What if the AlphaGo AI is so advanced, that it is toying with its opponent, and letting him near-win? What if it has the strength of "the Hulk" with the subtlety of Black Widow?
This is completely misunderstanding the nature of "soft AI" contained in Alphago. You're clearly talking about hard AI, which AlphaGo is not.
AlphaGo doesn't doesn't know that there is a world outside Go. It doesn't learn. It doesn't know the meaning of 'taunt.' It doesn't know it's playing a game against the world's top human. It can't self-introspect, it can't change its programming, and it doesn't know a Go board looks like, and doesn't even know what Go is.
AlphaGo is a classifier. Given a set, it will score potential additions to the set based on previous data. It also has some code to coordinate that with a Go board, but that is mundane code, it's not even soft AI.
Note that when I say "p hacking" I don't mean that they intentionally lied, just that statistics is hard and sometimes these effects will come out. I think that the hacking technique they used (and I say this without getting my equipment to reproduce, so caveat emptor) is to accept only a very small difference, so if a gene made even the difference of a small fraction of an IQ point, they counted it. It is hard to separate counfounding factors at such a small level. In addition, they didn't give these people IQ tests, instead they used education level as a proxy for IQ, which is a very very rough approximation.
The authors of the paper are hedging a bit on the finding, saying it's a starting point for further research, here's a quote:
as the authors point out, the work was meant to be a foundation rather than the last word: "These findings provide starting points for understanding the molecular neurobiological mechanisms underlying intelligence."
Finally, previous studies that have found genes for intelligence later didn't hold up under reproduction. So in that sense, this study would be typical if the same thing happened to it.
Of course I'm open to the idea that I'm wrong and we should wait to see if the study gets reproduced before drawing any strong conclusions.
Races are artificial and are scientifically meaningless; genetics is reality.
This is worth repeating. 'Race' is largely determined by skin color and a small number of other features. It's not a very good way to measure someone's other genetic traits.
Also worth mentioning the effect here was very small, each gene just a small fraction of an IQ point. And I think they used education level as a proxy for IQ test, since they didn't test everyone in the sample set.
It's very possible. The effect found by each of these genes was very small, a fraction of an IQ point. At that small size, I would doubt that I had accounted for all confounding variables. Something as simple as hair color might be a confounding factor, and height certainly would be.
“If you try to predict height using the genes we’ve identified in Europeans in Africans, you’d predict all Africans are five inches shorter than Europeans, which isn’t true,” Dr. Posthuma said.
Jesus christ, if you DON'T trust any sort of institution then you might as well be a freaking anarchist/prepper and this whole "society" thing just isn't for you. Because what's the alternative?
Democracy exists exactly because we don't trust our institutions. So we have horribly expensive elections every few years. It's wasteful, but it keeps people who are too bad out of office.
It's interesting you bring that up, because the last book I read about neuroscience mentioned the work of amateurs recently in the field. Connectome, it's a good book, you should read it.
Yeah that's true, it's worth remembering that the new version could be different. I expect it to just be an improved version of the same, though.
The interesting thing about the way it operated was that it could tell you which move was likely best, but could not explain why. The same is actually true of human Go players. While locally best moves can be identified, the human selects the globally best move based to a large extent on feel.
I can't speak for Go, but if you start reading chess books written by grandmasters, you'll see a lot of explanations on what they are thinking. Although the grandmasters aren't neuroscientists, so they don't have a cohesive theory, you can start to piece together the way they think.
It doesn't have to be perfect, it just has to be good enough.
If you write crap code, it will slow you down, a lot. In fact, that can be one definition of crap code: code that slows you down.
Yeah. If it's a one-off thing, it's not going to mess up your codebase.
The sales team fucked that one up, and it's definitely nice of you to help them out. Just don't let them take advantage of you by making it a habit.
There are ways to deal with it.
Unless you're a lazy developer and spend half your time on Reddit, then it doesn't matter if your boss pressures you. The software won't get done before you estimated it.
Slashdot Mirror
What if the AlphaGo AI is so advanced, that it is toying with its opponent, and letting him near-win? What if it has the strength of "the Hulk" with the subtlety of Black Widow?
This is completely misunderstanding the nature of "soft AI" contained in Alphago. You're clearly talking about hard AI, which AlphaGo is not.
AlphaGo doesn't doesn't know that there is a world outside Go. It doesn't learn. It doesn't know the meaning of 'taunt.' It doesn't know it's playing a game against the world's top human. It can't self-introspect, it can't change its programming, and it doesn't know a Go board looks like, and doesn't even know what Go is.
AlphaGo is a classifier. Given a set, it will score potential additions to the set based on previous data. It also has some code to coordinate that with a Go board, but that is mundane code, it's not even soft AI.
The authors of the paper are hedging a bit on the finding, saying it's a starting point for further research, here's a quote:
as the authors point out, the work was meant to be a foundation rather than the last word: "These findings provide starting points for understanding the molecular neurobiological mechanisms underlying intelligence."
Finally, previous studies that have found genes for intelligence later didn't hold up under reproduction. So in that sense, this study would be typical if the same thing happened to it.
Of course I'm open to the idea that I'm wrong and we should wait to see if the study gets reproduced before drawing any strong conclusions.
Democracy IS an institution. Doubly so for the US electoral college, the system of representation, and the government we elect through democracy.
Yes, and it's been rife with corruption. Don't you know your history? Any time the people stop paying attention, corruption will sneak in.
Yeap :)
Uber's about to get slapped down really hard by the law.
Probably not since the average IQ 200 years ago was less than five points below the average today (see the Flynn affect).
The paper is probably false, with a P value < .05. Sad but true.
Races are artificial and are scientifically meaningless; genetics is reality.
This is worth repeating. 'Race' is largely determined by skin color and a small number of other features. It's not a very good way to measure someone's other genetic traits.
Also worth mentioning the effect here was very small, each gene just a small fraction of an IQ point. And I think they used education level as a proxy for IQ test, since they didn't test everyone in the sample set.
It's very possible. The effect found by each of these genes was very small, a fraction of an IQ point. At that small size, I would doubt that I had accounted for all confounding variables. Something as simple as hair color might be a confounding factor, and height certainly would be.
“If you try to predict height using the genes we’ve identified in Europeans in Africans, you’d predict all Africans are five inches shorter than Europeans, which isn’t true,” Dr. Posthuma said.
Bush was smart. He didn't just cut taxes for the rich, he cut them for everyone. No one wanted to raise taxes again after that, not even democrats.
Jesus christ, if you DON'T trust any sort of institution then you might as well be a freaking anarchist/prepper and this whole "society" thing just isn't for you. Because what's the alternative?
Democracy exists exactly because we don't trust our institutions. So we have horribly expensive elections every few years. It's wasteful, but it keeps people who are too bad out of office.
I know plenty of history.
I guess you're going to repeat it.
You're wrong. Re-calibrate your intuition.
It's interesting you bring that up, because the last book I read about neuroscience mentioned the work of amateurs recently in the field. Connectome , it's a good book, you should read it.
The interesting thing about the way it operated was that it could tell you which move was likely best, but could not explain why. The same is actually true of human Go players. While locally best moves can be identified, the human selects the globally best move based to a large extent on feel.
I can't speak for Go, but if you start reading chess books written by grandmasters, you'll see a lot of explanations on what they are thinking. Although the grandmasters aren't neuroscientists, so they don't have a cohesive theory, you can start to piece together the way they think.
It doesn't have to be perfect, it just has to be good enough.
If you write crap code, it will slow you down, a lot. In fact, that can be one definition of crap code: code that slows you down.
Another might be "code that is hard to read"
ok, that's fine if you want to do it that way, but if it's affecting your code quality, then you're doing it wrong.
Same with XP or anything else that denies you need good people to get good results. Framework nazis rarely get shit done in my experience
These are some good sentences. The key is to take your people and make them better, not to hope some process will save you.
While these situations are not the norm,
Yeah. If it's a one-off thing, it's not going to mess up your codebase.
The sales team fucked that one up, and it's definitely nice of you to help them out. Just don't let them take advantage of you by making it a habit.
There are ways to deal with it.
Unless you're a lazy developer and spend half your time on Reddit, then it doesn't matter if your boss pressures you. The software won't get done before you estimated it.
It's entirely developments fault. Most managers will ask, "How long do you think it will take to get this done?"
The days when tech was dominated by people who wanted to be there are long gone.
Which is unfortunate. Those people make things bad for the rest of us. If they figured out how to enjoy it, things would be better.
They don't work. The main use case succeeds, but trailing along are all the bugs in every corner case that customers hate.
Yeah. The human pruning algorithm is actually really amazing in how quickly it works. Tree searching is slow and hard for humans, but oh well.