prions

" Not coincidentally, twins are at strong risk of being objectified. First, we get reduced to our appearances: what matters most about us is how our bodies hit the senses. Second, we get treated as fungible: interchangeable with our fellow objects. Third, in many cases, our subjectivity is ignored. People don’t deny, exactly, that twins have rich and distinctive inner lives, just as sexists don’t deny, exactly, that hot young women do. But those lives often aren’t what’s interesting about us, and they’re easily glossed over. "

Some people cant work and interact with others in a respectful manner. Just because you're in an open office doesn't mean that others DON'T have the right to a low-disruption work environment. It gets really tiring to have to constantly fight off rude people just so I can get a few minutes of uninterrupted focus.

"It doesn't bother me, therefore it doesn't bother anyone" is such a common trope.

"Many of these companies are willing to spend big dollars renting high-end Manhattan real estate, rather than the older office stock in Queens that Amazon was prepared to lease."

You're right about embedded though. Most telecom hardware is a race to the bottom and there were a lot of layoffs at Nokia when I was there.

These articles panning formal education don't fully grasp how difficult it is to get your foot in the door as a non-cs person whose also working full time.

I also wanted to do something more technical than just web dev, which most career changers seem to land up in. Being a grad student opened a lot of doors into more technical roles in companies that are _only_ reserved for grad students.

Grad school is expensive, but I went in with a solid plan. I mitigated some of this by going to an in-state school with a good CS program and commuting to class. I was very focused in my course choices and was a 4.0 student, which unlocked even more opportunities.

I was able to land two good internships (one in embedded development and one in machine learning) and used that experience to springboard into a high paying role in NYC. I have no problems living alone in the city, paying my loans, and still saving a significant amount of money per paycheck.

This is completely anecdotal but what I'm arguing is that grad school can be a good idea if you have a solid plan of how to get the most out of your education. The people I see who haven't gotten their money's worth came into school and just drifted through while assuming that opportunities would come to them just by virtue of being a grad student.

Forget the lousy microphones and the dinky interstitial stock music — the thing that derails most podcasts is the blab. There are two kinds, more or less. The first is that soft, inquisitive staccato popularized by Ira Glass on “This American Life,” the source from which so much pod-voice appears to have sprung.

The ubiquitous NPR Voice/Sound is really what turns me off from podcasts/audiobooks/etc. While I love NPR, their sound is absolutely grating to listen to. Too HD, rife with extremely loud and detailed plosives and sibilants. I don't want to listen to the speaker suck the spit back down their throats in excruciating detail after every sentence.

NPR's sound quality is too detailed. It's like if you watched a video of someone talking and could see the aftermath of a popped zit smack in the middle of your screen.

If Apple made IEM-like AirPods, I'd definitely jump back in.

This was big news last week and gave a lot of legitimacy to the allegations.

  Huawei China allegedly continued to pester Huawei USA   employees to dig up the intel, and they   in turn continued to ask T-Mobile folks about Tappy's design. One of the Huawei engineers, known in the indictment only as "A.X.", took and sent unauthorized photos of the robot to Huawei China, but it wasn't enough, and they asked for more.

When you're designing a real-world engineering project, the entire specifications are defined legally (through national, state and local laws) and technically in manuals/books. Many engineering specifications will describe the work done to a T before you even need to think about it i.e. "water main shall be constructed of 12'' coated DIP at depth no less than 2 ft". A lot of the challenges are managerial and logistical.

All of this is on purpose. "Traditional" engineering disciplines are more mature and have the constraint of being safe for the general public. There isn't much room at all to creatively deviate from what's already specified.

I've found software design to be a lot more technically demanding in regards to designing and building things. There's a lot less precedent, more moving parts and many different ways to do one thing.

I also notice a lot of dismissive comments about "black box models" or the simple solutions of just parsing out whitespace. My two cents:

1. Models with hand crafted rules perform WORSE than learned representations, especially when you have an end-to-end model with pre trained embeddings. This is shown by one of the seminal papers on this model, Ma and Hovy (2016) https://arxiv.org/pdf/1603.01354.pdf.

"However, even systems that have utilized dis-tributed representations as inputs have used theseto augment, rather than replace, hand-crafted fea-tures (e.g. word spelling and capitalization pat-terns). Their performance drops rapidly when themodels solely depend on neural embeddings"

2. Human speech and human written text are messy. Having a rule for human speech will inevitably lead to a massive list of rules and exceptions to those rules.

3. This model is multi domain, meaning that you don't just need rules for one domain, but rules for multiple domains and interactions between those domains. Considering Amazon's hefty amount of data, it's much more efficient to learn these represntations though a machine learning model rather than constantly playing cat-and-mouse with keeping your hand crafted rules.

While I don't do academic research, I spend a non-trivial amount of time reading papers and keeping abreast of new research that is relevant to my job. If the research seems relevant enough, I'll dig deeper and implement the research into a workable model. The software engineering portion really kicks in when I productionize this model and make it a usable product.

So in regards to your question, if you want to do fundamental, academic research then go for your PhD. For the kind of work I do currently, then it's more about the group you work with in industry.

Frozen 2 is in both Top Stories and Business, but not in Entertainment. There's some objectivity at play here, but I'd rather see articles about movies only in Entertainment, with the ability to not see that topic.

>I don't care if the Chinese government has my data, I'm not a Chinese citizen

>The US spies as well

>The US did something wrong in the last 100 years, therefore you shouldn't complain that China did it last week

Can you clear up what you mean by a hidden seed? My understanding of GANs were that the generator learns the distribution of the input data without ever directly observing it.

https://en.wikipedia.org/wiki/Category:Military_Superfund_si...

https://www.sciencedirect.com/science/article/pii/S016770120...

Elmo embeddings are especially interesting. Dynamic Bernoulli embeddings also deserve a mention. https://arxiv.org/pdf/1703.08052.pdf

One of their latest updates added a really frustrating regression where emoji autocompletes are spellchecked on ios, correcting it nonsensical words. It took some digging to find their bug tracker and I noticed it's set to their lowest priority. Why was this allowed to be released in the first place?

As a dev, you can pick up some popular ML frameworks and learn the basics relatively quickly. The difficulty in this field comes from the amount of theoretical knowledge you need to interpret your results.

All these data science bootcamps/learn quick schemes are like teaching a blind person to drive a racecar. He can work the pedals and the steering wheel - but has no idea where he's going.

Generative models like GANs are fascinating, but very temperamental to train. Some of the findings in this paper mirror my own observations - increasing the complexity of the GAN adds a lot of instability. My solution was to keep things as simple as possible. I spent a lot of effort trying to increase the size of the network to get better results, but in the end my smallest implementation worked the best.

This bit is interesting: "Without progressive growing, all layers of the generator and discriminator are tasked with simultaneously finding succinct intermediate representations for both the large-scale variation and the small-scale detail. With progressive growing, however, the existing low-resolution layers are likely to have already converged early on, so the networks are only tasked with refining the representations by increasingly smaller-scale effects as new layers are introduce"