There are two very distinct types of cinnamon…
5,948 karma · joined June 23, 2017
On the other hand, neither one is guaranteed to be optimal for continuous thrust maneuvers, which with electric prop are increasingly common.
When you perform a finite sample reconstruction, this is essentially the unstated approximation you’re making.
There are many signals of practical interest that can be approximately reconstructed with a finite truncation of the series. Note, however, that any signal that has only a finite length, eg has a uniformly zero amplitude after some time t_final, does not have a finite bandwidth, and cannot be exactly reconstructed by any sampling scheme. This is the case whenever you stop sampling a signal, eg it is always the case whenever you step outside the mathematical abstraction and start running real code on a real computer. So any signal reconstructed from samples is always approximate, except for some relatively trivial special cases.
- the signal being sampled has to be stationary
- you have an infinite number of samples
In that case, a sampling frequency of 2N+epsilon will perfectly reproduce the signal. Otherwise there can be issues.
There is indeed a list of rejected papers. The system logs all of them. Generally they're recycled, updated, and published elsewhere.
In order to publish a research paper, it has to be reviewed for suitability for public release. This process is more than a little silly, because it requires seven levels of review, of which exactly one - my immediate supervisor - will have any idea what the paper is about. But fine.
There used to be a paper form. You'd fill it out and either route it around for signatures, or if you had a time crunch, walk it around yourself. Eventually they replaced the paper form with a web form, so now there's an automated queuing system that emails people when they have a paper waiting to be reviewed.
The web form has all of the same info as the paper form, with one addition. They scanned the paper form and turned it into a pdf, and they make you fill out both the web form AND the pdf version of the original paper form. So to sign off on a paper, you now have to download the pdf, digitally sign it, upload it again, and hit the "Approve" button on the web form.
Because God help us if anybody does an audit and we don't have all of the forms correctly signed.
I'm an ML researcher and a sheep farmer. My lambs are up on their feet, nursing, thirty minutes after birth. Once they have enough blood sugar they are capable of running to keep up with mom. They come prepackaged with a fully functional quadruped locomotion scheme and associated path planning and obstacle avoidance and fully functional vision, touch, and audio processing algorithms. And this is with basically zero embodied learning.
The difference between sheep and humans is, I think, that humans actually "learn" on three timescales instead of two, unlike every other living thing. Sheep learn on evolutionary timescales, through natural selection, and on the timescale of an individual sheep lifetime. But humans learn on the timescale of society as well, in between the other two. I believe that the difference between us and animals isn't so much our dramatically increased intelligence as it is the ability to pass detailed descriptions of what we know on to others, e.g. to language. The amount of information in the world has increased exponentially since the Renaissance. Societies advance on a characteristic timescale of about a century, as far as I can tell. Much faster than evolution, but quite a bit slower than individuals.
In ML, we are doing the same thing as sheep, basically. We have two "learning loops". One is the continual development of new DNN architectures, which corresponds to what evolution does; the other is the training of these architectures on data, which corresponds to what individuals do. But the outer one still mostly proceeds at the speed of human cleverness, not computation. But we are just... relying on the fact that society, and the data it produces, is there for the consumption of ML. We do not have any ideas for speeding up the production of that data except for the hope that we already have enough of it to kickstart GAI. If we do, we'll go through the singularity. If we don't, we won't.
But we can at least solve the "evolutionary learning" problem for ML. We'll need to bring back something like genetic algorithms, or make DNN architectures somehow differentiable so they can be efficiently evolved.
Then it got bought by RIM (BlackBerry). The source code went away and the owners stopped responding to the user community and the whole thing basically became irrelevant.
If you’re paying two billion $ for something you become very very interested in test design and test results.
Also, safe mode isn’t really the same as a BSOD. It’s a mode where the spacecraft decides something is wrong and disables a lot of functionality and focuses on pointing the solar panels at the sun and the antennas at the ground. It does not cease functioning - if that happens, you’ve probably lost your spacecraft. It is therefore VITALLY IMPORTANT that safe mode works, and a smart program manager tests the hell out of it.
I am a spacecraft engineer. I don’t see anything in the linked article indicating that they are actually running Windows - the BSOD claim is tongue-in-cheek, or at least that’s how I read it. I also don’t know of anyone anywhere that runs Windows on a spacecraft, with the exception of laptops used by astronauts. Typically one runs vxWorks, or maybe QNX. Some experimental (high risk, low cost) systems run Linux. Older spacecraft don't run any OS at all, everything is running on bare metal, and that may be true for a handful of current spacecraft as well.
Windows is used in some places by ground controllers, but these days they tend to be running Linux a lot more often.
By design, 1Password always makes you re-authenticate every time you lose focus on the app. But Face ID (or Touch ID) makes reauthenticating a lot less painful.
There is a lot of discussion in a lot of threads about the design of the robot to water "from the top" by spraying the leaves instead of watering directly on the roots, and whether that's a good or bad thing, and whether the designers of the robot thought about it.
Here's the problem with watering the leaves: yes, plants ultimately get their water from rain. But under normal conditions, the rain comes in sporadically in large quantities -- not every day -- and soaks into the soil, which is where the plants actually pick it up. Flood irrigation does largely the same thing. Spray irrigation doesn't attempt to water the soil that deeply, it tends to give the plants just what they need for the next 24-48 hours, and that encourages wilt and fungal infections.
Also, domesticated vegetable crops are far more susceptible to wilt and fungal infections than natives, and than grain crops, which are at the end of the day grasses. So you can in the same garden have perfectly healthy corn but all of your melons and squash have such bad fungal infections that the leaves are literally white. You can criticize the selection of vegetables for yield and not hardiness, but the fact is this is where we are with vegetable crops.
This is an interesting project, but IMHO it isn't practical, and there isn't any way to make it practical. The X-Y gantry design, for gardening, has a number of intractable problems, watering from the top being just one of them. Another is that the design doesn't scale. You can't make this thing handle a 25 by 100 foot grade bed, which is the size you'd need to even start making a serious dent in the nutritional needs of one person. It can't really weed, and there's no way to modify the design to make it weed effectively; you'd have to add degrees of freedom to the gantry so that it could reach down to soil level and grasp roots (or, alternatively, to very selectively apply an herbicide). Garden crops grow to dramatically different heights; micro greens will be a few inches about the soil, zucchini will be three feet high, tomatoes can be 4-5 feet, and corn depending on cultivar can be as much as 9 feet tall.
And finally, watering and weeding, if you know what you're doing are actually the easiest parts of the problem. Preparing the bed so you don't have to weed is a lot more work. To do that, you plant your crops and then apply large amounts of mulch. If you've never prepared beds, shoveled dirty barn straw for mulch or tried to wrangle weed barrier cloth on a hot, humid day, you haven't lived, my friend. That's the physically hard part. THe mentally hard part is diagnosing problems in your crops before they become problems. Noticing that those shiny weird insects flying around are squash vine borer. Looking at the underside of leaves and seeing squash beetle eggs or going around your tomatoes with a blacklight looking for cutworms.
If you want to apply robotics to gardens, you either need a low mobile base, or you need to carefully lay out rows with fixed spacing, and have a high mobile base that can clear the height of the crops, and can take a variety of attachments, e.g. tillers to handle weed control. Which means you need think about monocropping. Which starts to look like the mid 20th century basic garden tractor, the International Harvester Farmall Cub, just with maybe an electric power plant and an autonomy appliqué kit. THis makes sense because the mid 20th century was the last time people in North America practiced gardening as a survival mechanism, and the Farmall Cub was the result of 50 years of practical design by people who knew how to garden when it counted.
This has not been my experience, my agency hires new PhDs at the GS-13 level and BS/MS graduates at the GS-12 level. But YMMV.
"It's possible to find a job at some agencies working on stuff that feels very rewarding and very important, but you will always constantly feel like the bureaucracy is constantly fighting you."
Concur.
Of course, government jobs are very stable, which can be an advantage or not depending on what you want out of life. We find that most of the engineers we are able to recruit have been to industry or academia and found the instability and poor life balance to be incompatible with having a family. So my "new" recruits tend to be a little older, married, with a child or two. On the other hand, many of my actual younger hires eventually get headhunted and leave for industry to make 2x more than I can pay them. Then they find out that on an hourly basis they're making roughly the same amount, because their new jobs expect 20-40 hours of unpaid overtime.
Also, you're actually expected to work 40 hours a week and generally not more, and you're expected to take all your vacation.
It takes a while to get to that level, or you need to get hired directly into a GS-15 equivalent position. But it's doable, especially at an agency that does technical work (eg NASA, NOAA, DOE, parts of the DOD, etc.)
We had an official manager, a "branch head", who was worse than incompetent. He couldn't find his butt with both hands, but he also thought he was God's gift to management, and would forcefully and emphatically make bad decision after bad decision.
Eventually, he had screwed up the group's major program so thoroughly it looked like a sure fire failure, and he found another job, and didn't bother to tell anyone; he just stopped showing up for work one day.
The level of management above him had bigger problems to deal with than replacing him, so they made sure we had a competent secretary and left us alone for two and a half years. It turned out to be arguably the most productive period of my professional life. My buddy and I took over business development. THe team turned the big project around and made it a rousing success, and grew the funding from it by two orders of magnitude.
The point being: there's bad management, that acts randomly or not at all; and then there's really bad management, which takes up your days with constantly changing orders, fixing relationships with customers or sponsors that they've screwed up, and levying time-taxes in the form of training, reports, and morale boosting exercises.
If given a choice between bad management and really bad management, pick bad management.
Fifteen years later, my buddy runs the place and I'm the senior scientist.
This incident is exactly what precipitated my switch to Fastmail.
Unfortunately, I did it by making them gmail accounts. Google without warning closed both accounts when my daughter was 12 for being under-age. I lost everything. I tried to appeal to get them to unlock the accounts long enough for me to get the contents out, but talking to a human being at Google is famously impossible.
rm -rf *
Yep; I actually did it. By mistake, as a junior in a CS research lab.
I’ve been working on robot control using physics-based computing devices for about a year. Not yet able to write down the problem clearly enough to attempt a solution.
Otherwise, if NASA issued stock, you should consider shorting it.