Posts
Comments
* Hinton's daughter reading this: "Oh dad - not again!"
I don't know what this was a reference to, but amusingly I just noticed that the video I wanted to link was a 2007 lecture at Google by him (if it's the same Geoffrey Hinton): https://www.youtube.com/watch?v=AyzOUbkUf3M
In it he explained a novel approach to handwriting recognition: stack a bunch of increasingly small layers on top of each other until you have just a few dozen neurons, then an inverted pyramid on top of this bottleneck, and train the network by feeding it a lot of handwritten characters using some sort of modified gradient descent to train it to reproduce the input image in the topmost layer as accurately as possible. After the network is trained, use supervised learning with labeled data to train a usual small NN to interpret/map the bottleneck layer activations to characters. And it worked!
I find it interesting, especially in context of your comment, because:
- Unsupervised learning meant that you could feed it a LOT of data.
- So you could force it to learn patterns in the data without overfitting.
- It appeared uninspired by any natural neuronal organization.
- It was a precursor to Deep Dream - you could run the network in reverse and see what it imagines when prompted with a specific digit.
- It actually worked! and basically solved handwriting recognition, as far as I understand.
And so it felt like a first qualitative leap in technology in decades, and a very impressive at that, innovating in weird and unexpected ways in several aspects. Sure, it would be another ten years until GPT2, but some promise was definitely there I think.
According to the links from the Scott's post, Rob Reich's position is that we should tax charitable donations at the exact same rates as all other spending, with an exception for under $1000/year donors getting 25% back. No more, no less.
I personally think that this is a blindingly stupid idea because it assumes that everyone who donates more than that will donate even more to compensate for the government taking a lion's share of their donations, because he sort of got himself into a frame of mind where he sees donations as more of a privilege to change the world according to one's wishes given to the donors, not as a lifeline for the recipients.
But nothing in the two articles about his position I read suggests anything more sinister than that misguided plan, which even makes sense on his own terms.
Keep in mind that a lot of user-submitted programs will try to do the same (because writing a step-by-step interpreter is hard), so they would keep evaluating each other spawning watchdog threads every time, so, um, your watchdog thread would be badly outnumbered.
The easy fix for you would be to run your watchdog in a separate process, but players wouldn't have this ability, which might make things either more interesting or boring (ruling out all strategies using eval). Maybe a specially designed restricted subset of Scheme with time-restricted eval would be a better choice?
By the way, after looking at your payout matrix to see what should I do if I see the opponent using eval, looks like you have in fact created a version of PD with three choices. Not incentivizing the third choice doesn't really help because a program still has to consider the possibility that the other program chooses "other" due to a bug, in which case it should choose Defect.
I suggest you implement the standard PD by declaring that anything that is not Cooperate is Defect. The only downside would be that you'll see somewhat more programs using all their allotted 10s, but you'll see a lot of those either way. At least you'll be able to say that this competition was about the actual PD.