skwb

When I first saw this going around before seeing it on here, my first thought that it's still a bad take because how else are you form your inference without data collection? You're not just going to be given your conditional probability; you're almost certainly going to have to go out to collect the data to construct this (and the data collection being the regulatory item covered).

I think that the hot take machines that are twitter/online communities have a certain amount of mindrot where poorly researched topics make headline news concerns me (which tbf is in itself a hot take I suppose).

Bioinformatics PhD student here: any plans to move to histology field? Or are you focusing more on the R&D space for now?

The public health and medical professional communities.

Not when they're in bad faith.

Has there been any major updates to using TF for AMD GPUs recently? Or is it still too much of a headache to do it?

I know you're being snarky but to be fair there is the Wisconsin Nation Air Guard. :)

There's a world of difference between "occasional incidental radar profiling of F-35 from the OpFor perspective" and "Consistent friendly profiling of F-35 as they land at Incirlik air base".

I tend to agree with this sentiment.

Lets take the Turkey case for example. For all intents and purposes, they are a close NATO ally since the Cold War. They were part of the F-35 program, where the specific electronic and radar profile are paramount to the tactical advantage of the weapon system. But Turkey decided they wanted to buy into the Russian S400 antiair system as well (given that their alternatives like Pariot is frankly quite lackluster).

Now lets put aside the potential of hidden Russian code providing telemetry back to the motherland about what is going on. Similar performance information about F-35 detections could easily be acquired through non technological means as well through human intelligence. The entire premise of the F-35 program is essentially that OpFor doens't really have great knowledge in the radar profile of what the F-35 looks like, meaning it could lead to an inefficient deployment of resources/unknown vulnerability in radar coverage.

So much of modern air war doctrine is premised on this notion, that the concern of what nation buys into what system creates systemic risk. If it's a narrow program that is intended to understand potential risk, I think it is easily justified. If it is to try to win contracts for American companies for a profit in a form of state sponsored industrial espionage a la China, I am fully against.

My gut instincts (based off of nothing but pure speculation and being a ML developer for several years) says no.

For one, the M1 engine has 16 cores, vs over 1,500 for a 1660 series NVIDA card. I know it may not be apples to apples, but I have a very hard time believing it would be able to keep up with even the most marginal card for training.

Not unless you're writing the backend CUDA code for TF honestly. Most of the time you're accessing your data/models/structure through the TF API so you really aren't doing too much that directly interfaces with the GPU.

To put it another way, when you're developing ML models (99.9% of the time), it's not like people are writing if statements in CUDA.

Yeah, exactly this. I mostly am using cloud VMs for training, but sometimes just need to be able to mess around on my local machine for my deployment pipeline. I don't really care if it takes ~10 seconds to do inference on a handful of images, I really just need to tinker with the interfaces to ensure everything is hooked up correctly.

IMHO it is missing the fully sampled (the ground truth 60 fps) solution. It's sorta hard to discern the quality of this work without this benchmark.

Mostly for dust. A lot of the smoke/dust particulate matter from the 9/11 attacks has been linked to nasty cancers, even among civilians simply fleeing from ground zero.

I actually had an earthquake bag, supplied with several N95 masks. Around march when COVID started becoming "real", I felt very fortunate to be able to retrieve them and send them to my parents. This has in my view vindicated the necessity of having a plan and basic supplies in case disaster strikes.

I think you're completely correct. The San Diego ship fire basically totaled a amphibious assault ship (helo carrier) was also apparently caused by a welding accident.

Yeah but pharma in my personal experience is VERY conservative. A lot of them still use SAS as their bread and butter for analysis. And guess what? All their older trials that they have to maintain are all written in SAS. It's basically how banking got stuck in a lot of legacy COBOL code, pharma got stuck in SAS. Want to do healthcare economic modeling? Buy a license for TreeAge (and it ain't cheap!). And it's not like you cannot do complex analysis in it, it just doesn't have a sleek interface. The "biggest" technology disputer in pharma I would argue would be Salesforce! They support creating all tons of custom input interfaces for patient consenting and record keeping. Enterprise ain't supposed to be sexy!

I just sorta worry if there is correct culture fit in pharmaceuticals when there is a prevailing mentality of "no one ever got fired for buying Microsoft".

Yeah, if we had such cut and dry "rules" about medicine, then we could automate a lot of what doctors do. Medicine is as much about the art as it is the science. You come to understand the limitations of different measuring techniques, and learn to know when you need to rely on each for more accurate representations.

I completely agree. My snarky comment was in reference to GitHub's marketing emphasis a few years ago on "social coding".

https://docs.github.com/en/github/getting-started-with-githu...

https://medium.com/glitch/github-glitch-and-the-future-of-so...

Not necessarily so! Gig (and self-employed) workers were included to participate in UI under CARES act, who normally are not eligible for UI benefits [1].

You are correct however in clarifying my statement above.

[1]. https://www.marketplace.org/2020/06/05/when-does-the-expande...

Because we don't have any efficient way for states to figure out exactly how much people were making before. Lawmakers have been on a campaign to defund, stifle, and keep unemployment insurance hard to access. This has resulted in ancient systems (even by government standards) that makes it very hard to figure out how much people make.

I cannot find the specific reference for it, but I believe I heard people saying some of these systems were having a hard time paying out over $1,000 a month because they were so old and outdated.

This is the case of "don't let the enemy of the perfect be the enemy of the good" that is repeated a lot in policy making. Sure, we could create a more perfect solution, but what are the costs associated with crafting that perfect solution under the specified given time frame?

https://www.cnbc.com/2020/06/19/outdated-technology-may-slow...

https://www.marketplace.org/2020/06/05/when-does-the-expande...

The $600 UI benefits were mostly done as a means of practicality, because our unemployment systems were so out of date to efficiently calculate someone's actual wage. It's sorta a general broad stroke implimentation that figured that sure, some people may be "overpaid" compared to their regular wage, but the political necessity of passing it ASAP was so great that any delay would be immensely costly.

| give them a better foothold in Social Networking than Linked-in provides.

So you're saying that GitHub isn't a social network for coding?

Gah! You're right. Thank you for pointing out my prior typo.

There was a really good recent interview with FCC Ajit Pai from the Neoliberal Project , where he discusses a lot of excitement of these small satellites to potentially enhance rural internet connectivity (an issue that seems close to his heart). He also mentions how evolving regulatory framework around small satellites, and how the FCC will become the de facto regulator for satellites despite the FCC primary mandate being the radio spectrum.

Listening to him, you can tell he's like a lot of other crony type Trump appointees; he seems to genuinely care about enhancing communication technologies in the US. I'm not saying I see eye to eye with him on all issues, but I think it's important for federal regulators to see the costs and benefits of policies, even if we fervently disagree with the final conclusion.

Link to the interview: https://www.youtube.com/watch?v=FUeNZsTajJQ

I work in the medical imaging space, specifically with implementing deep learning into clinical practice. I see a lot of people making a lot of fuss about what type of network or loss function to use. I would argue that this focus is misguided 90% of the time. Sure, maybe using a very specific network architecture and custom loss can edge you out by a 2-3% performance gain. But is that making or breaking the fundamental clinical application? I would argue that it usually is not. Instead, I've seen how much of the deep learning in medical imaging is driven by the quality and diversity of source data, which in the medical space can often be scarce for a number of reasons.

I'm reminded about this tweet, which emphasizes that a lot of your performance is going to be down to how good your datasets are [0].

[0]. https://twitter.com/lishali88/status/994723759981453312

Really? The message I got from American Factory was that there are stark cultural divides between Chinese and American manufacturing. It's been a while since I saw it, but they made a point to emphasize some of the poor financial decisions the Chinese bosses made during refurbishing (I think an overly expensive stage was built).

If I recall correctly, most of the moonshine blindness was not due to improper distillation technique (i.e. not cutting your distillation head), but rather that methanol was often used to dilute ethyl alcohol.

Perspective for (2): it's because no one in graduate training really cares about code quality. Your PI focuses more of your attention on scientific writing, and so there's little to not time to polish your work. The incentives just don't support this work at the graduate training level.

Here in Ca, not if you want to buy alcohol!

|What the boss really didn’t understand was the software you needed to buy to make the computer useful.

I would almost change it to say "what the boss didn't really understand was how to use the computers to make my employees more productive". I think the specifics in how you automate is incredibly important, and there are many cases of how it can fail. 737 Max MCAS system is a prime example of how ramming automation through without consideration of "human factors" leads to massive failure.