thearn4

Federal research (i.e. civil servants at NASA, NOAA, DOE, etc.) tends to work closer to that model, but not 100%.

I'm not a medical professional, but the whole discussion of handoff risk always seemed to me like it was side-stepping the real issue presented, which is poor documentation and/or communication between peers. Instead, the premise is offered by the AMA that handoff risk can only be minimized by insane shift lengths.

I think ESP8266 boards (running something NodeMCU) are becoming the more popular replacements for Arduino. They're super cheap, and have integrated Wifi.

Cool thing about them is that you can write in C-like AVR code (like Arduino), or micro versions of Lua, Python, and other choices.

This is a good point, and a relevant one when discussion Common Core with people. At the surface level, I don't see anything in the new mathematics curriculum that I would consider negative, but the biggest complaint I hear is that parents can't understand it, and some teachers barely do either. But that hints at a deeper problem.

It concerns me a bit that at some point, there are fundamentals concerning simple yet effective approaches that may be forgotten (or a least not taught) in favor of black box neural nets, like Kalman filtering, PID control, etc.

Institutional memory is something of a weak point in the software industry compared with other fields.

Linear algebra is the ultimate common variable in technical computing and applied mathematics at large.

I'm pretty close to Lake Erie, the sewer drain of the other Great Lakes. Invasive species are pretty bad here too. But the agricultural runoff and season algae blooms have been the most recent headache in the western portion of the lake.

Serious question, why do the major tech companies put up with this and not slowly migrate elsewhere? I get it's a bit of a chicken vs. the egg game when it comes to where the work is & where the talent is, but the bay area seems like it's at such a critical mass.

I feel like Python is the closest that we've come to this, though it is imperfect.

So for someone who has tinkered on the periphery of Java development (but does not work with Java/JVM languages professionally), it isn't entirely clear to me what the purpose of modules is for developers, over the encapsulation already provided by packages. Is it to simplify deployment?

I reached this point in graduate school, though it took a few years for it to feel that natural.

FWIW, I'm a GED holder, but later finished college & PhD. My linkedin profile is 100% complete, and I find zero utility in the platform. You really aren't missing much there.

Do you feel the same way about visibility in OOP? E.g. private data and methods?

Does DL provide mechanisms for feeding back into theory? As in, does a successful deep convolutional neural network provide a means to extract enough from it's structure and behavior to potentially not NEED the CNN for prediction in a future iteration? Gradient projects can be used to gather a total derivative quantity, and compare sensitivities across inputs. We can regularize to prevent overfitting with cross validation, L-curves, etc. But what about hypothesis generation?

For many of us that have dipped our toes in the ML tooling but don't have a great application for it in our work areas, this would is the kind of thing that we would like. A NN that predicts well, AND has a well understood methodology that gives us actual insight and not just a black box.

Maybe what I have in my head is a deterministic gradient-based analog of an evolutionary algorithm? I'm not sure.

Count me among the "moon first, then mars" crowd.

Are you saying that you think IRBs are a waste by design, or that the way IRBs are typically implemented makes them so as a side effect?

> professors are getting paid to do next to nothing

Speaking from the part-time professors perspective: occurring in parallel to this is a steady growth in course section sizes. 30 students used to be the norm for some courses, but 80-90 is the new normal just a few years later. Some of these tools are designed to help manage this. But the actual process of planning and delivering a lecture, providing individual feedback on written assignments (like exams), etc. is suffering.

Teaching effectively to a large section is harder than you might think, especially when teaching is only a fraction of your job & you have no TAs. But half-assing it is actually not too hard I guess. And you are seldom rewarded for the former. Thankfully for me and for my students, it is mostly a side hobby for me and not a living.

I taught a calculus class the the department had mandated be managed with MyMathLab. There is a kernel of a good idea in these subject-specific course content management systems, key among them is they give a reasonable way to actually manage homework for large course sections. I tended to grade HW very lightly, I just want them to have their head in the space a small bit every few days, and MML was a way to make that manageable for a section of 70+ students without losing my mind.

But the execution is so half-assed that it not only turns off students from the platform, but it runs the risk of turning them off to the actual subject matter as well. The whole thing had the stench of static requirements doc to a sub-contractor to a sub-contractor development pipeline with little to no end-user (that is, students, not department administrators) feedback in the development cycle.

It's a bit like self-driving cars in a way. The execution is everything, and the half-way hybridized situation is the most dangerous for the end-user. And the missed execution to me smells a bit like an arms race between Pearson, Wiley, etc. to get their platforms out and entrenched with individual departments before each other.

I'm teaching a data structures class w/java this coming fall semester, and am putting together a reference text/notes in markdown to provide for free, since it seems really silly to require a paid text for this stuff.

Every single item that contributes to the learning objectives of this course is incredibly Google-able, but as new students to the field, they do need at least some curation and direction to keep from getting overwhelmed. They need a reference resource that they can trust which is academic enough to be true, yet reasonably conversational. Textbooks used to fill this role, but publishers priced and walled-gardened themselves out of relevance.

Side note: if you're an professor who has done something like this in the past, I'd love to compare notes with you.

For what it's worth, I've had luck moving some people's anti-science opinions by meeting them at their level of understanding, and offering some compassion. Might help that they are relatives and not total strangers though.

I'm extremely skeptical that this could in any way be possible, but is it remotely possible that she herself did not know the extent of the fraudulence of the tests?

1. Aerospace

2.1 Access to energy / energy density of fuels (batteries included). This is the case across a wide range of industries and problem areas of course, not just transportation. But incremental optimizations in efficiency have lead to squeezing more performance out of the margins, but no Moore's-law type growth will ever happen without some kind of energy breakthrough.

2.2 Going forward, tightly coupled systems will be the norm. The traditional tube-and-wing aircraft with bolted on nacelles is a bit of a dead end for civil aviation. Systems to enable a more complex design workflow (e.g. graph based dataflow with accurate gradients) will be more paramount.

3. Research into the next generation of energy storage materials, and improved large-scale gradient-based numerical optimization algorithms.

Yeah, the strange "data structures and algorithms are either everything or nothing" polarization in the opinions on tech interviewing is strange.

I agree with the parent comment, but also agree that "soul" could come from writing, voice talent, assets, etc. Deus Ex and Mass Effect are two games I can think of that offer an amazing experience, despite the actual programmed game being fairly shallow. They were unique and well written, which was enough.

FFT and the convolution theorem are some of my favorite tricks in math. A toy I made a few years back:

https://github.com/thearn/game-of-life

I'm in N.E. Ohio, just outside the Cleveland metro area. The croplands here aren't at large as what you'd see in central OH, but plot sizes of around a hundred acres aren't too hard to spot.

That's a good point that autonomy could be a big deal for some consumers, and I bet larger plot size is a factor in that.

There is a fair amount of competition in this space. Around me, I see a lot of Kubota, New Holland, etc. Not too many John Deeres actually.

they could probably have just depressurized their suits and would have lost consciousness extremely quickly.

Teaching something is a great way of reinforcing the material for yourself, since to end up having to conceptualize the material in a few different ways to meet your students at their own level of understanding. I think it's cool that UCLA is giving undergraduates that opportunity.

I'm curious, are you saying that America should not have responded militarily to the attack at Pearl Harbor, or to the declaration of war on it by Germany?