It's much easier to work with LISP. It's functional, trees are easy to reason about. We're using an even more simplified language in this work. It's kind of like the MNIST for progsynth research. Scaling this to real and useful programming languages and domains is still non-trivial, and is a major drawback of our work here.
Yeah, this is true! These are more like expressions rather than programs. We were mostly following the language used by previous work, https://arxiv.org/abs/1906.04604
Couldn't this be used to do HTML generation from designs? Especially when combined with multiple viewport sizes at the same time, generating a fluid HTML layout would be pretty awesome.
Changing tokens in a program is not differentiable. For me, the key idea is that you can train a neural model to suggest edits to programs by randomly mutating nodes. And when you run this neural model, you get to make edits that are syntactically correct (i.e., a number will only replace a number etc.) according to a context-free grammar.
Adding to the thoughts before, letting the virus spread and mutate among the unvaccinated to create a new variant sounds like a reset on the progress we've made. With the Delta we got lucky, our vaccines still largely work. I'm not sure how long the luck will hold on with more variants.
I'm all for leaving people alone to their own decisions. But with this, their decisions greatly affect me and society at large.
You won't stop mutations. It's a pandemic, world-wide - and short of blocking all international travel, freight/shipping, etc - if there is a new mutation somewhere in the world, it will eventually land here in the US.
Herd immunity - either reached by vaccinations, previous-infections, or both - is all we can hope for.
Good news is, we don't need unvaccinated people's cooperation to reach herd immunity - they will eventually have natural immunity anyway.
Of course, we can never stop all mutations. However, for mutations to be useful for the virus, they do need to happen at an evolutionary scale. Especially because we're lucky that SARS-CoV-2 mutates much slower than Influenza and HIV.
If the rate of transmission reaches the elusive herd immunity, the time-frame for the virus to evolve into evading vaccines grows significantly longer, even with a non-zero mutation rate.
Lastly, antibodies created via the mRNA vaccines have a broader binding affinity to different variants than those from natural immunity [1].
Again, the goal is to not eradicate, but to scale it down to the point where it is statistically not a threat.
Your original point was we got lucky with the Delta variant because our current vaccines still work.
I'm saying it doesn't matter - even if we had a 100% vaccination rate, some new variant will develop someplace that's being ravaged right now (say, India) and spread to the US eventually anyway. If our vaccines still work - great... if they don't, well.. where are we then?
Lastly, the numbers already indicate it's statistically not a threat to the overwhelming majority of the US population as-is.
I wasn't talking just about the US, I think we need very high levels of vaccination everywhere to make it statistically not a threat. We need to control the spread of the virus to levels where it's not mutating at evolutionarily scales, _everywhere_, and the only way to effectively do that is vaccines.
It's not a threat at this very moment, for this very point in time, but since we're allowing the virus to spread at such evolutionarily scales, inside and outside the US, a new variant popping up that evades vaccines is statistically very possible, and pose a significant threat to my own and society's well being.
I'm trying to argue that this is not a personal choice issue, rather a public health one.
I've engineered it to be very inexpensive to run. I work on high efficiency video codecs, so one benefit is that I get to sometimes test my codec's and get real user feedback.
I personally don't find virtual stuff fun, the pandemic is really sad, so wanted to make something available for public use. :)
Hey HN! I am a TA for a large undergraduate class at MIT. Teaching virtually definitely comes with its challenges, but the biggest challenge for me was holding office hours.
We used to have a queue system, with hundreds of students trying to get the first spot, with me (along with the other TAs) trying our hardest to keep the queue moving quickly, but algorithmic discussions need time and frequent pauses to think.
I experimented with lots of alternatives, but none of them scaled to hundreds of students (inexpensively) or had strong permission systems. We wanted a clear separation of permissions between the Instructors, TAs, and Students. Having a multiplayer code editor is a nice extra feature that has been useful to help students debug their code.
If you're in a similar position, or just want better breakout rooms, feel free to use the system. The default cap is 100 users to prevent abuse, but I can happily raise it if you reach out.
The difference here is that this library is a one-line drop into your Python back-end to automatically generate the client.
Of course, the route you mentioned is much more powerful and complete, but here I'm trying to offer a way to very quickly build APIs without learning how to integrate any of those tools.
We bought a bunch of ov5647 camera modules from AliExpress. A quick search on the site yields a couple of listings at about $3.26 per camera, here is one of the listings, https://bit.ly/2W0Xui6
Thank you! That's super cool :) The fastest strobe we used was a 500ns strobe, which is a 1/(2e6)th of a second exposure. The Pi cameras use the OmniVision OV5647, dynamic range of 67 dB @ 8x gain.
I'm pretty curious as to what you've discovered / worked on so far or have any interesting CV ideas you'd like to see?
thats ln(10^6.7)/ln(2) = 22.257 bits, that's unheard of for camera sensors I have looked up, so I assume the reported values are "gamma" compressed before digitization.
Would you have interest in characterizing or collaborating on characterizing the exact compression function for the V1 and V2? Relevant literature seems to be Steve Mann's comparametric equations: https://en.wikipedia.org/wiki/Comparametric_equation
(I am interested in high dynamic range, and high monochromatic or color bit depth for an experiment, which will progress much faster if I can start out with a higher dynamic range and bit depth sensor, I will need to oversample to observe a phenomenon, and every bit of increased depth a sensor has compared with another sensor would mean the experiments can be run 4 times as fast...)
I'd probably like to have a go at replicating some of the Blue River, Bilbery ideas but on a budget. That is sensing weeds or crop issues on the move therefore requiring global shutter. Uniform light throughout the day/night is another issue I am thinking through.
Working through fastai when I get a chance but my biggest issue is finding domain specific datasets.
