I recently ended up in an interesting conversation with a university faculty member where we found that many of the most endowed universities could essentially make undergraduate education free and make back the difference (and often more) on straight interest accumulated from leaving their money in a bank.
Given that they are probably making more than the current interest rates by investing, I wonder why students still pay so much for tuition at these universities in particular.
Because they can charge that much, and people are willing to pay the amounts they pay. Wouldn't be very exclusive now would it, unless you're charging $100k a year?
>>You can save time by learning from the mistakes of others.
>Only barely. For important life decisions/paths, people only learn from their own mistakes (and often not even from those).
>Heck, Computer Science as a field itself forgets its own collected wisdom every new generation hits the market, and re-invents BS that other eras have tried and buried with another fad name.
Your last statement sounds like it's in favor of "learning from the mistakes of others" in that we could avoid re-inventing failed "BS" if we learned from others' mistakes. Are you saying that the idea that folks in CS re-invent work which has previously failed is evidence that it is not possible to learn from the mistakes of others?
>Are you saying that the idea that folks in CS re-invent work which has previously failed is evidence that it is not possible to learn from the mistakes of others?
No, I'm not dealing in absolutes ("not possible").
I'm merely saying that the reality that "folks in CS re-invent work which has previously failed" is supportive argument in how people in general don't learn from other's mistakes.
Of course it IS possible. It's just that it's rare.
There is also a fantastic video where Cliff Stoll describes the Friden EC-132 calculator and shows off its acoustic delay line here: https://youtu.be/2BIx2x-Q2fE
I remember being in high school when Google Plus came out and thinking it would be incredibly popular. I totally missed the mark, and it's funny because while I thought it would take off with other people, I never really used it myself.
I touched on this before in my rambling. Google botched it, not the product itself as the initial offering of G+ was actually really really good. Far better than Facebook, a return to basics in a clean interface with Circles. Groups being a feature Facebook later added.
When Google launched G+ it was actually one of two new products the primary one being Google Accounts. Before that point you needed to use a Gmail account to sign up on any of Google's services and it made everything a big pain when you had more than one Gmail.
Now with a single Google Account you could sign in to Google's entire ecosystem, including all of your Gmail accounts at the same time. It also came with a free, ad-free, totally simple, you-don't-have-to-use-it social network called G+.
Google should have marketed this all honestly. Instead they started talking up G+, and hid Google Accounts underneath it. This was a huge mistake. The general public felt like they were being tricked, because they were, albeit far less nefarious reasons than was assumed. Nobody wanted to give their name, to sign up to G+, and therefore rejected both G+ and the Google Accounts system.
This compounded because Google really needed everyone to switch to the new account system. Otherwise they were doomed to continue supporting both Gmail authentication and Google Accounts authentication forever. So Google began pushing for people to create Google Accounts, as you would imagine. People then saw this as Google trying to force them to give their full name, trust was already gone.
Amazing to me people trusted Facebook over Google because at the time Google so far had a very good record protecting private information. Now that's all gone of course but at the time they had an incredible amount of good will which was seemingly not reciprocated and now Google is a slimy gross privacy succubus the same as Facebook is. Probably they just threw their hands up and said 'fuck it'.
Google tried to further reduce complexity in their myriad of different platforms. They now had two social networks G+ and Youtube. G+ was going to eventually absorb and replace Youtube, it would have been pretty elegant really. The first step was to have the Google Accounts system absorb Youtube accounts. People resisted this again because at this point good will was gone. Eventually Google gave up on everything.
Now G+ is a cesspool of tacked on features, advertising, just a bad smell entirely. It did have a lively base for a short while. Things really went downhill after the Youtube consolidation because G+ suddenly had what was notoriously one of the worst communities on the internet, Youtube comments.
Google has I believe still not successfully moved everyone over to Google Accounts. Youtube was never successfully merged into G+. G+ is pretty much dead at this point. Nothing went well at all and now Google has to support a million different things that all barely work.
Meanwhile Youtube hasn't seen a facelift in about a decade while Google figures out what it wants to do with it.
I had a really hard time getting into Vim until I found out everyone I knew who was using it had mapped something else to escape. Once I did the same, it was much easier to get started.
For me it was the same. Using ESC is really ergonomically awkward for me. There are also people using Ctrl-C (which works by default) - but I prefer the 'jj' or similar method.
It's also useful to use Bash in vim-mode and also remap the esc-key, e.g.
set editing-mode vi
set keymap vi
$if mode=vi
set keymap vi-insert
"jj": vi-movement-mode
$endif
Something like this - you can then search Bash history via '/', etc.
I'm assuming by iterate you mean "generate outputs of a GAN which preserve some order that you can iterate through."
The answer is yes, but not in the way that one might think. Despite the fact that we seed the GAN with input noise, there is no guarantee that the GAN makes use of this at all. This is a theme with GANs: we often want to imbue them with prior knowledge that we think is important, but is easily ignored by the GAN. In this case, we want to generate samples from p(x|z), where x is in the space of our data (often images, in this case passwords), but provided it gets good results according the the loss function, your GAN may learn p(x|z)=p(x). This is fine if you don't care about enforcing some relationship between input and generated samples, but here we do.
One solution is to use InfoGAN (https://arxiv.org/abs/1606.03657), which adds a term to the loss function that the mutual information between a latent code and the generator output must be high. Your latent code might be drawn from a uniform distribution on [-1,1], and the generator output will be conditioned on this code. This being continuous, it's questionable what "iterate" might mean. On a computer, maybe you iterate through every possible float (as someone mentioned), but if you want to generate N different samples, you could also discretize this distribution to N values on the given interval, each with probability 1/N and sample from this PMF.
I haven't seen it posted yet here, but for anyone unfamiliar with mathematics in juggling (or anyone with an hour to kill watching something entertaining), there is a lecture/performance by Allen Knutson (https://www.youtube.com/watch?v=38rf9FLhl-8) that I found to be a great intro (before I even knew I was looking for one!) and got me excited about the subject.
All I could find on these archives were summaries of the letters, is there anywhere where we can view the actual documents that are archived (either scanned or OCR)? I admit, I had a hard time navigating that page, so I may have missed something obvious.
Edit: My bad, after reading the Conditions Governing Use I found this: "Photocopies and photographic copies of material in the archive can be supplied for private study purposes only, depending on the condition of the documents." So it looks like these written summaries are all we can get?
It's great seeing that more and more data is being collected about this all the time. I'm a huge proponent of this tech.
What I wonder when I see these statistics, though, is whether all miles are really equal? For example, are Tesla drivers more comfortable using Autopilot in "easy" driving situations? Is there really a one-to-one correspondence in the distribution of the kinds of miles driven with Autopilot on vs. normal cars?
Furthermore, the metric commonly cited is "fatalities ever N miles." Are there fewer fatalities because Autopilot is great, or because Teslas are safer in general? Has there been a comparison between fatalities with/without Autopilot strictly on Teslas? Even then, it seems to me we are subject to this potentially biased notion of "miles" I mentioned previously. The Wikipedia article you mentioned cites a 50% reduction in accidents with Autopilot, but the citation is to an Elon Musk interview. I haven't yet seen anything official to back this up, but if anyone has any information on this, I'd love to see it!
Isn't that easily countered by comparing Tesla Model S' overall rate of accidents vs another similar vehicle, with similar safety rating, including all self-driven and human-driven miles? There should be a proportional reduction.
Yeah, I think so! That's exactly why I mentioned the accident rate reduction cited in the Wikipedia article shared above.
I'd love to see official work that explores that angle (rather than a claim from an interview, which is what the Wikipedia article refers to), I just haven't seen any document/study about it yet.
