1,352 karma · joined March 25, 2019
It looks fantastic for the compute budget
Communism is just not compatible with freedom. When free to choose, societies do not become communist, they embrace production and trade. You cannot achieve what communists want without extensive use of force
To whom is that fine? To the people going from $2/hour to zero it certainly isn't.
I honestly cannot grasp how some people can _forcibly_ remove other people's _options_ then pat themselves on the back as some sort of armchair savior.
Yes, for some people a bad option could still be their best option. What sort of moral superpower is that that enables you to forbid a contract between two consenting adults A and B such that A wants to work for X and B that's willing to pay X, but you as a C that just won't let it happen and is willing to use force to stop it from happening
To me the answer is no.
Now feed it Doom haha
Jokes aside this is pytorch so this is compiled to C++ or cuda, the problem likely comes from the different functions that are called for += vs +
If on the other hand it's printing money (as I'd expect from the trillions of free money in recent months) it should only get worse. What I observe is more consistent with money printing than disruption of supply chains
I'm sure our demigods will gladly enlighten us with their wisdom to distinguish between misthoughts and correct thoughts.
So that when journalists claim to be following the science there will be no dissenting voice in research to disagree.
Those who defend opposing ideas will have no leg to stand on and will rightfully be labeled science deniers
Sometimes they're buying decades worth of debugging/testing
Would other companies that support pornhub/across in any way also be considered participants? Water supply/electricity/ internet providers/whoever sells them dildos/food
Categorical > one hot encoding > deal with new categories in test time (sklearn does this, but it's really slow and clunky)
Numerical > either figure it out the data distribution for each column and normalize by that or normalize everything by z score. Found an outlier?? Oops, every feature collapsed to 0
Can you that for 10 features? Sure, now try it again with 500, it's not fun
Ok, now that you've done all that you can begin training and possibly get some reasonable result.
Compare that with tree models: data>model>results
Wake up, decide it's a good day to be exploited.
Kiss your wife goodbye as you enter your car "bye honey! On my way to being exploited by Uber!"
Drive away and open your exploitation app. Accept a ride, go to the person: "who's the person that's going to participate in my exploitation together with Uber today? Oh, there he is!". Drive him to the place he needs to be, drive away.
"Ok, do I want to be exploited some more today?"
Decide to be exploited on a few more rides and return home.
Now being serious, if you're an Uber driver and let's say that in the short term that's your best option, why would it be better for him to have this option striped from him? Why do you think anyone else should make that call except him?
Blaming bad thermal design on "use case" does not make sense to me.
If you'd like to run some intensive numerical programming on your laptop just because, I would find it outrageous for it to shutdown due to thermals. Can other hardware designers use that excuse too? "Oh your software use case is not covered by the cpu voltage regulators", "the power supply was not meant to be on all the time", "the power cable should not be connected for that long", "you should not use 90% of your ram for 4 hours straight, it was not designed for that use case"
"Oh the average end user only uses the car 1h/day. This 5h long trip was not the expected use case, that's why it overheated"
If only they did with their lives exactly as I told them to