gradys

7 ms·

91.
by gradys 5y ago
Maybe just a Python notebook with a Postgres client library and some helper functions to keep the amount of Python in the main body to a minimum?
92.
by gradys 5y ago
What is ridiculous exactly? That it's deflationary? Or the intuition that it should rise in the face of USD inflation (all else equal)?
93.
by gradys 5y ago
You'd think an intrinsically deflationary asset like Bitcoin would move differently in the face of USD inflation. Any explanation for why this might be?
94.
by gradys 5y ago
Shout out to the attrs library. If you find yourself using namedtuple a lot, it's worth considering despite it being an external dependency. https://www.attrs.org/en/stable/
95.
by gradys 5y ago
I don't have comment on the legacy of supply-side economics and and how successful e.g. Reagan's economic policy was, but it looks to me like we have more of a supply-side issue right now than demand-side. We have shortages of sem
96.
by gradys 5y ago
Why is the no diamond constraint useful here?
97.
by gradys 5y ago
Seems like you should sue?
98.
by gradys 5y ago
It works much better when using immutable data and mostly side-effect-free functions as is heavily encouraged by Clojure.
99.
by gradys 5y ago
But maybe with 5 million real examples sampled from the distribution CIFAR-10 was sampled from they would in fact see a difference. Maybe the generative model is capturing only a limited slice of the diversity that the ideal model would rea
100.
by gradys 6y ago
The computer as a tool for an individual to use is a perspective that comes naturally to us now, but note that this article was written before even the Alto and the Apple II, and a decade before the IBM PC. Interactive computing was a fring
101.
by gradys 6y ago
Do you think there's any chance that this was actually what a majority of Californians wanted? You seem very confident that you know better what the voters wanted than they did. While differences in spending surely have some effect, sp
102.
by gradys 6y ago
My understanding of the term "neoliberal" doesn't include relentlessly lowering taxes or shrinking government. Maybe you mean "neoconservative" or "libertarian"? To me, neoliberalism is better characterize
103.
by gradys 6y ago
Yep, it would definitely be difficult to justify running it in production. Accuracy would need to be higher as you said or it would need to be applicable to more tasks such that you can take other models out of production. This kind of mode
104.
by gradys 6y ago
One key idea here is to use a very large number of parameters (model weights), but only use some subset of the parameters on each example. The parameters are divided up into blocks called "experts", and then some subset of experts
105.
by gradys 6y ago
Mixture of experts architectures like this are a specific design decision to increase the parameter count but use and update those parameters sparsely. Sure, that design decision doesn't fit all scenarios, but it fits some, and it has
106.
by gradys 6y ago
How do I make sure I hear about Invisible Sun when it comes out?
107.
by gradys 6y ago
I guess maybe I'm alone then, but I don't hate using Amazon.
108.
by gradys 6y ago
That optimization makes the 3D games more complex, not less. What Factorio does is impressive and surely comes with its own challenges, but it isn't necessarily more complex.
109.
by gradys 6y ago
Any evidence to back up the bot speculation? Normal social media memey exuberance seems sufficient to explain this to me.
110.
by gradys 6y ago
This is a great tip, thanks! Is this essentially importance sampling? Reminds me of a technique for off-policy RL.
111.
by gradys 6y ago
Sorry, my question wasn't very clear. I guess what I meant to ask is whether this works when eating food that doesn't have as as much acid as a straight lemon. Sounds like the answer is yes. Also, tangential, but I think the advic
112.
by gradys 6y ago
With the party trick version of this (eating lemons or limes straight), I was always concerned about the acidity on my teeth. Is that not a concern here because you just don't use as much acid?
113.
by gradys 6y ago
I guess the question here is who is more like those doctors in this scenario: you, or the people running the randomized controlled trials?
114.
by gradys 6y ago
I made a tool for visualizing text datasets in 2D or 3D: https://nebulate.ai It runs a machine learning model in your browser to convert the text into points in a high dimensional space, and then it projects those points down to
115.
by gradys 6y ago
A Twitter account with this vibe: https://twitter.com/dmitricherniak Would love to hear about others.
116.
by gradys 6y ago
I'm a machine learning engineer with a focus on data efficient NLP. I want to build tools that make it dramatically easier to apply ML (especially NLP) to real world problems like: - Routing customer support requests - Understanding fr
117.
by gradys 6y ago
If you mean 2008, yeah, you're right, but those were mostly loans to relatively low income people buying homes, not loans against billionaires' unrealized capital gains.
118.
by gradys 6y ago
Why does taking private loans against his private assets count as socializing risk? The banks he gets these loans from understand the risk and do not represent "society".
119.
by gradys 6y ago
Thank you! We have very similar interests! Especially the first one and the interactive sound recognition one. Any other work in IML you'd recommend?
120.
by gradys 6y ago
I had a similar reaction, and also signed up. Upon reflection, I think there actually is a chance for it to be the best. Global poverty continues to shrink. Depending on your politics, you might be excited about the next US administration.

More ›