eachro

Another thing to keep in mind about hospitalizations is that many people have avoided going to hospitals during the pandemic for non-covid related visits for fear of contracting covid at the hospital and/or adding to hospitals caseload. It could also be the case that people are re-evaluating their risk profiles and no longer taking the same precautions with respect to hospital visits. I'm not sure what is really happening but wanted to emphasize how tricky it can be to interpret covid related metrics.

Medicare For All I'm assuming

15k in just tuition does not seem outrageous. Suppose a year is 30 weeks, and you spend 10 hours a week in classes (not even counting the other benefits that college/a college campus may offer), thats about 50 dollars/hr you're spending for lecture. For some classes/fields it's clearly a much better deal. But 40k+ in tuition? Yea agreed, that's pretty outrageous.

I don't know whether it would or wouldn't. I was just replying to the parent comment which said this:

> But what Yang doesn't do here or in his book is make a powerful case to fiscal conservative spelling out just how universal income will avoid demolishing the economy through inflation. Until this happens, UBI is going nowhere.

I wonder if there are any economic studies on whether social security increased inflation when it was first enacted. It's not the ideal comparison because of the sheer scale of UBI vs social security but it would be something to get notion of the impact.

Can anyone point to a place where knowledge graphs are used effectively in industry/research/personal hobby?

Does the added regulation, oversight help prevent the problems that led to the 08 crisis? Also curious to know what/how these learnings manifest themselves outside of goverment imposed regulations.

Ray Tracing in a Weekend(https://raytracing.github.io/books/RayTracingInOneWeekend.ht...) is an incredible guide to work through to write a basic functioning ray tracer.

It's also pretty common for people to create new accounts to post things they wouldn't otherwise post on their mains, isn't it?

Accessible to whom? What makes tensorflow/keras more accessible than pytorch?

I don't think it's necessarily a matter of "interesting" but whether it's an ideal way to showcase these guests (which I surmise is what outlace is trying to get at). The guests he's had on his podcasts is astounding, no doubt about that. But suppose you knew nothing about Susskind/Kahneman/Knuth/Stroustrup/etc's prior work and listened to these podcasts, could you have deduced (given the conversation/questions asked) the mountain of work they'd produced to be worthy of their accolades in the first place? I don't think I could've and in some way that's a letdown and a bit of a missed opportunity. In the end it's Lex's podcast and he should have w/e conversations he wants with his guests. That's alright too.

What tech companies are based out of the LA area?

I think it depends on what you want to focus on. If you want to do deep learning, fast.ai is probably the best resource available. Jeremy Howard and Rachel Thomas (the two founders) have poured quite a lot into fostering a positive, supportive community around fast.ai which really does add quite a lot of value.

If you want to really understand the fundamentals of machine learning (deep learning is just one subset of ML!), there is no substitute for picking up one of the classic texts like: Elements of Statistical Learning (https://web.stanford.edu/~hastie/ElemStatLearn/), Machine Learning: A Probabalistic Approach (https://www.cs.ubc.ca/~murphyk/MLbook/) and going through it slowly.

I'd recommend a two pronged approach: dig into fast.ai while reading a chapter a week (or at w/e pace matches your schedule) of w/e ML textbook you end up choosing. Despite all of the hype of deep learning, you really can do some pretty sweet things (ex: classify images/text) with neural nets within a day or two of getting started. Machine learning is a broad field, and you'll find that you will never know as much as you think you should, and that's okay. The most important thing is to stick to a schedule and be consistent with your learning. Good luck on this journey :)

I'm not entirely sure how helpful it would be to see these algorithms since many of them rely on machine learning models underneath the hood (IE: a model that predicts the likelihood a user will click a given video). Is it useful for the average person to know that some gradient boosted tree/deep learning model spat out a probability estimate? More information certainly does not hurt, but I doubt the average citizen/gov regulator can do much with this knowledge. These are not the kind of static bfs/dfs/quicksort/etc algorithms we learned in undergrad that can be dissected so easily. The ML models that underlie the recommendations/ranking "algorithms" are constantly changing based on the data they're trained on. Does this constitute an algorithm change?

Disclaimer: this is just my understanding of how these newsfeed type algorithms work based on conversations with friends who work on these teams at FB/GOOG. Please do correct me if I got something wrong here.

Just out of curiousity, can you say a bit about how it influences your every day thinking? And why is measure theory so essential to really understanding probability? I sort of understand why it's necessary to have the language of measure theory and be able to talk about the measures/probabilities of uncountable sets but don't really understand beyond that. I have the equivalent of an undergrads understanding of measure theory after numerous gos at it but I haven't ever been able to piece it all together to have a cohesive understanding of the area the way I do for say linear algebra.

In the first problem they told you what numbers to put in position 1 and 2. In the second problem you want to find the specific numbers for position 1 and 2 that yield the desired output.

I'll second this. The style is clunky and reads as though the author were trying too hard to make every sentence entertaining which mostly detracts from the work. I know Stephen Merity is a serious researcher and the content here is legit given his body of work. But the style /prose in this preprint reminded me a lot of the some of garbage Siraj Raval peddled. Again, to reiterate, I am not commenting on the substance, only the style.

I'm super curious what the ML looks like behind the scenes. Do y'all think it's just a GAN thats been fed tons of logos that already exist? That seems unlikely–I can't imagine there are nearly enough logos for that.

Learn linear algebra properly. Like many folks on HN, I really enjoyed Axler's Linear Algebra Done Right, but ymmv. I got a ton out of going through each of the proofs + practice problems and really taking the time to work through the solutions. It wasn't enough for me to just read the text.

There's a linear algebra lurking everywhere in the realm of applied math (some people like to joke that machine learning is really just linear algebra) so it really is worth your time to have a firm understanding of it.

Suppose Facebook or Twitter were incorporated as a non-US company. Would the US gov still be able to regulate them?

36k miles on a 12 year old car seems super low. Seems like a great deal! Did you have to get it fixed up? I'm wondering why it only costed $2200.

Regardless of gender, the name should be in the headline.

How is AMD doing in the GPU for deep learning business? Have they completely ceded that market to Nvidia?

The $$ they can get from anywhere else in industry given how hot research level ML experience is (and how scarce this talent is). I imagine the appeal of FAIR is much more about the academic freedom. But sure the $$ doesnt hurt.

What solvers is he using? Is it really just some sort of simulated annealing? I would have expected A* search at a first glance.

I remember buying some of these inidivudal sour patches candies at a local deli for 10 cents a pop about 10 years ago. I wonder what the price is now. Probably a quarter I'd guess.

I agree that 2500x48hrs is probably a reasonably cost to pay for these kind of sweet results. But it is a bit prohibitively expensive for an ML hobbyist to try to replicate in their own free time. I wonder if there is some way to do this w/o all the expensive compute. Pre-trained models is one step towards this, but so much of the learning(for the hobbyist) comes from struggling to get your RL model off the ground in the first place.

Yea, intro is a bit misleading. Maybe the first 6 chapters(~130) pages is enough for a solid intro.

Lapham's Quaterly is pretty great https://www.laphamsquarterly.org/

Eh, but if you've taken a machine learning course, you should have seen the notion of false positive/false negative when you cover any kind of classification technique.