> I went to the library and did a small experiment; what could I do if I were to start with nearly nothing? Ignoring my library card sitting in my pocket, I went up to the librarian and asked for access to the public computer. She handed me an access code to use, and I sat at one of the workstations and logged in. Using the default browser, I now had access to free resources like FreeCodeCamp or Open Source Society University to learn anything I needed. I could make an email account for free and use that to make accounts elsewhere. You can't install anything on a public computer, but sites like Replit provide an online IDE to create almost anything. You can make a GitHub account to manage your code, contribute to others, even host projects thanks to GitHub Pages. You can make a free Heroku account and have a backend server to build on. Everything needed for success, although not perfect, was there. Nothing stops you from trying to start a small business, applying for software jobs, or making silly games as long as you're determined to make the most of what you have rather than feeling paralyzed at what you don't.
This seems to be the functional crux of the argument, but it's flawed, for at least a couple reasons.
(1) It presupposes a background in CS and software engineering, as well as knowledge of the ecosystem of tools available online. At some point in your life, you have to have had the time and stability in life to learn all that, and you need time and stability to work towards your goals in the present using those tools.
(2) It's too narrow in scope. This is shown as one example of a general philosophical mindset, but in fact software engineering is uniquely suited to starting out from nothing in this way. You'd be hard-pressed to find analogous examples that require any interaction with the physical environment.
If you're already into software, and you want to learn more about it or build something new, you're right that we have a lot of free resources available online to do so. And I agree that it's good to structure your life, when you can, so that pursuit of wealth isn't important, and you just happen to do so along the way. But I think it's difficult to translate that into a more general point about simply adopting a different outlook on your circumstances to effect meaningful change. There are big swaths of the human experience that that philosophy doesn't speak to.
Fortunately, given the purported scope of the hack, it seems we'll be able to actually quantify that. How many of the websites weren't hosting violent or extremist content? If this is real, we should be able to get an exact percentage.
To put it another way that's a little less accusatory but still encourages the right mindset - assume any comment on a service from a user without verified ID was written by a troll trying to provoke a reaction.
If you're playing games through Steam and haven't checked out Proton in a while, give it another try! The only game I haven't been able to get work on Ubuntu 16/18/20 is Fallout 3.
So there's some arguing over terminology, and then the main technical point seems to be that you can reverse-engineer a training dataset from the "virtual machine" built by training a neural network.
The decompilation process doesn't just use the neural network though, if I understand correctly it also uses logs from the final training epoch that include error and weight update data. Does this somehow smuggle the training dataset back into the VM? To me, if you're making a statement about the nature of existing ML systems, the statement "reconstruct the source dataset from the cryptic matrices that constitute the software executed by them" would imply that this is possible from trained networks alone.
A chef friend in DC tried this recently. They say marinated + fried they are good, like mild shrimp. Roasted however, and you get a snack that is crunchy on the outside with a "goosh" as you bite in. Not recommended.
Right? If someone runs for some local position with no requirements to get on the ballot, can they start posting incitements to violence against DeSantis or whoever? Writing "go shoot this guy" might run afoul of other laws, but if you just post about how a lot of people are saying so-and-so has a child sex dungeon and somebody ought to do something about it, that's still free speech right? Or can you still get kicked off for, say, posting objectively false information or roundabout encouragements of violence? That was basically where the "deplatforming" controversy started in the first place.
Isn't this a process failure? At least for that hypothetical 18%. If you have a product roadmap with well-defined milestones and good weekly accountability, it should be clear very quickly if someone is underperforming. There may be people who don't want to work remotely, but that's again why having a flexible+hybrid WFH approach seems like the best of all worlds.
Maybe instead of a centralized repository, individuals could be given the ability to create their own partial sets of recommendations that could be aggregated by a user. Instead of trying to globally solve trust and designation of expertise, you explicitly define the group of people whom you designate experts and whose assessments you trust.
NVIDIA reps at my place of work have no trouble getting us A100s. Maybe this is too cynical but I think they learned their lesson after the Pascal generation was released - everybody bought the GTX series because performance/$ was way above what Tesla offered, even if Teslas were within budgets. This time around they can actually price segment for enterprise customers!
There's cause for optimism! Basically, it'll take a big chunk of 2021 to vaccinate everyone (at best), immunity may take a while to kick in, and while vaccines may be 90ish% effective, continuing to follow transmission reduction measures will help further mitigate spread while COVID is still prevalent, so keep it up for a couple months after the bulk of people start receiving vaccines. Not amazing, but also not indefinite, and I believe this has been the messaging for quite a while now.
Thanks for sharing! This is all good to know, I'll be reaching out to see if there's any possibility of these groups helping to address some of the issues within our institute at the NIH. Right now it's unclear to what extent that can assist an intramural research program.
Yeah this meshes with my experiences working at the NIH. Lots of interest in big data, AI, etc. etc., and even financial incentives for new hires in those areas, but absolutely no coordinated approach to recruitment, retention, or project management. A few labs or support groups end up with great teams, but mostly tech knowledge is redundantly siloed in small mediocre groups across dozens of teams institutes-wide. And that's not even touching on IT issues...
Nice find! Time to put together a pipeline... (1) take a few photos of a thing you want to model, (2) use a segmentation algorithm to mask out the background, (3) run a NeRF algorithm, (4) use marching cubes to generate a mesh.
Right, I'm aware of the structure from motion stuff that's already out there and its large view requirement. I've had a similar experience. Hence my question about using something like PixelNeRF instead.
It'd be great if you could use something like this for quick-and-dirty 3D asset generation, but is there any way to convert neural radiance fields to meshes? It'd be cool if you could take a photo or two of an object and get a 3d mesh out quickly.
For a recent publication, I had to create a neural network architecture diagram and tried it two ways - one by hacking someone's python code to create an isometric 3d diagram (https://i.ibb.co/HdBvTFk/fig2.png), the second using draw.io to create a 2d diagram (https://i.ibb.co/kmrfpVK/fig2a.png).
The 3D figure looks more appealing to my eye, but I went with the 2D one instead. The issue with architecture SVG generation code that I've seen is that it's inflexible once you want to do anything outside common layer computations or arrangements. I found someone's code that had a U-Net example and then manually had to go in and modify a whole bunch of stuff to get it to do just what I wanted. On top of that, I opted to do a bunch of postprocessing in an image editor anyway. More importantly, the 3D diagram is harder to parse for anyone looking to re-implement the model. Most (but not all) of the information is there, but you have to work for it. With the 2D diagram, the architecture is fully specified and, I believe, very easy to read (though I'm open to feedback). It's not as pretty, but ease of manual creation and the boost in clarity are more important.
NN-SVG looks like a better UI around a beautiful parametrized isometric 3D architecture diagram generator, but it also looks like it suffers from the drawbacks I have experienced in the past. Can it handle residual or skip connections? How easy is it to modify? If you are creating publication-quality architecture diagrams daily, automating it may save you time, but if you've gotta write a bunch of code every time your model's components are tweaked and only need pretty diagrams once and a while, you might be better off doing it by hand. Moreover, whether you use a 3D or 2D diagram, I think it's worth favoring a format that aids reproducibility.
Does it matter that Venus has a lot of extra energy "lying around", so to speak? On Mars everything is very cold, but a floating cloud city on Venus would have some crazy thermal and chemical gradients to tap below it. I can imagine tons of engineering challenges in those conditions, but it seems like the analog of "geothermal" energy on Mars would be less useful than Earth's, and could be considerably more useful on Venus.
Please, HN readers, gather round for my important announcement. It has come to my attention that the article's author, Rusty Guinn, has been chaining young women in a dungeon beneath his home at 1234 Main St. Please do not turn your attention away from this announcement, like THEY would have you do. Please do not listen to the fake news media telling you I am not credible. Rusty is kidnapping these women, and the liars and the elites won't lift a finger to stop him. Something needs to be done! Who will do it? I know many thousands of you will read this, could somebody be a hero? I don't know, i hear a lot of people saying so though. All i know is if this comment is moderated, it's proof that I'm onto something, and the HN mods don't trust you to make up your mind yourself.
Yeah, federal employee doesn't usually mean "integral and knowledgeable part of governmental power structures", often enough it's just an average person who wants stability and not a lot of challenge.
I agree with the effort to find privacy solutions, but it seems like enforcement of surveillance bans like this will be an uphill battle as long as the physical video hardware remains present. 2020's cutting-edge facial tracking systems will be 2022's Intro to PyTorch tutorial, available to anyone with a $500 GPU. As long as video is being captured and stored, how do you realistically prevent it from being subject to analysis somewhere?
It's a very different world but i relate to this a ton coming from biomedical EM. This article does an awesome job explaining what makes biomedical computer vision so hard.
There's huge opportunity for automation, a wide range of vision challenges, and very cool rewards if challenges can be met. We're working now on building 3d nanoscale structural models of COVID vs. control blood clots and modern hardware leaves tons of data on the table for better software to take advantage of. The relevant vision problems are very challenging, and even just within EM, robust computer vision that isn't built up around individual narrow high-visibility problem domains (like neural circuit tracing) has a long way to go.
It's a terrible rule of thumb, especially because it can be deliberately manipulated. I say HN user johntfella maliciously killed 12 puppies, and i heard a lot of other people say the same thing. He claims that didn't happen. The truth is probably somewhere in the middle.
This seems like six totally unrelated techniques, any one of which would be its own project type, and only some of which actually involve machine learning. Is there any context for this? It seems unmotivated.
I don't think this changes anything. Deployed networks typically use only inference from pretrained weights, and those weights are what get transferred for model "updates". You can have all your devices using weight array W0 for a neural net architecture, spend a million compute hours training that net on cutting edge systems to produce a much better weight array W1, then upgrade all the deployed devices by sending them W1 which will be the same size as W0.