xamuel

33 ms·

301.
by xamuel 7y ago
>You shouldn't be able to see exact activity with details, no reason for that. If I urgently need to meet with someone today and I see they've blocked off the whole afternoon for a 4-hour meeting, it could make a big difference
302.
by xamuel 7y ago
By human beings. And he did not contribute to AI capable of contemplating human research like a human does, because no such AI exists yet. When it does exist (if ever), it will surely not consider this person or his hundred lab workers to h
303.
by xamuel 7y ago
Sounds like someone reeeeeallly good at nudging envelopes and writing grant proposals. Will be mostly forgotten a couple hundred years from now. Academic playwrights thought Shakespeare was a joke in his day.
304.
by xamuel 7y ago
What revolution in games do you have in mind? Is there any particular video game that is making really interesting use of ML?
305.
by xamuel 7y ago
I'd rather not say just on the slim chance of compromising confidentiality. So instead here are my favorite dialogues so far (of the ones I've read): Theaetetus, Philebus. I found Phaedrus and Symposium a little overrated. Haven&#
306.
by xamuel 7y ago
We sweep the ancients aside at our peril. I've been reading the works of Plato during part of my commute for the past few months. And for the past six months or so I've been one of several peer reviewers for a certain philosophy-o
307.
by xamuel 7y ago
In the context of political opinions, I was thinking people like Adam Smith, Karl Marx, etc. But you're right, even those people stood on the shoulders of giants, they didn't just independently get struck by lightning one day and
308.
by xamuel 7y ago
The list of people whose opinions are "really their own" is pretty much limited to feral children and certain philosophers/scientists
309.
by xamuel 7y ago
Not really, unless I'm missing something. They have thematic similarities, I suppose.
310.
by xamuel 7y ago
The argument is for deterministic environments--no initial condition dependence. If initial conditions are allowed, then you can consider the following environment: initial condition is an arbitrary source-code; the environment proceeds to
311.
by xamuel 7y ago
That core point is like a variation on a lemma in a paper I recently published (Lemma 2 in [1]): There can be no single deterministic interactive reward-giving environment which can, by its lonesome, serve as a good proxy for general intell
312.
by xamuel 7y ago
>Studying various technical disasters is something of a hobby/professional interest What would be some recommended reading in this area? Not limited to airplanes
313.
by xamuel 7y ago
The paper isn't intended to be a manual for how to practically compare agents. It will indirectly help there, I hope, by making people realize that they're looking at an election problem, and that there's a big existing liter
314.
by xamuel 7y ago
The computability problem can be addressed by using Levin complexity instead of Kolmogorov complexity, an approach which you can read here: http://users.monash.edu/~dld/Publications/2010/HernandezOral... It s
315.
by xamuel 7y ago
Right, the environments are not uniformly distributed. In fact, the paper actually defines not one single intelligence comparator but an infinite family, parametrized by a hyperparameter which is, essentially, a choice of which environments
316.
by xamuel 7y ago
As others pointed out, the NFLT only applies if the environments are uniformly distributed. In the paper, they are not uniformly distributed.
317.
by xamuel 7y ago
The #1 thing is it led to an elegant notion of how to compare intelligence (or rather, a parametrized family of intelligence comparators). There's a theorem called Arrow's Theorem, which basically says there's no good way to
318.
by xamuel 7y ago
So if you want "Carmack's Theorem" or "Noelsusman's Theorem" to appear in high school math textbooks 500 years from now, this is the best time to jump in. (Assuming there are still textbooks 500 years from now.
319.
by xamuel 7y ago
The paper addresses this. Rather than (futilely) attempt to nail down the "one true electorate", the paper defines an infinite family of comparators, depending on a hyperparameter which is, essentially, the choice of which environ
320.
by xamuel 7y ago
The "voters" are not humans, but environments. A simple example of an environment would be a room with two buttons, one of which rewards you when you press it, and one of which punishes you when you press it. Independently of anyt
321.
by xamuel 7y ago
I've been dabbling in AGI and it seems like the field has a lot of low-hanging fruit. I'll bet Carmack can offer some significant contributions. I'll take an opportunity to plug a paper I recently published on comparing relat
322.
by xamuel 7y ago
I meant in my example about undiagnosed lead in Flint. Once you notice that the AI detects people in Flint under-perform, why would you want to modify the AI to avoid the AI detecting that? Your cancer example doesn't answer the questi
323.
by xamuel 7y ago
When would you want to hack the AI? Not a rhetorical question. Reading through the comment thread I'm slowly shifting more towards being on the fence, as opposed to having a "don't hack the AI" position. But rather than
324.
by xamuel 7y ago
Hypothetical question. Suppose no-one ever noticed the lead in the Flint water supply. People put test scores into an AI and the AI emotionlessly concludes that people from Flint perform worse than average. Should you hack the AI to force i
325.
by xamuel 7y ago
You seem to have this idea that people pick up the Bible, read "Jesus walked on water", and go "Wow, praise God!" In actuality, what matters isn't the claims the Bible makes, but the actual words that Jesus and the
326.
by xamuel 7y ago
I remember playing this game for hours and hours as a little kid, on an old DOS computer. A fun meta-game is to try and just keep the game going as long as possible, which means trying to find the right balance of keeping at least one compe
327.
by xamuel 7y ago
I wonder whether the author of Conflict was trying to make an uber-realistic simulation, or a fun game. If the latter, the game was a great success, it's a fantastic game. At some point you have to let chess be chess, you shouldn'
328.
by xamuel 7y ago
What we need is for someone to apply deep neural networks to astrology. Sure, star-signs never worked in the past, but do you realize how powerful GPUs are these days? Now please give me a $500mil seed round
329.
by xamuel 7y ago
Possible plot twist: the girls were into astrology because it gave boys an easy way to approach and initiate conversation with them ;)
330.
by xamuel 7y ago
Or neural nets trained to detect cancers from patient photographs, that end up learning "if the patient has a ruler next to them, then they have cancer"

More ›