect that this goes to 0, quite rapidly.

How many Sudokus can you get with 9 digit primes horizontally and vertically?

Not a single one. Which is quite obvious when you consider that you can't have a 2, 4, 6, or 8 in the bottom row. But you have to, to have a Sudoku, by the definition.

It's a bit analogous situation here.

Open thread, September 18 - September 24, 2017

Say, that we have N-1 lines, with N-1 primes. Each N digits. What we now need is an N digit prime number to put it below.

Its most significant digit may be 1, 3, 7 or 9. Otherwise, the leftmost vertical number wouldn't be prime. If the sum of all N-1 other rightmost digits is X, then:

If X mod 3 = 0, then just 1 and 7 are possible, otherwise the leftmost vertical would be divisible by 3. If X mod 3 = 1, then 1, 3, 7 and 9 are possible. If X mod 3 = 2, then just 3 and 9 are possible, otherwise the leftmost vertical would be divisible by 3.

The probability is (1/3)*(((1+2+1)/5))=4/15 that the first digit fits. (4/15)^N, that all N digit fit.

Actually, we must consider the probability of divisibility by 11, which is roughly 1/11, which further reduces 4/15 per number to 40/165. And with 7 ... and so on.

For the divisibility with 3, we render out not only one permutation of N-1 primes but all of them. For the divisibilty with 11, some of them.

It's quite complicated.

https://www.oecd.org/going-digital/ai/principles/

Either I have no clue, either ...