pavanky

Isnt this assuming coal / gas powerplants ?

Wouldnt solar / Wind have a smaller CO2 foot print and hydro electric be more efficient ?

Just a heads up

1) OOP concepts don't mean everything.

2) Forcing Whiteboard interviews is bad. Some people prefer writing code (on computers) to think. Some people prefer whiteboards.

Please dont write yourself off because of these things.

> I also suppose that I'm an ok "hacker" in that I can get very interested / fixated on certain problems, although my solution is more likely to be a complete mess.

This is a good skill set to have and build on. If you care about code quality or improving your skills, try to find people who can mentor you in those directions.

Thats not what they said. If it was the average, the sum of all plans (42 @ 0.81) would be 34.02 which is less than the most expensive plan at 108.

If anyone is else is interested in thought experiments on how other forms of intelligence / sapience may evolve, try to read "Children of Time" and "Children of Ruin" by Adrian Tchaikovsky.

Also if anyone else has other books that follow similar themes, please recommend!

I think it has a harder time if more than one object is present in the image no matter what you draw.

Even if the other stuff you draw is just a random line segment a bit close to the actual picture.

> and immediately messed it up by mixing ordered maps with arrays into a big soup, but hey, PHP

And here I thought Lua was the only language insane enough to do something like that.

Just realized Lua tables aren't ordered, but it still mixes arrays and maps / tables into a single data structure.

The weights arent averaged unless the training step is synchronous. Even then most of the times it is the gradients that are added up rather than the actual weights.

For inference, I dont think there are many papers that claim direct average of weights perform better than any single model. It is usually the output that is accumulated in some way.

The post no longer exists?

Isnt the GIL part of the implementation rather than the language?

There is no single Hindu philosophy. This feels like a mix and match of multiple different versions.

Yes, more than one infact.

We dont farm and eat as many carnivores. Octopus and some fish are the only exceptions. The non ethical issues raised do not generalize easily.

Did you intentionally remove the first part of the paragraph ? They are talking about the environmental impact of octopus farming which creates artificial strain on the environment.

> The amount of feed needed to sustain and grow an octpus is three times the weight of the animal itself and, given that octopuses are carnivorous and live on fish oils and protein, rearing them risks putting further pressure on an already over-exploited marine ecoystem.

The calories spent on raising and consuming animals higher in the food chain is exponentially more than consuming those lower in the food chain. This is the point they are trying to make. It wasnt about the ethics of eating Octopus itself, but the means in which they are being raised / procured.

I dont care who or what is dictating them, as long as they are consistent. If someone is making the claim that it is OK to eat anything as long as they are delicious, I want to know what their stand on cannibalism is.

It isnt a competition. The point I was trying to make is that India not only has highly fertile land, but also a lot of it to theoretically feed its people without relying on others.

As you said, US has only a 3rd of the population. This is probably the reason they are the largest exporters of food in the world.

FYI India has more arable land than USA at the moment: https://data.worldbank.org/indicator/AG.LND.ARBL.HA?end=2016...

Alternatively, we need to see how well these algorithms can fake videos from different angles in a consistent manner. If we have enough enough people recording, it may just be enough to prove something is not a fake.

I traveled to India in 2017 and 2018. I had problems getting HSPA+ most of the time. Had no problems with local carriers.

You dont go from beginner to proficient in any language in 1 year, much less C++.

Not everything in the compute pipeline is going to be converted to fp16 operations. Anytime you are doing accumulation or exponentials, you would have to have it in fp32.

There was a good talk from NVIDIA at last years GTC: http://on-demand.gputechconf.com/gtc/2018/presentation/s8923...

Here is another relevant blog post: https://devblogs.nvidia.com/mixed-precision-training-deep-ne...

EDIT: Also not everything in the training loop is a matrix multiplication where tensor cores are useful.

I have friends who keep switching between three (not counting the made up language only those two understand).

This isn't uncommon among multi lingual people.

Knowing Apple, I wouldn't be surprised if they come after the ctrl key next..

It depends on the type of vector operation(s) you are doing and the machine you are on. For one off vector operations it is never worth it to make a transfer to the gpu.

If you are going to have a lot of temporary vectors as part of a larger algorithm, it is usually beneficial to copy the inputs once, do all the computations on the gpu, and copy them back.

> Possibly attractive for people who just like deep C++ platform building, which is an internal drive not often found in people wanting to solve business problems with ML models.

And the world has plenty of people who are not interested in trying to solve business problems with ML models, but are rather interested in the engineering side of ML. I am one of those people. My current work (at cortex) involves improving the time taken to train the types of models not usually described in research papers.

If anyone is interested in machine learning but wants to build high performance training / serving platform, message me over twitter (@pavan_ky), Cortex is hiring.

Before anyone flips out, he actually merged the code: https://lkml.org/lkml/2018/6/5/774

    Side note: I've merged it, and it's going through my build tests, 
    so it's really not that I hate the code.

    But I really find that kind of one-sided rationale that ignores reality unacceptable.

    And I find it dangerous, because it *sounds* so "obviously correct" to people who don't know any better. 
    If you don't know that gcc explicitly says that you should use unions to do type punning to avoid aliasing issues, 
    you might believe that union type punning is a bad thing from that commit message.

    So it's dangerously misleading, because lots of people have a dangerous reverence for paper over reality.

    In programming, "Appeal to Standards" should be considered a potential logical fallacy. 
    Standards have their place, but they definitely have their caveats too.

I was happy paying a subscription fee to whatsapp when it was still a thing. I dont know if facebook has ever made money off of ads shown to me.

Same goes with google and the subscription I pay for google music.

I mean it wouldnt be that big of a problem if they adopted Vulkan, but they are pushing Metal :-/

Why aren't they using LLVM :-/

not sure what the battery life on x230 is. I usually get 6-7 hours atleast on my current chromebook.

Besides this was an impulse purchase. I was tired of my home laptop and wanted something so I went to best buy and purchased it within half an hour. I can't say the same for x230.

I needed a machine to ssh from and have a browser. $250 acer chromebook with crouton mostly works.

This is not my primary machine however. I only use it at home to ssh into my desktop once in a while.