mxkopy

10 ms·

91.
UBCO study disproves the simulation hypothesis(news.ok.ubc.ca)
23 points by mxkopy 11mo ago|45 comments
92.
by mxkopy 11mo ago
The metaphor doesn’t work. It’s not the security of the package that’s in question, but something like whether the delivery person is getting paid enough or whether you’re supporting them getting replaced by a robot. The issue is in the con
93.
by mxkopy 11mo ago
Slightly related, there’s ways to differentiate linear programs ( https://github.com/cvxpy/cvxpylayers ), which might allow one to endow deep neural networks with some similar reasoning capabilities as these sorts of sol
94.
by mxkopy 11mo ago
I disagree, I think the leap of faith is to believe that something in our brains made of physical building blocks can’t be replicated on a computer that so far we’ve seen is very capable of simulating those building blocks
95.
by mxkopy 11mo ago
Right, I’m not the most well read on this stuff either, so I’m wondering now if existing architectures operate on this > efficient frontier of parsimony against say time complexity. As you mentioned before regularization approximates par
96.
by mxkopy 11mo ago
Ah, what I was hoping to get at is that true intelligence might not have these big gaps between explanation-lengths that parsimonious TMs do. And there’s also the question of deduction; having a few redundant “theorems” on hand might make d
97.
by mxkopy 11mo ago
Are ”noisy” inputs here at all related to ones where their Kolmogorov complexity is their encoding length? I don’t know how much I buy the idea that intelligence maximizes parsimony. Certainly true for inductive reasoning but I feel like th
98.
by mxkopy 11mo ago
I agree, gradient descent implicitly assumes things have a meaningful gradient, which they don’t always. And even if we say anything can be approximated by a continuous function, we’re learning we don’t like approximations in our AI. Some d
99.
by mxkopy 1y ago
Not affiliated with anyone, but I think the likes of OptNet (differentiable constraint optimization) are soon going to play a role in developing AI with precise deductive reasoning. More broadly I think what we’re looking for at the end of
100.
by mxkopy 1y ago
Thats the same thing, no? https://en.wikipedia.org/wiki/Entropy_(information_theory)?w...
101.
by mxkopy 1y ago
In this post I go over training LPs using cvxpylayers, but I think the more interesting discussion lies at the end, where I argue that current AI models lack the ability to reason counterfactually in discrete settings & that LPs can bri
102.
Learning sudoku by doing gradient descent on a linear program(mxkopy.github.io)
3 points by mxkopy 1y ago|1 comments
103.
by mxkopy 1y ago
What if the IDE is a LeapFrog 2-in-1 Educational Laptop
104.
by mxkopy 1y ago
The differentiating factor is that Norway wasn’t colonized.
105.
by mxkopy 1y ago
Scrolling TikTok I get cool videos from underground artists, memes and skits Scrolling HN I find hateful stances about hyperspecificties, overconfident self righteousness, and new peaks of misanthropism
106.
by mxkopy 1y ago
If C is the same for all K, then can’t he extrapolate the trend from K = 2 (which you’re right, I did misread) to all other values of K? That seems to be his justification for the claim you’re arguing. I do think some more rigor is needed h
107.
by mxkopy 1y ago
In the graph you’re referencing, K = 2 never reaches C = 0.2. K = 3 only reaches C = 0.3 before getting cut off. I’m not even sure why that would be a problem, because he’s projecting N basis vectors into K dimensions and C is some measure
108.
by mxkopy 1y ago
“If it’s just the south” like the south isn’t the larger of two cultures that define the U.S. historically. Again, it’s not out of pocket for the US to be blatantly racist. You say these are extreme beliefs (and they are objectively) but in
109.
by mxkopy 1y ago
Check my edit. It’s not a straw man because the poor uneducated south never left their attitudes towards slavery behind (or otherwise rediscovered/were reintroduced to them for political benefit). Believe it or not, people actually hav
110.
by mxkopy 1y ago
Under RTO new grads enjoy less competition because they’re willing to relocate
111.
by mxkopy 1y ago
Nice, my undergrad thesis used this stuff. I truly believe if P=NP the proof will be an mILP solving HC or similar
112.
by mxkopy 1y ago
For every addiction you enable with more free time there’s an overworked but capable and loving parent on the other side of the equation. That’s why your argument isn’t really a rebuttal but a counterfactual based on an opinion. The reason
113.
by mxkopy 1y ago
I guarantee that a 20 hour workweek would fix this problem without having to invade anyone’s privacy, but we can’t have that for obvious reasons. My friends with healthy attachments to social media had healthy and present parents. You have
114.
by mxkopy 1y ago
You can look up the figure pretty easily but from what I’ve glanced it seems to be related to his and his son’s crypto schemes, touted through official WH channels
115.
by mxkopy 1y ago
You misunderstand. From a billionaire’s POV a real contribution looks like taking some responsibility for the average person’s wellbeing even if it has nothing to do with their business interests, simply because they have the power to do th
116.
by mxkopy 1y ago
You should be forced to invest into all aspects of the social fabric that makes your success possible. Your reward is being able to be a successful philanthropist
117.
by mxkopy 1y ago
Well, in terms of architectural complexity you have to wonder what something intelligent is going to look like, it’s probably not going to be very simple, but that doesn’t mean it can’t be readily interpreted. For the brain we can ascribe s
118.
by mxkopy 1y ago
From what I’ve seen neurology is very readily interpretable but it’s hard to get data to interpret. For example the visual cortex V1-V5 areas are very well mapped out but other “deeper” structures are hard to get to and meaningfully measure
119.
by mxkopy 1y ago
This statement is also supported by the fact that despite the relatively high taxes in NYC billionaires will jump through all loops necessary to effectively still live there just not on paper
120.
by mxkopy 1y ago
There’s the hierarchical reasoning model https://arxiv.org/abs/2506.21734 but it’s very new and largely untested Though honestly I don’t think new neural network architectures are going to get us over this local maximu

More ›