blackbear_

15 ms·

91.
by blackbear_ 1y ago
Then it seems that blocking ads is the more honest thing to do! Otherwise the company placing the ad would be unfairly paying money for a service not actually delivered. This also makes the market more efficient, as blocking ads is a clear
92.
by blackbear_ 1y ago
Note that the token embeddings are also trained, therefore their values do give some hints on how a model is organizing information. They used token embeddings directly and not intermediate representations because the latter depend on the
93.
by blackbear_ 1y ago
I've heard compactness described as one such concept. https://math.stackexchange.com/questions/485822/why-is-compa...
94.
by blackbear_ 1y ago
You are right, that works if you assume that every number can be represented as a decimal string. That is indeed true for real numbers, but not for hyper-reals ( https://en.m.wikipedia.org/wiki/Hyperreal_number ), which
95.
by blackbear_ 1y ago
The fact that there are no numbers in between is not obvious at all, and has to be proven formally! In fact, there is a (rational) number between any two distinct real numbers, therefore your proof attempt only works if you assume that 0.99
96.
by blackbear_ 1y ago
It definitely works in the short term, but what matters is the long term effects: how will people and businesses adapt, and will that lead to a better society? Case in point: At least as far back as the 17th century, many people believed th
97.
by blackbear_ 1y ago
How do you test "real creativity" and "critical thinking" in a way that is both scalable and reliably tells apart those who get it and those who don't?
98.
by blackbear_ 1y ago
> Let’s define spam ads: they’re the ads you didn’t ask for and don’t care about, but they still show up in your feed. Every ad is spam. I do not want ads, never, anywhere. Content has no influence on this simple fact.
99.
by blackbear_ 1y ago
This idea was already picked up and elaborated in the 50s and 60s by the Frankfurt school: > Consumption of the easy pleasures of popular culture, made available by the mass communications media, renders people docile and content, no mat
100.
by blackbear_ 1y ago
> Error establishing a database connection Oh, the irony.
101.
by blackbear_ 1y ago
Yes, you need to flash their image. They have a web installer from the browser and the process is relatively straightforward: https://grapheneos.org/install/web
102.
by blackbear_ 1y ago
> Is it still possible to initialize an Android phone without a Google account? Totally, get a Pixel phone and put GrapheneOS on it. You get state of the art hardware and the latest Android hardened for privacy and with optional Google s
103.
by blackbear_ 1y ago
On the contrary, reporting changes relative to the standard deviation of a control group frees you from scales and their meanings, because it relates the observed change to the normal spread of scores before the intervention. In this way, y
104.
by blackbear_ 1y ago
It does indeed have two peaks: https://en.m.wikipedia.org/wiki/Inverse_distribution#Recipro... As the mean of this distribution does not exist (it's "infinite") the Monte Carlo estimates aren't reli
105.
by blackbear_ 1y ago
If X is normal and centered around 0, then the average of 1/X does not exist (math speak for "is infinity" in this case). In these cases Monte Carlo simulations are not reliable because they give high variance estimates (math
106.
by blackbear_ 2y ago
The ease of cheating/creating fakes surely influences how much cheating/fakes are in circulation, and while we can tolerate a little, excessive amounts will be disruptive. So many technologies moved from obscure curiosities to mas
107.
by blackbear_ 2y ago
Productivity at work is well correlated with enjoyment of work, so the industry better look for people who enjoy programming. The realization that productive workers aren't just replaceable cogs in the machine is also a bitter lesson f
108.
by blackbear_ 2y ago
I can't say for "most" apps but I've never had any issue with banking nor digital identity apps.
109.
by blackbear_ 2y ago
> due to everyone wanting to export to the US to obtain Dollars, the US has a truly unique negotiating position in a trade war Seems like the plan relies on the key assumption that this will not change. But why is this the case?
110.
by blackbear_ 2y ago
> in traditional statistics, you can understand the parameters. I also think that this is the key differentiator between ML and stats. Statistical models can be understood formally , which means that not only we know how each parameter
111.
by blackbear_ 2y ago
Hypocrisy is a feature, it makes clear than they are above and you are below.
112.
by blackbear_ 2y ago
Not sure which other topics you mean, but "1000 exercises in probability" should keep you busy for a while (one can find the PDF online). For other math oriented riddles, check out "The colossal book of short puzzles and prob
113.
by blackbear_ 2y ago
> Musk says that Grok-3 is a “maximally truth-seeking AI — even if that truth is sometimes at odds with what is politically correct. https://www.theverge.com/news/614218/elon-musk-xai-big-brain... https:/
114.
by blackbear_ 2y ago
Sounds like you would enjoy being involved in developing the technical strategy of a company or department, which basically means staying up to date with new trends in technology and planning their adoption in house.
115.
by blackbear_ 2y ago
I think that 14 years old is too young to specialize. Your free time, which for now is plenty but will become less and less, would be better spent exploring and learning ad many things as possible according to your personal interests rather
116.
by blackbear_ 2y ago
That's indeed an interesting issue. Perhaps one could argue that PII data stored in the model weights is equivalent to encrypted data, where the prompt plays the role of the decryption key (IANAL)
117.
by blackbear_ 2y ago
Sure initialization helps, but are there also results about long term training dynamics? Even the paper you suggested had to use some sort of normalization to keep things stable
118.
by blackbear_ 2y ago
And so vanishing gradients are not a thing anymore?
119.
by blackbear_ 2y ago
No mystery there IMO. It works well because linear projections are the basic building block of modern neural networks.
120.
by blackbear_ 2y ago
Sounds like federated learning [1]. There is lots of research on that, and therefore a lot of interest too. This is not my field so I will just point to some surveys [2,3] to get you started. [1]: https://arxiv.org/abs/

More ›