chmullig

I feel like this massively overstates the advantages of an AWS to satellite link. They aren’t going to put tiny regions in space, so it’s just about data transit. Nobody would suggest AWS should start a cell phone company or fiber ISP just for bundling that with AWS. IP, possibly with a VPN, is a very solved issue for AWS access. It’s how 100% of their customers use AWS now, after all.

Which projects do you think have particularly high quality and easy to work with data? I was using that but it’s so messy...

Something like sendgrid?

I haven’t fully explored their portal. However I find it really hard to square with other data, such as a 90% reduction in NYC subway data.

https://iquantny.tumblr.com/post/616488508041936896/nyc-subw...

This is a cool article, and I appreciate the actual testing.

I also heartily recommend this implementation using conceptnet embeddings.

https://github.com/commonsense/codenames

Grab the latest mini.h5 and go. https://github.com/commonsense/conceptnet-numberbatch/blob/m...

I’m pretty sure iPhones do put it in. You should exif dump an iPhone XS image. INSANE what’s in there.

The orgo night performance is one of the big events of the semester. Every student knows when and where it is, so nobody will be surprised or pranked.

As others have said, this is a big loss for what little school community there is, and is very in keeping with university administrator style.

Actually need to go PA to right near OAK to SFO within just a few hours in the next few days. If they had a 9:30pm OAK->SFO I would have probably booked it at that price. $215 compared to $100 with Lyft...

I've been enjoying some minor woodworking as a purely hobby. I agree, it's great fun, and I think there is some common idea along with building software.

Wondering what you've found to be a reasonable amount of income vs work for sale? I'm not really interested in making money, but covering costs and feeling satisfaction would be good.

It's strangely beautiful, especially for something educational. A slow moving version of the animations as a wallpaper would be cool.

From the source article: Employing a catalyst made of copper nanoparticles embedded in spikes of carbon, the team found that electricity applied at just 1.2 volts was sufficient to convert CO2 suspended in water into ethanol.

From the abstract: Herein we report a common element, nanostructured catalyst for the direct electrochemical conversion of CO2 to ethanol with high Faradaic efficiency (63 % at −1.2 V vs RHE) and high selectivity (84 %) that operates in water and at ambient temperature and pressure.

It's not that the cell is small. It's that it has a single respondent representing it, so his weight is increased enough to be the equivalent of 30 respondents.

Yup, that's literally the definition of VCs. Instead of a single LP having to choose one startup to invest their $10M into, they buy shares in the dozens of startups that the VC then invests in. The risk and reward is shared between all the LPs in that fund.

Gordon, Zettelmeyer, Bhargava, and Chapsky "found some evidence that PSA’s as controls would lead an advertiser to draw the incorrect conclusion about the true lift of an ad campaign."[1]

[1] You can't tell me not to cite! ;)

Well that's a mistake.

Quite true for google. Less (but still kinda) for Facebook IME.

Web conversations like this are terrible. Clicks are a very small part of what advertisers are buying, even if that's how it's measured. Just like sometimes when you're at a cafe you buy a cup of coffee, but actually what you want is the nice space to work and an internet connection for a couple hours.

(Some) advertisers are actually smart. They do extensive, scientific research around attribution and effectiveness. Even if you don't click on ads, you're influenced by them (yes, you, too). They can track that, and it's a large part of what they're paying for. Just like TV ads.

No, steam heat is still quite common in NYC. It's just those valves work approximately 0% of the time.

Interesting. They talk about plain old Metropolis Hastings, which is pretty questionable.

Anyone excited about this, I highly recommend checking out Stan; it's under active development, actually works with real problems, and is used in the real world. With NUTS and HMC they've really made good on their promises, and quite soon they'll have meaningful ADVI support. See this former discussion: https://news.ycombinator.com/item?id=10244771

This is an aside, but I've noticed a lot of the best Ars Technica content is coming out of their UK arm lately.

It's definitely done. A traditional method is that the longer you've been at a place, the more vacation you earn. Another technique I've seen is to have a divide, eg "VP and above get an extra 5 days per year."

Facebook, Google, etc commonly use the contractor technique, where anyone you want to work for you but not give benefits becomes an independent contractor, or works for a vendor. Eg, the cooks in the Facebook kitchens.

That link is to the version hosted on a MyCPU. There's a bar at the bottom advertising a backup: http://mycpu.thtec.org/www-mycpu-eu/index1.htm

Wasn't that sort of what medium was pushing originally?

I interned at Facebook a few years ago, so I'll speculate with an ounce of information.

There's a lot of openness, including around issues like this, at the company. There's weekly all hands meetings, where a developer could directly ask Mark about his opinion on a feature like this. The user experience, including creepiness, was almost certainly discussed. They may have misjudged in this case, but that's fallibility, not because it was ignored.

I don't know about entirely stuck, but I've definitely spent a long time in the weeds on questions, writing code on the whiteboard that was clearly sub-suboptimal. I've still gotten offers. It's all about managing the interview, communicating, asking for a hint or two. Not guaranteed, but definitely possible.

Not just may, they definitely will be making investments they'd rather not because of signaling. They obviously think that's a worthwhile trade for the upside on the next Dropbox.

I feel like I've read literally this exact subthread on HN twice before. Am I going insane?

I love the concept of this. Essentially it's just letting you enter the ad auction for yourself, with a monthly spending cap (I imagine it's a little trickier under the hood, but maybe not that much trickier).

I signed up yesterday after a friend mentioned it, and it's near. It's definitely a minority of ads, but it's still nicer to see cat pics than other shit.

oh neat!

In my experience, Stan performance is decent with many models, particularly if all relevant operations are vectorized and the data sets are "reasonable." However it's easy to accidentally walk off a cliff, and write a model that takes days to fit. Additionally the real time output is a little lackluster, so it's hard to know how you're doing until it finishes (I hear they're working on that for ShinyStan).

I haven't done any HMMs or CRFs with Stan, but don't see why you couldn't do them. Passing in the data likely requires some tricks with arrays and indexing, but it's totally doable. Probably unlikely that you'd beat standard, custom algorithms, but if your HMM was part of a larger model, it might make sense.