DeepMind Lab2D vs gvgai Updated +Created
At twitter.com/togelius/status/1328404390114435072 called out on DeepMind Lab2D for not giving them credit on prior work!
This very much looks like like GVGAI which was first released in 2014, been used in dozens (maybe hundreds) of papers, and for which one of the original developers was Tom Schaul at DeepMind...
As seen from web.archive.org/web/20220331022932/http://gvgai.net/ though, DeepMind sponsored them at some point.
DeepMind project Updated +Created
Deepmind soccer simulation Updated +Created
  • From Motor Control to Team Play in Simulated Humanoid Football
Video 1.
From Motor Control to Team Play in Simulated Humanoid Football by Ali Eslami (2023)
Source. Likely a reupload by DeepMind employee: www.linkedin.com/in/smalieslami.
Video 2.
DeepMind’s AI Trained For 5 Years by Two Minute Papers (2023)
Source. The 5 years bullshit is of course in-game time clickbait, they simulate 1000x faster than realtime.
Military-industrial complex Updated +Created
This is notably what the United States emerged to be after World War II. But it was likely what Nazi Germany also was, and many other superpowers.
Ciro Santilli feels that much more relevant would be to also include academia as in "military-industrial-academic" complex, the Wikipedia page actually mentions precedents to this idea.
The addition of congress/politicians is also relevant.
But hey, the name wouldn't sound so slick with three parts.
It is basically in this context that American science and technology flourished after World War II, including notably the development of quantum electrodynamics, Richard Feynman being a prototypical example, having previously worked on the Manhattan Project.
Millennium Prize Problems Updated +Created
Ciro Santilli would like to fully understand the statements and motivations of each the problems!
Easy to understand the motivation:
Hard to understand the motivation!
MuJoCo getting started Updated +Created
Tested on Ubuntu 23.10;
git clone https://github.com/google-deepmind/mujoco
cd mujoco
git checkout 5d46c39529819d1b31249e249ca399f306a108ac
mkdir -p build
cd build
cmake ..
make -j
Now let's play. Minimal interactive UI simulation of a simple MJCF scene with one falling cube:
bin/basic ../doc/_static/hello.xml
Test soure code: github.com/google-deepmind/mujoco/blob/5d46c39529819d1b31249e249ca399f306a108ac/sample/basic.cc. The only thing you can do is rotate the scene with the computer mouse it seems. Mentioned at: mujoco.readthedocs.io/en/2.2.2/programming.html#sabasic
Some more interesting models can be found under the model/ directory: github.com/google-deepmind/mujoco/tree/5d46c39529819d1b31249e249ca399f306a108ac/model E.g. the imaginary humanoid robot DeepMind used in many demos can be seen with:
bin/basic ../model/humanoid/humanoid.xml
A very cool thing about that UI is that you can manually control joints. There are no joints in the hello.xml, but e.g. with the humanoid model:
bin/simulate ../model/humanoid/humanoid.xml
under "Control" you move each joint of the robot separately which is quite cool.
Video 1.
Demo of MuJoCo's built-in simulate viewer by Yuval Tassa (2019)
Source.
There's also a bin/record test executable that presumably renders the simulation directly to a file:
bin/record ../doc/_static/hello.xml 5 60 rgb.out
ffmpeg -f rawvideo -pixel_format rgb24 -video_size 800x800 -framerate 60 -i rgb.out -vf "vflip" video.mp4
Mentioned at: mujoco.readthedocs.io/en/2.2.2/programming.html#sarecord but TODO that produced a broken video, related issues:
Nintendo 64 Updated +Created
This is the one that hit Ciro Santilli the hardest, coming in at the point in which he started to discern between games and the real world a little better. His parents bought it for him during a trip to Disney World in Florida in 1996 (?), since electronics were much cheaper in the USA.
So as Ciro became older, and turned into a software engineer, he started to become more and more morbidly curious about "N64 internals": tool-assisted speedrun, how the devkit looks like, how games were developed for it, hardware leaks, etc.
Luckily Ciro's mind is not interested enough by that useless shit for Ciro to seriously study it himself. But that's what YouTube is for, right? Why do useless stuff when other more useless people can do it for you?
The console has only 4 MB of RAM memory. It is quite incredible what can be done with 8 MB, from the point of view of a 2020 worls where 16 GB laptops are the norm.
ImageNet Updated +Created
14 million images with more than 20k categories, typically denoting prominent objects in the image, either common daily objects, or a wild range of animals. About 1 million of them also have bounding boxes for the objects. The images have different sizes, they are not all standardized to a single size like MNIST[ref].
Each image appears to have a single label associated to it. Care must have been taken somehow with categories, since some images contain severl possible objects, e.g. a person and some object.
In practice, the ILSVRC subset of ImageNet is the most commonly used dataset.
Official project page: www.image-net.org/
The data license is restrictive and forbids commercial usage: www.image-net.org/download.php. Also as a result you have to login to download the dataset. Super annoying.
The categories are all part of WordNet, which means that there are several parent/child categories such as dog vs type of dog available. ImageNet1k only appears to have leaf nodes however (i.e. no "dog" label, just specific types of dog).
A major model that performed well on ImageNet starting on 2012 and became notable is AlexNet.
ImageNet subset Updated +Created
Subset generators:
Unfortunately, since ImageNet is a closed standard no one can upload such pre-made subsets, forcing everybody to download the full dataset, in ImageNet1k, which is huge!
Independence (mathematical logic) Updated +Created
A theorem is said to be independent from a set of axioms if it cannot be proven neither true nor false from those axioms.
It or its negation could therefore be arbitrarily added to the set of axioms.
Infinitesimal generator Updated +Created
Elements of a Lie algebra can (should!) be seen a continuous analogue to the generating set of a group in finite groups.
For continuous groups however, we can't have a finite generating set in the strict sense, as a finite set won't ever cover every possible point.
But the generator of a Lie algebra can be finite.
And just like in finite groups, where you can specify the full group by specifying only the relationships between generating elements, in the Lie algebra you can almost specify the full group by specifying the relationships between the elements of a generator of the Lie algebra.
This "specification of a relation" is done by defining the Lie bracket.
The reason why the algebra works out well for continuous stuff is that by definition an algebra over a field is a vector space with some extra structure, and we know very well how to make infinitesimal elements in a vector space: just multiply its vectors by a constant that cana be arbitrarily small.
Insanely active poster on Stack Overflow 4chan post (2023-07-03) Updated +Created
archive.ph/Dd3aC web.archive.org/web/20230709141533/https://desuarchive.org/g/thread/94445084/#94448535 desuarchive.org/g/thread/94445084/#94448535
Title reply because they can't Ctrl+F: How Ciro Santilli manages to write so much
Most of the thread went into pro/anti gay trashtalk due to Ciro using Gay Putin at the time on his Stack Overflow profile as a useless way to protest the Russian invasion of Ukraine.
Some comments:
How does this guy manage to be so active on Stack Overflow? I feel like this disgusting avatar is on at least a quarter of all the active posts.
The answers are always pretty good though.
I came across this schizo's github once, but I had forgotten his name
Obviously severe autism. Also racism homophobia Looks like everything is ok if it's Russia/Chinese...
Reply: it is publicly known that Putin is homophobic as fuck and hates that picture. Therefore we use it. If Putin were heterophobic, we'd post him as hetero.
The only new information:
Reminds me of Xah Lee.
Install Conda on Ubuntu Updated +Created
Tested on Ubuntu 20.04:
mkdir -p ~/miniconda3
wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh -O ~/miniconda3/miniconda.sh
bash ~/miniconda3/miniconda.sh -b -u -p ~/miniconda3
rm -rf ~/miniconda3/miniconda.sh
Add to your .bashrc:
PATH="$PATH:$HOME/miniconda3/bin"
and then to use it on a shell e.g. with Python 3.9 create the environment with:
conda create -y -n mytest3.9 python=3.9
and then use it with:
eval "$(command conda 'shell.bash' 'hook' 2> /dev/null)"
conda activate mytest3.9
Now you can use python and pip normally from inside that mytest3.9 environment.
At that time, the exact installer under latest appears to have been: repo.anaconda.com/miniconda/Miniconda3-py311_23.11.0-2-Linux-x86_64.sh
Deep tech Updated +Created
Ciro Santilli is a fan of this late 2010's buzzword.
It basically came about because of the endless stream of useless software startups made since the 2000's by one or two people with no investments with the continued increase in computers and Internet speeds until the great wall was reached.
Deep tech means not one of those. More specifically, it means technologies that require significant investment in expensive materials and laboratory equipment to progress, such as molecular biology technologies and quantum computing.
And it basically comes down to technologies that wrestle with the fundamental laws of physics rather than software data wrangling.
Computers are of course limited by the laws of physics, but those are much hidden by several layers of indirection.
Full visibility, and full control, make computer tasks be tasks that eventually always work out more or less as expected.
The same does not hold true when real Physics is involved.
Physics is brutal.
To start with, you can't even see your system very clearly, and often doing so requires altering its behaviour.
For example, in molecular biology, most great discoveries are made after some new technique is made to be able to observe smaller things.
But you often have to kill your cells to make those observations, which makes it very hard to understand how they work dynamically.
What we would really want would be to track every single protein as it goes about inside the cell. But that is likely an impossible dream.
The same for the brain. If we had observations of every neuron, how long would it take to understand it? Not long, people are really good at reverse engineering things when there is enough information available to do so, see also science is the reverse engineering of nature.
Then, even when you start to see the system, you might have a very hard time controlling it, because it is so fragile. This is basically the case of quantum computing in 2020.
It is for those reasons that deep tech is so exciting.
The next big things will come from deep tech. Failure is always a possibility, and you can't know before you try.
But that's also why its so fun to dare.
Stuff that Ciro Santilli considers "deep tech" as of 2020:
Dense and sparse matrices Updated +Created
A good definition is that the sparse matrix has non-zero entries proportional the number of rows. Therefore this is Big O notation less than something that has non zero entries. Of course, this only makes sense when generalizing to larger and larger matrices, otherwise we could take the constant of proportionality very high for one specific matrix.
Of course, this only makes sense when generalizing to larger and larger matrices, otherwise we could take the constant of proportionality very high for one specific matrix.
Depth of a quantum circuit Updated +Created
This is an important metric, because it takes some time for the quantum operations to propagate, and so the depth of a circuit gives you an idea of how long the coherence time a hardware needs to support a given circuit.
Derivation of the Dirac equation Updated +Created
The Dirac equation can be derived basically "directly" from the Representation theory of the Lorentz group for the spin half representation, this is shown for example at Physics from Symmetry by Jakob Schwichtenberg (2015) 6.3 "Dirac Equation".
The Diract equation is the spacetime symmetry part of the quantum electrodynamics Lagrangian, i.e. is describes how spin half particles behave without interactions. The full quantum electrodynamics Lagrangian can then be reached by adding the internal symmetry.
As mentioned at spin comes naturally when adding relativity to quantum mechanics, this same method allows us to analogously derive the equations for other spin numbers.
Video 1.
Deriving The Dirac equation by Andrew Dotson (2019)
Source.

Unlisted articles are being shown, click here to show only listed articles.