A recent trend has been to use a language model to do the scoring: https://arxiv.org/abs/1904.09675
274 karma · joined February 28, 2015
A few weeks ago I spun up wikisurvey.org to try and make a place for Wikipedia-like collaboration on research. There's nothing on there yet, but I'm hoping it will enable low-latency collaboration once we get the ball rolling.
I think he's excited because he thinks that in this framework we can just "throw compute" at the problem to enumerate the possible theories of everything, and automatically rule out the ones that don't align with reality.
Personally, I'm quite bullish on the resurgence of SVMs as SOTA. What did it for me was Mikhail Belkin's talk at IAS.[1]
[1] https://m.youtube.com/watch?index=15&list=PLdDZb3TwJPZ5dqqg_...
However, the approach of understanding math through code is still very helpful, I think. Personally, implementing things that are fuzzy mathematically provides immense clarity once I write them down.
For example, a simple monty hall simulator[1]. Or implementing matrix multiplication multiple ways to understand why each is equivalent[2], and why multiplying A(BC) can sometimes be faster than (AB)C[3].
I am not sure why this helps me. It may be because I was "raised" as a coder, and so that is how my brain works. But I also think that implementing something in code is very close to constructivist mathematics, in spirit. You cannot prove anything if you cannot construct (/implement) it.
[1] https://github.com/mitchellgordon95/implementing-paradoxes/b... [2] https://github.com/mitchellgordon95/implementing-paradoxes/b... [3] https://github.com/mitchellgordon95/implementing-paradoxes/b...
It's multifaceted. It might be fine if you own a large enough home that you can have a dedicated working space, with a door.
Even without kids, I find it extremely difficult to work from home without a dedicated space. If I try to work from the living room, I end up doing living room things (like playing video games).
For some people, though, the idea of engineering your environment to change your behavior is either impractical or unheard of, which is what I think gives these people the general impression that "working effectively from home is impossible."
I was a little disappointed when I read this. When I read the title, I was hoping for something based on natural language processing summarization, which would aggregate many stories into cohesive chunks.
The methods discussed there take a different angle at the problem.
It has been my go-to editor for many years now; I highly recommend it.
Friends are peers that have similar life experiences and values (and humor). Sometimes people like that are hard to find.
https://medium.com/tensorflow/fitting-larger-networks-into-m... https://medium.com/huggingface/training-larger-batches-pract...
https://github.com/Innectic/isnt-thirteen
Incredible.
That's not true! We can also use it for undermining democracy, making everything look like a Van Gough painting, and...
> This algorithm is similar to [...] what self-driving cars use to "imagine" road scenarios.
I understand the fear here, but I don't appreciate rage-bait sub-headlines that contradict the rest of the story. Obviously GANs can be used for many things.
Also, I'm sure there are a few bad actors who will use this for things like revenge porn (which is Not Good), but I suspect the driving demand behind this application is horny teenagers who think Daenerys from GoT is hot, rather than malicious people who want to "oppress women." Although to be fair, this is the effect the technology has, whether intended or not.
I think we have similar opinions about what the tool is, and what it's not. I just must have not expressed myself in a very clear way. I did spend about two days hacking on this, just to get a basic POC working. But in the end, I couldn't figure out how to put right angle markers on my vectors. I also submitted an issue about vectors getting smaller when the norm was less than 1.
I'm just saying, the title makes it seem like it's a general purpose tool, but it's not.
But like some of the other comments have mentioned, it's not a general purpose tool... it's more like he open-sourced his dot files.
You can certainly make a few things work, but it would be nice to see a community of interested individuals come together to flesh out the abstractions, make sure everything is documented and works as advertised. As it is now, the class hierarchy is a little... misshapen. It's not clear what classes are responsible for what functionality, and how everything fits together in a nice, orthogonal way.
1. The documentation is sorely lacking.
2. Which makes the inheritance hierarchy (which is already a little unintuitive) really hard to grok.
3. There's this weird home-rolled way of doing default property inheritance that took me a long time to wrap my head around.
In the end, I ended up coding a short interactive thing[1] for my blog using Paper.js. It's a great idea, but could really use some love and maybe a stronger philosophy to guide development. Right now it feels more like a few hacks thrown together.
[1] http://mitchgordon.me/math/2019/03/18/multiplying-complex-nu...