sixo

8 ms·

61.
by sixo 1y ago
It probably is a lot like that! I imagine it's a matter of specializing the networks and learning algorithms to converge to world-model-like-structures rather than language-like-ones. All these models do is approximate the underlying m
62.
by sixo 1y ago
Roughly, actual intelligence needs to maintain a world model in its internal representation, not merely an embedding of language, which is a very different data structure and probably will be learned in a very different way. This includes t
63.
by sixo 1y ago
This feels like it's not accounting at all for changes in enforcement.
64.
by sixo 1y ago
This is a nonsensical opinion by a person who doesn't know what they're talking about, and probably didn't read the article. These models are tools, and LLM products bundles these tools with other tools, and 90% of UX amounts
65.
by sixo 1y ago
You really missed my point. All your objections either: 1. Also apply to the present day pricing of property tax: - "How do the police and fire departments set the fees?" - "Are they fixed for everyone in their area&qu
66.
by sixo 1y ago
Ha I did. But only because, your arguments only make sense if it's not really required, or if it's implemented so poorly as to be effectively not required. I don't see the point of criticizing a policy idea with an argument l
67.
by sixo 1y ago
Those are all good arguments one would have to deal with to implement this thing well! I'm not sure the case for one-big-bundle-of-property-taxes is any better, though. Where does THAT number come from? > Every discussion I've
68.
by sixo 1y ago
Oh, I know, that's the destabilization I mentioned in the first case. I don't think you do this in isolation. Ideally you rationalize the property tax while compensating for its progressive/welfare effects with other policies
69.
by sixo 1y ago
We get around both of those by requiring buildings be insured, just like cars. That word was in my original post, but you missed it and gave the standard counter-arguments to a different (and bad) idea instead. This isn't meaningfull
70.
by sixo 1y ago
It seems like you could straightforwardly unbundle the property tax into a bunch of specific fees and insurances. We're comfortable requiring car insurance, why not fire and police? This would probably make it more palatable to MAGA, b
71.
by sixo 1y ago
Just like in real life. Wait a sec...
72.
by sixo 1y ago
Location: Currently NC, but I'd like to move back to NYC. Open to in-person in either state. Remote: Open to full remote or hybrid in NC/NYC Willing to relocate: Possibly within the East Coast. Technologies: Python, SQL, some Scal
73.
by sixo 1y ago
The usual calculus here is that they don't capture ALL the upside of that training, but they capture plenty of it.
74.
by sixo 1y ago
You are spamming this and similar paragraphs all over this thread, and you come off as a crackpot.
75.
by sixo 1y ago
It's not even worth mentioning this problem unless you talk about how the result depends on the data generating process. If you take it to be something like "you randomly sample from families with two children, discarding any with
76.
by sixo 1y ago
ah, I mean, working in the sense of OP: that a system which overcomes the "engineering hurdles" is actually feasible and will be successful. To be blocked merely by "engineering hurdles" puts QC in approximately the same
77.
by sixo 1y ago
On the contrary, we have one working example of general intelligence (humans) and zero of quantum computing.
78.
by sixo 1y ago
> which means we don't learn anything usefully actionable from the test and shouldn't apply it. This just isn't true. In practice any such screening model can ALWAYS improve with more data—basically because the statistical
79.
by sixo 1y ago
huh, well, if it's true that "no amount of data will reduce them" then my point is wrong, but I highly doubt that's true.
80.
by sixo 1y ago
Well, you missed my point. I'm not talking about "you look at the MRI and see something and say it's a positive", I'm referring to the process of reading MRIs as like a statistical model (even if in practice it exis
81.
by sixo 1y ago
The MRI analogy is not good. The false positive risk is only against the present-day distribution of MRIs mostly taken of symptomatic patients; if we had the dataset of "annual MRIs for everyone" we would very quickly recalibrate
82.
by sixo 1y ago
In the space of floats, you are dividing by zero. To map back to the space of numbers you have to check. It's nice, though; inf and NaN sentinels give you the behavior of a monadic `Result | Error` pipeline without having to wrap you
83.
by sixo 1y ago
Skill issue. There are other ways to get those results, but being an asshole is the lowest-hanging, and is nearly free if the people around you don't have the self-respect to walk away.
84.
by sixo 1y ago
At one point in my life I liked crafting code. I took a break, came back, and I no longer liked it--my thoughts ranged further, and the fine-grained details of implementations were a nuisance rather than ~pleasurable to deal with. Whatever
85.
by sixo 1y ago
That determines the cost effectiveness to make it worth it to train one of these models in the first place. Using someone else's weights, you can afford to predict quite inefficiently.
86.
by sixo 1y ago
No, it's what you do if your model architecture is capped out on its ability to profit from further training. Hand-wrapping a bunch of sub-models stands in for models that can learn that kind of substructure directly.
87.
by sixo 1y ago
Based only on this comment thread—because I have no interest in adjudicating the actual dispute here—I see playing out in your post, for about the 1000th time in my life, the motte-and-bailey of "prosecuting rape culture". The OP,
88.
by sixo 1y ago
https://pyformat.info/
89.
by sixo 1y ago
oh. I did not interpret that blob as a cow. Thanks.
90.
by sixo 1y ago
A few questions occur to me while reading this, which I am far from qualified to answer: - How much of this structure survives if you work on "fuzzy" real numbers? Can you make it work? Where I don't necessarily mean "fu

More ›