< Earlier Kibitzing · PAGE 4 OF 39 ·
Later Kibitzing> |
Jan-10-18 | | WorstPlayerEver: Anand did not know they spent $25M. Speaking of brute force. |
|
Jan-12-18 | | zborris8: <So many uninformed comments.> Proving that artificial intelligence is no match for natural stupidity! |
|
Jan-13-18 | | zborris8: AlphaGo has an IQ of 47.28, but a 6yr old has 55. Paper:
https://arxiv.org/ftp/arxiv/papers/... AI will reach an IQ of 10,000 in the upcoming years. |
|
Jan-27-18
 | | Ron: Here is the academic paper of AlphaZero:
https://arxiv.org/pdf/1712.01815.pdf
I suggest this to be chessgames.com Quote of the Day: "The parameters θ of the deep neural network in AlphaZero are trained by self-play reinforce- ment learning, starting from randomly initialised parameters θ. Games are played by selecting moves for both players by MCTS, at ∼ πt. At the end of the game, the terminal position sT is scored according to the rules of the game to compute the game outcome z: −1 for a loss, 0 for a draw, and +1 for a win. The neural network parameters θ are updated so as to minimise the error between the predicted outcome vt and the game outcome z, and to maximise the similarity of the policy vector pt to the search probabilities πt. Specifically, the parameters θ are adjusted by gradient descent on a loss function l that sums over mean-squared error and cross-entropy losses respectively,
(p,v)=fθ(s), l=(z−v)2 −π⊤logp+c||θ||2
where c is a parameter controlling the level of L2 weight regularisation. The updated parameters are used in subsequent games of self-play." |
|
Jan-27-18 | | ChessHigherCat: And to think parents forbid their child to engage in self-play for fear of learning disabilities! |
|
Jan-27-18
 | | alexmagnus: <zborris8> An AI programmed specifically to pass IQ tests once scored 150. So much for the bad old IQ. |
|
Jan-27-18 | | john barleycorn: <alexmagnus: <zborris8> An AI programmed specifically to pass IQ tests once scored 150. ...> Well, there are books out there to train yourself. to break 150 in an IQ test just requires a good memory.... |
|
Jan-27-18 | | ChessHigherCat: <AlphaGo has an IQ of 47.28, but a 6yr old has 55.
Paper:
https://arxiv.org/ftp/arxiv/papers/...That notion of "Absolute IQ" in the paper you cited seems suspect to me, because IQ is generally supposed to be measured as a function of age, which means that the average 6-year-old will always have an IQ of 100. |
|
Jan-27-18
 | | alexmagnus: The original definition of IQ was "mental age divided by biological age times 100, with mental maturity reached at 20". By this, a 6-year-old has an "absolute IQ" (that is, measured as if he were an adult) of 30. |
|
Jan-27-18 | | john barleycorn: <alexmagnus: The original definition of IQ was "mental age divided by biological age times 100, with mental maturity reached at 20". By this, a 6-year-old has an "absolute IQ" (that is, measured as if he were an adult) of 30.> ????? |
|
Jan-27-18 | | ChessHigherCat: <alexmagnus> Thanks for the formula. I still find the whole concept of "absolute IQ" dubious, though. I mean, take the case of 3-year-olds. You could devise a special test for them (of the figure-out-how-to-get-your-greedy-litte-hands on the Gerber's type) and then map their performance on a Bell curve and assign an IQ of 100 to the mean (i.e., average, not nasty :-) and the rest based on the standard deviations, but there's no way you could have them take a standard test designed for 20-year-olds because they don't even know how to read. Anyway, it all depends on the type of test. If it involves a lot of interpretation of verbal instructions computers suck absolutely. |
|
Jan-27-18 | | ChessHigherCat: <John Barleycorn> I think he means an average 6-year-old with a relative (normal) IQ of 100 has an "absolute IQ" of X relative to the average 30-year-old with a relative (normal)
IQ of 100 (although why should it be 30 if the age of "mental maturity' is supposed to be 20?)
According to the usual IQ formula, a 6 year old who solves problems as well as the average 9-year-old has an IQ of 150: https://www.merriam-webster.com/dic...
medical Definition of IQ. : a number used to express the apparent relative intelligence of a person based on the results of standardized testing: as. a : the ratio of the mental age (as reported on a standardized test) to the chronological age multiplied by 100. Not only are 6-years-old incapable of reading the instructions of standardized test for adults but the extrapolation of IQ to computers seems absurd because computers have nothing like "age" so the whole approach seems misguided to me |
|
Jan-27-18 | | john barleycorn: <ChessHigherCat: <John Barleycorn> I think he means an average 6-year-old with a relative (normal) IQ of 100 has an absolute IQ of X relative to a 30-year-old (although why should it be 30 if the age of "mental maturity' is supposed to be 20?) ...> Yes, he was not very precise in his statement. and then who wants/needs/whatever compare a 6 year old to a 20 year old? Like *we* all know that for instance the ability for logical thinking is developed after puberty and that there are some early talents and some who will never learn it. |
|
Jan-27-18
 | | alexmagnus: The "mental maturity reached at 20" clause is given in the sense that past 20 those original tests should be read as if the testee were 20. The original IQ tests were devised to detect mentally disabled children. This is why it was defined through age. |
|
Jan-27-18 | | morfishine: <ChessHigherCat> I agree about the approach being misguided. I've taken a bunch of IQ tests on the internet for fun. They all have some disclaimer about interpretation. The weird thing is despite the disparate testing companies, I score virtually the same every time. So if there's nothing concrete to be interpreted, at least the meaningless results are consistent LOL ***** |
|
Jan-27-18 | | john barleycorn: <morfishine: <ChessHigherCat> I agree about the approach being misguided. I've taken a bunch of IQ tests on the internet for fun. They all have some disclaimer about interpretation. ...> I guess that help you a lot. |
|
Jan-27-18 | | john barleycorn: <alexmagnus: The "mental maturity reached at 20" clause is given in the sense that past 20 those original tests should be read as if the testee were 20. ...> Ah yes. I am 61 now. what would that result in? |
|
Jan-27-18 | | ChessHigherCat: Anyway, what would the questions look like in the IQ test for computers.Take a question that would be easily understood by most Americans with an IQ of 80: John drives a vintage 1941 Rolls Royce and Joe has a 1975 Oldsmobile. Which one is probably richer? I guarantee you no supercomputer will understand that question. You can say, well it's culturally biased but exactly what culture does a computer have? To eliminate the "humanoid-bias" you would have to ask questions like "What is the cube root of 17 divided 96?", but that's reducing the whole notion of human intelligence to the level of a pocket calculator! |
|
Jan-29-18 | | Big Pawn: I wonder if they will have Alpha Chess play anymore matches? The result here was stunning, no question about it, and I found these games to be the most interesting games in who knows how long. It's not every day you see Stockfish getting beaten up like this. Has anyone heard any news about A0 since this match? |
|
Jan-30-18
 | | AylerKupp: Those are are interested in the IQ of AlphaZero or artificial intellingence in general might be interested in these papers: https://arxiv.org/ftp/arxiv/papers/... and https://arxiv.org/ftp/arxiv/papers/... . And when you're through reading them, please explain them to me in terms that I might be able to understand. |
|
Jan-30-18
 | | keypusher: < john barleycorn: <alexmagnus: The "mental maturity reached at 20" clause is given in the sense that past 20 those original tests should be read as if the testee were 20. ...>
Ah yes. I am 61 now. what would that result in?> Decades of mental decline. I'm right behind you. |
|
Jan-30-18 | | john barleycorn: <keypusher: ...
Decades of mental decline. I'm right behind you.> Good to hear that I have followers ... |
|
Feb-03-18 | | zborris8: <Ayler Kupp IQ papers> Beat you to it! AlphaZero (Computer) (kibitz #78) |
|
Feb-11-18
 | | AylerKupp: <<zborris8> Beat you to it!> You say that as though beating me at anything (except perhaps verbosity) was a hard thing to do ?! But, as a "reward" for beating me, you get to explain the paper to me. :-) |
|
Feb-11-18 | | WorstPlayerEver: What I feared became true. Computers are just senseless machines which rely on calculations. The engine. I based my observation on the materialistic mind of its developers to figure out their weaknesses. Needless to state I had to prevail ☺ |
|
 |
 |
< Earlier Kibitzing · PAGE 4 OF 39 ·
Later Kibitzing> |