Regarding 2x faster than pytorch being a condition for tinygrad to come out of alpha:
Can they/someone else give more details as to what workloads pytorch is more than 2x slower than the hardware provides? Most of the papers use standard components and I assume pytorch is already pretty performant at implementing them at 50+% of extractable performance from typical GPUs.
If they mean more esoteric stuff that requires writing custom kernels to get good performance out of the chips, then that's a different issue.
The expression f(z) = \sum_i 1/(z-\lambda_i) is called Stieltjes transform and is heavily used in random matrix theory and similar expressions are used in other works such as Batson, Spielman and Srivastava. This is all to analyze the behavior of eigenvalues which is exactly what they were trying to understand. I'd be very surprised if Aaronson doesn't know about this.
I think it is not just the library but the huge costs associated with storage, encoding and bandwidth. YouTube has innovated significantly to make it as cheap as possible to run such a service and it is likely that it would take an enormous amount of money for any competitor to replicate it.
(Disclaimer: I work at Google but no connection to YouTube)
Netflix is definitely trying its hardest to get into the Indian market and I do think they are gaining some ground recently. They have picked up a lot of new telugu movies and has basically become essential if you want to stream new movies. They also have mobile only plans at 199Rs but not sure how many such subscriptions they sell but for me at least the 649Rs plan is not too bad even if my parents watch at most 1-2 movies a month on Netflix.
If they are able to get even 2 million telugu households over to subscribe the next few years, that would be a revenue of about $150 million which I think is probably much less than the amount of money they spend on telugu content.
I didn't want to associate this account w/ my real name but now that you mentioned it wasn't right of me to not point that out. I added a disclaimer.
The posted algorithm and the one mentioned in my paper are very similar. It is just that the cumulative sum computation is parallelized in the posted website.
Exactly lol. Not sure why everyone is taking the contribution of lean in finding the error so seriously.
I would be more impressed when some mathematician finds a more severe error in their proofs with the help of theorem provers (meaning a mistake in their own intuition).
I recently visited France. Before flying from the US, I purchased an Orange e-sim online and activated it while at the airport in the US. As soon as I landed in Paris, I was connected to Orange network and was able to use internet instantly.
God! I hope both WhiteHat Jr. and their parent company Byju's crash and burn. They are crap companies which make poor and gullible parents enroll their kids in one of their programs for extortionate amount of money. If the parents cannot afford to pay all at once, they give a "loan" to the parents and collect a part of it each month even if the kids aren't satisfied with the education that is being provided. Note that none of the kids in India use these apps as the only way to study. They attend regular schools and use these apps as an additional source. Byju's employs 1000s of people whose only job is to go visit as many houses as they can and trick the parents into subscribing for their services by falsely claiming that the only way their kids can succeed is when their education is supplemented by these apps. Many innocent people who want to do all they can to make their kids' lives better fall into the trap and then keep paying the "loan" amount back which could be a substantial part of their monthly income.
Meta and Disney invested a lot of money into these companies and they use it to bully anyone who tries to bring these issues up by shutting down their YouTube channels and filing defamation law suits [1,2]. The entire online education industry is unregulated and resort to shitty antics to extort money. I don't understand why these companies need to exist when we have other free and amazing resources online like Khan Academy which teach all topics to all ages of people.
FAQs can be useful if they are used as a questionnaire to test our own understanding of the content. But FAQs being the only source of information is pretty terrible and very lazy. Lately, on many Indian government websites, I’ve seen that they straight up just have a list of FAQs and nothing else which infuriates me a lot.
As an Indian who recently moved to the US, this list captures a lot of things that surprised me. One of the biggest things I was shocked by is the number of choices in US supermarkets - 100s of varieties of chips, 4-5 different sizes of eggs along with a choice of white/brown/cage-free/organic, 4 types of milk (skim/1%/2%/whole), 10s of types of bread etc. Back in India, we have considerably fewer choices in each of the ones I mentioned even in the bigger supermarkets.
"we will understand that Apple is not willing to make the changes necessary to allow us to provide Android customers with the option of choosing their app store and payment processing system."
Did they intend to write iOS there? That's a pretty big typo in such an important email.
Yes. iPhone XR already had great battery life and 11 Pro and 11 Pro Max have more than 4-5 hours additional battery life compared to Xs and Xs Max. My 11 Pro can easily last 1.5 days with heavy usage.
Any two matrices AB and BA have same eigen values. Say v is an eigen vector of AB with eigen value lambda : you have ABv = lambda v. Now BA(Bv) = B(ABv) = B(lambda v) = lambda*(Bv). So, Bv is an eigen vector of BA with eigen value lambda.
Can they/someone else give more details as to what workloads pytorch is more than 2x slower than the hardware provides? Most of the papers use standard components and I assume pytorch is already pretty performant at implementing them at 50+% of extractable performance from typical GPUs.
If they mean more esoteric stuff that requires writing custom kernels to get good performance out of the chips, then that's a different issue.