trextrex

I've used a similar solution called TalkYard [1] which provides a interesting combination of commenting and forums, and has super-helpful support.

[1] https://www.talkyard.io/

About 90% of the complaints there seem to be from users who have obviously not read the instructions in the linked page. About 10% seem to be from users who are being charged after cancellation which is a bit worrying.

The problem with a failed experiment, at least in machine learning, is that it's not always clear what caused the failure. Was it not enough training, was there some small "trick" the model could have used, were the hyper parameters off etc.

There are an infinite set of configurations that could fail, and it's not sufficiently useful to know they failed without understanding why they failed. And analysing failures in a useful way is an extremely difficult and fundamental problem. On the other hand, a successful experiment is an extremely rare event and hence interesting by itself.

> Replacing forward and inverse kinematics with machine learning is extremely simple

Could you point towards a good paper or review paper on using machine learning models to learn forward and inverse kinematics? How far can one get with plain old LSTMs?

I meant to say that you would use that scheme while you were working on your lessons, and at the end of it all, do a single pass of renumbering all the lessons with an appropriate number scheme.

It seems like the author's file numbering problem could have been easily solved with a slightly different numbering scheme -- reserve a couple of digits for additional content. So start with 0010, 0020, 0030 ... and if you wanted to add something between 0010 and 0020, it becomes 0015 and so on. Then you would have to renumber the files only at the very end of the process.

Did you actually read the article? It's about a lot more than just the Venezuelan coup.

This link has a pretty good explanation of why the supply-demand model for the job market doesn't work: https://www.bloomberg.com/opinion/articles/2018-04-05/supply...

Out of curiosity, can you explain, or point to some writing about why you think the theories in the book are close to "complete bunk"?

> There's a lot of good evidence that correlates race and IQ. Even when controlling for as many factors as possible. To the best of our current scientific knowledge, this is a thing.

Care to point to actual data supporting that statement?

Similarly the population of many of the Nordic countries is comparable to mid to large cities in US/Asia.

They charge for the service, although IIRC they do have a free version with limits on number of bookmarks.

After increasing problems with pinboard, I recently cancelled my subscription and moved to ggather.com. It's been pretty good so far.

I have been using ggather.com for the past few months and liking it quite a bit, especially it's bookmarking and organization interface as well as performance, as compared to pinboard. The only thing I miss is having an android app.

Flattr's model seems really well suited for conveniently contribution back to open source. Does Flattr not work for software anymore? I'm trying to figure out why it's not more popular as a way to contribute to open source software.

According to the Wikipedia article on movable type [1], it existed long before Gutenberg, in East Asia. It seems like Gutenberg's contribution was to introduce some additional innovations and popularize it in Europe.

[1] https://en.wikipedia.org/wiki/Movable_type

This reminds me of other similar attempts that have happened in the past to help "assimilation" and "integration":

https://en.wikipedia.org/wiki/Canadian_Indian_residential_sc...

https://en.wikipedia.org/wiki/Stolen_Generations

Having interacted with immigrants from many different places, I find for a vast majority of them, the environment does in fact determine their culture. Are there a few for whom it doesn't? Sure. In the same way there are criminals even among people who grow up in such an environment.

Have you ever considered that what you consider "western values" are in fact values that emerge in a society that lives in conditions where there is political stability and economic opportunity? That people's "culture" is affected by the environment around them as a general pattern?

While it's a good idea in principle to publish failures, in practice it's a bit more tricky. So a particular model didn't work. Does that mean the model is fundamentally flawed? Or that you weren't smart enough to engineer it just right? Or that you didn't not throw enough computing power at it?

In a vast error landscape of non-working models, a working model is extremely rare and provides valuable information about that local optima.

The only way publishing non-working models would be useful would be to require the authors to do a rigorous analysis of why exactly the model did not work (which is extremely hard with our current state of knowledge, although some people are starting to attempt this).

Agreed. The hype surrounding machine learning is quite disproportionate to what's actually going on. But it's always been that way with machine learning -- maybe because it captures the public's imagination like few other fields do.

And there are definitely researchers, top ones no less, who play along with the hype. Very likely to secure more funding, and more attention for themselves and the field. Which has turned out to be quite an effective strategy, if you think about it.

The other upside of this hype is that it ends up attracting a lot of really smart people to work on this field, because of the money involved. So each hype cycle leads to greater progress. The crash afterwards might slow things down a bit, particularly in the private sector. But the quantum of government funding available changes much more slowly, and could well last until the next hype cycle starts.

While I don't completely disagree with you, how would you propose researchers go about the problem?

If anything, machine learning is applied to real world problems these days more than it ever was.

For better or worse, AGI is a hard problem, that's going to take a long time to solve. And we're not going to solve it without exploring what works and what doesn't.

Many concerns in this paper, especially about deep learning being data hungry, limited capacity for transfer, and integrating prior knowledge are addressed by recent papers in meta-learning and learning-to-learn [1,2,3 and many many others] both in the supervised and reinforcement learning contexts.

In the case of meta-reinforcement learning there has been recent work [5] which seems to indicate that this mechanism is very similar to how learning works in the brain.

In fact, my group recently published a paper [4] about learning-to-learn in spiking networks (making it biologically more realistic) and showing that the network learns priors on families of tasks for supervised learning, and learns useful exploration strategies automatically when doing meta-reinforcement learning.

While I don't claim this is the right path to AGI, it's a very promising and new direction in deep learning research which this paper seems to ignore.

[1] http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.5.32...

[2] https://arxiv.org/abs/1611.05763

[3] https://arxiv.org/abs/1703.03400

[4] https://arxiv.org/abs/1803.09574

[5] https://www.nature.com/articles/s41593-018-0147-8 (Preprint: https://www.biorxiv.org/content/early/2018/04/06/295964)

I just attempted to try GitKraken, and while it looks quite good on paper, I find it perplexing (and a deal breaker) that it asks me to login (with Github or Gitkraken) before using the open source and free desktop client.

As far as I know, ArchLinux is the only distribution where python points to python3 (and IIRC the change predated PEP394)

Star Trek (the various series) certainly did a good job communicating the vastness of space to the viewer.

Very cool. It's interesting how powerful the recurrent network becomes with the addition of the learned hebbian term. For context, even without the Hebbian term, recurrent networks can learn to learn to do quite interesting things (Hochreiter et al. 2001).

Shameless plug -- our lab recently ported LSTMs to spiking networks without a significant loss in performance, and showed that learning to learn works quite well even with spiking networks (Bellec et al. 2018).

So it seems like this method of learning to learn could provide a extremely biologically realistic and fundamental paradigm for fast learning. The addition of the Hebbian term neatly fits in with this paradigm too.

Hochreiter et al. 2001: http://link.springer.com/chapter/10.1007/3-540-44668-0_13

Bellec et al. 2018: https://arxiv.org/abs/1803.09574

Agreed. There is a very condescending implication that in "developing countries" everything universally is shitty.

I implemented a library that does exactly this -- https://github.com/IGITUGraz/SimRecorder (In case anyone finds it useful). It supports storing data in both hdf5 and redis (although I wouldn't recommend redis for storing large numpy arrays)

There was one theoretical paper to enable learning more complex distributions in the encoding layer(s) while still using a factorized gaussian or similarly simple distribution to derive the learning rules [1]. This approach should, in theory, lead to generation of sharper images, although that particular paper doesn't show examples of it.

[1] https://arxiv.org/abs/1505.05770