Incoming links: Deep learning
Check out: OurBigBook.com, the best way to publish your scientific knowledge. It's an open source note taking system that can publish from lightweight markup files in your computer both to a multi-user mind melding dynamic website, or as a static website. It's like Wikipedia + GitHub + Stack Overflow + Obsidian mashed up. Source code: github.com/ourbigbook/ourbigbook.
Sponsor me to work on this project: 100k USD = I quit me job and work on it one year full time. Status: ~144k / 200k USD reached: 1st year locked-in, 2nd year stretch goal open at 200k USD. 1M USD = I retire and do it forever. How to donate: Section "Sponsor Ciro Santilli's work on OurBigBook.com".
I reached 100k USD after a 1000 Monero donation, so I quit my job for 1 year starting 1st June 2024 to solve as many STEM courses as I can from a world leading university to try and kickstart The Higher Education Revolution. If I reach 200k USD, then I'll do it for two years instead. A second year greatly improve chances of success: year one I solve a bunch of courses, year two I come guns blazing with the content and expand further.
Mission: to live in a world where you can learn university-level mathematics, physics, chemistry, biology and engineering from perfect free open source books that anyone can write to get famous. More rationale: Section "OurBigBook.com"
Explaining things is my superpower, e.g. I was top user #39 on Stack Overflow in 2023[ref][ref] and I have a few 1k+ star educational GitHub repositories[ref][ref][ref][ref]. Now I want to bring that level of awesomeness to masters level Mathematics and Physics. But I can't do it alone! So I created OurBigBook.com to allow everyone to work together towards the perfect book of everything.
My life's goal is to bring hardcore university-level STEM open educational content to all ages. Sponsor me at github.com/sponsors/cirosantilli starting from 1$/month so I can work full time on it. Further information: Section "Sponsor Ciro Santilli's work on OurBigBook.com". Achieving what I call "free gifted education" is my Nirvana.
This website is written in OurBigBook Markup, and it is published on both cirosantilli.com (static website) and outbigbook.om/cirosantilli (multi-user OurBigBook Web instance). Its source code is located at: github.com/cirosantilli/cirosantilli.github.io and also at
cirosantilli.com/_dir and it is licensed under CC BY-SA 4.0 unless otherwise noted.To contact Ciro, see: Section "How to contact Ciro Santilli". He likes to talk with random people of the Internet.
GitHub | Stack Overflow | LinkedIn | YouTube | Twitter | Wikipedia | Zhihu 知乎 | Weibo 微博 | Other accounts
Besides that, I'm also a freedom of speech slacktivist and recreational cyclist. I like Chinese traditional music and classic Brazilian pop. Opinions are my own, but they could be yours too. Tax the rich.
Let's create an educational system with:
- no distinction between university and high school, students just go as fast as they can to what they really want without stupid university entry exams
- fully open source learning material
- on-demand examinations that anyone can easily take without prerequisites
- granular entry selection only for space in specific laboratories or participation in specific novel research projects
I offer:
- online private tutoring for:
- any STEM university course
- passionate younger STEM students (any age) who want to learn university level material and beyond. Can your kid be the next Fields Medalist or Nobel Prize winner? I'm here to help, especially if you are filthy rich! I focus moving students forward as fast as they want on and on producing useful novel tutorials and results
Let your child be my Emile, and me be their Adolfo Amidei, and let's see how far they can go! I will help take your child:and achieve their ambitious STEM goals!- into the best universities
- into the best PhD programs
- educational consulting for institutions looking to improve their STEM courses
- do you know that course or teacher that consistently gets bad reviews every year? I'll work with the teacher to turn the problem around!
- are you looking to create a consistent open educational resources offering to increase your institutions internationally visibility? I can help with that too.
My approach is to:For minors, parents are welcome to join video calls, and all interactions with the student will be recorded and made available to parents.
- propose interesting research projects. The starting point is always deciding the end goal: Section "Backward design"
- learn what is needed to do the project together with the student(s)
- publish any novel results or tutorials/tools produced freely licensed online, and encourage the student to do the same (Section "Let students learn by teaching", digital garden)
I have a proven track of explaining complex concepts in an interesting and useful way. I work for the learner. Teaching statement at: Section "How to teach". Pricing to be discussed. Contact details at: Section "How to contact Ciro Santilli".
I am particularly excited about pointing people to the potential next big things, my top picks these days are:I am also generally interested in:
- 20th century physics, notably AMO and condensed matter
- the history of science, and in particular trying to look at seminal papers of a field
Ciro Santilli's amazing Stack Overflow profile
. Ciro contributes almost exclusively by answering question he Googles into out of his own need, and never by refreshing the newest question of big tags for low hanging fruit! More information at: Section "Ciro Santilli's Stack Overflow contributions".
Ciro Santilli's Stack Overflow stats
. Further methodology details at: Figure "Ciro Santilli's Stack Overflow stats".Introduction to the OurBigBook Project
. Source. OurBigBook Web topics demo
. Source. The OurBigBook topic feature allows users to "merge their minds" in a "sort by upvote"-stack overflow-like manner for each subject. This is the killer feature of OurBigBook Web. More information at: docs.ourbigbook.com/ourbigbook-web-topics.OurBigBook dynamic article tree demo
. Source. The OurBigBook dynamic tree feature allows any of your headers to be the toplevel h1 header of a page, while still displaying its descendants. SEO loves this, and it also allows users to always get their content on the correct granularity. More information at: docs.ourbigbook.com/ourbigbook-web-dynamic-article-tree.OurBigBook local editing and publishing demo
. Source. With OurBigBook you can store your content as plaintext files in a Lightweight markup, and then publish that to either OurBigBook.com to get awesome multi-user features, or as a static website where you are in full control. More information at: docs.ourbigbook.com/publish-your-content.Top Down 2D continuous game with Urho3D C++ SDL and Box2D for Reinforcement learning by Ciro Santilli (2018)
Source. More information: Section "Ciro's 2D reinforcement learning games". This is Ciro's underwhelming stab at the fundamental question: Can AGI be trained in simulations?. This project could be taken much further.
My Bitcoin inscription museum by Ciro Santilli
. Source. Introductory video to Section "Cool data embedded in the Bitcoin blockchain". -------------------------------------
| Force of Will 3 U U |
| --------------------------------- |
| | //////////// | |
| | ////() ()\////\ | |
| | ///_\ (--) \///\ | |
| | ) //// \_____///\\ | |
| | ) \ / / / / | |
| | ) / \ | | / _/ | |
| | ) \ ( ( / / / / \ | |
| | / ) ( ) / ( )/( ) \ | |
| | \(_)/(_)/ /UUUU \ \\\/ | | |
| .---------------------------------. |
| Interrupt |
| ,---------------------------------, |
| | You may pay 1 life and remove a | |
| | blue card in your hand from the | |
| | game instead of paying Force of | |
| | Will's casting cost. Effects | |
| | that prevent or redirect damage | |
| | cannot be used to counter this | |
| | loss of life. | |
| | Counter target spell. | |
| `---------------------------------` |
| l
| Illus. Terese Nelsen |
-------------------------------------Code 1. .
Artist unknown, uploaded December 2014. Part of Section "Cool data embedded in the Bitcoin blockchain" where Ciro Santilli maintains a curated list of such interesting inscriptions.
This was a small project done by Ciro for artistic purposes that received some attention due to the incredible hype surrounding cryptocurrencies at the time. Ciro Santilli's views on cryptocurrencies are summarized at: Section "Are cryptocurrencies useful?".
YellowRobot.jpgJPG image fully embedded in the Bitcoin blockchain depicting some kind of cut material art depicting a yellow robot, inscribed on January 29, 2017.
Ciro Santilli found this image and others during his research for Section "Cool data embedded in the Bitcoin blockchain" by searching for image fingerprints on every transaction payload of the blockchain with a script.
The image was uploaded by EMBII, co-creator of the AtomSea & EMBII upload mechanism, which was responsible for a large part of the image inscriptions in the Bitcoin blockchain.
The associated message reads:This is one of Ciro Santilli's favorite AtomSea & EMBII uploads, as it perfectly encapsules the "medium as an art form" approach to blockchain art, where even non-novel works can be recontextualized into something interesting, here depicting an opposition between the ephemeral and the immutable.
Chiharu [EMBII's Japanese wife] and I found this little yellow robot while exploring Chicago. It will be covered by tar or eventually removed but this tribute will remain. N 41.880778 E -87.629210
At twitter.com/EMBII4U/status/1615389973343268871 EMBII announced that he would be giving off shares of that image on Sup!?, a Bitcoin-backed NFT system he was; making. In December 2023, he gave some shares of the robot to Ciro Santilli.
2010 Wayback Machine archive of starwarsweb.net
. This website was used as one of the CIA 2010 covert communication websites, a covert system the CIA used to communicate with its assets. More details at: Section "CIA 2010 covert communication websites".
Ciro Santilli had some naughty OSINT fun finding some of the websites of this defunct network in 2023 after he heard about the 2022 Reuters report on the matter, which for the first time gave away 7 concrete websites out of a claimed 885 total found. As of November 2023, Ciro had found about 350 of them.
2010 Wayback Machine archive of noticiasmusica.net
. This is another website that was used as one of the CIA 2010 covert communication websites. This website is written in Brazilian Portuguese, and therefore suggests that the CIA had assets in Brazil at the time, and thus was spying on a "fellow democracy".
Although Snowden's revelations made it extremely obvious to the world that the USA spies upon everyone outside of the Five Eyes, including fellow democracies, it is rare to have such a direct a concrete proof of it visible live right on the Wayback Machine. Other targeted democracies include France, Germany, Italy and Spain. More details at: USA spying on its own allies.
Diagram of the fundamental theorem on homomorphisms by Ciro Santilli (2020)
Shows the relationship between group homomorphisms and normal subgroups.
Used in the Stack Exchange answer to What is the intuition behind normal subgroups? One of Section "The best articles by Ciro Santilli".
Spacetime diagram illustrating how faster-than-light travel implies time travel by Ciro Santilli (2021)
Used in the Stack Exchange ansewr to Does faster than light travel imply travelling back in time?. One of Section "The best articles by Ciro Santilli".
Average insertion time into heaps, binary search tree and hash maps of the C++ standard library by Ciro Santilli (2015)
Source. Used in the Stack Overflow answer to Heap vs Binary Search Tree (BST). One of Section "The best articles by Ciro Santilli".
Top view of an open Oxford Nanopore MinION
. Source. This is Ciro Santilli's hand on the Wikipedia article: en.wikipedia.org/wiki/Oxford_Nanopore_Technologies. He put it there after working a bit on Section "How to use an Oxford Nanopore MinION to extract DNA from river water and determine which bacteria live in it" :-) And he would love to document more experiments like that one Section "Videos of all key physics experiments", but opportunities are extremely rare.
A quick 2D continuous AI game prototype for reinforcement learning written in Matter.js, you can view it on a separate page at cirosantilli.com/_raw/js/matterjs/examples.html#top-down-asdw-fixed-viewport. This is a for-fun-only prototype for Ciro's 2D reinforcement learning games, C++ or maybe Python (for the deep learning ecosystem) seems inevitable for a serious version of such a project. But it is cute how much you can do with a few lines of Matter.js!
HTML snippet:
<iframe src="_raw/js/matterjs/examples.html#top-down-asdw-fixed-viewport" width="1000" height="850"></iframe>Generative adversarial network illustrates well AI brittleness. The input looks obvious for a human, but gets completely misclassified by a deep learning agent.
Some of the earlier computers of the 20th centure were analog computers, not digital.
At some point analog died however, and "computer" basically by default started meaning just "digital computer".
As of the 2010's and forward, with the limit of Moore's law and the rise of machine learning, people have started looking again into analog computing as a possile way forward. A key insight is that huge floating point precision is not that crucial in many deep learning applications, e.g. many new digital designs have tried 16-bit floating point as opposed to the more traditional 32-bit minium. Some papers are even looking into 8-bit: dl.acm.org/doi/10.5555/3327757.3327866
As an example, the Lightmatter company was trying to implement silicon photonics-based matrix multiplication.
A general intuition behind this type of development is that the human brain, the holy grail of machine learning, is itself an analog computer.
These come with pre-installed drivers, so e.g. nvidia-smi just works on them out of the box, tested on g5.xlarge which has an Nvidia A10G GPU. Good choice as a starting point for deep learning experiments.
Conda is like pip, except that it also manages shared library dependencies, including providing prebuilts.
This has made Conda very popular in the deep learning community around 2020, where using Python frontends like PyTorch to configure faster precompiled backends was extremely common.
It also means that it is a full package manager and extremely overbloated and blows up all the time. People should just use Docker instead for that kind of stuff: www.reddit.com/r/learnmachinelearning/comments/kd88p8/comment/keco07k/
You also have to buy a license to use their repos if you are part of a large-enough organization: stackoverflow.com/questions/74762863/are-conda-miniconda-and-anaconda-free-to-use-and-open-source
As of 2023, apparently does not use deep learning nor GPUs:
A pair of Austrailan deep learning training provider/consuntants that have produced a lot of good free learning materials:Authors:
- twitter.com/jeremyphoward Jeremy Howard
- twitter.com/math_rachel Rachel Thomas
As of 2020's and earlier, humans were far far behind. As of 2020s and earlier, even an average personal computers without a GPU, the hallmark of deep learning beats every human.
Chess is just too easy!
This point is beautifully argued in lots of different sources, and is clearly a pillar of AGI.
Perhaps one may argue that our deep learning layers do form some kind of hierarchy, e.g. this is very clear in certain models such as convolutional neural network. But many of those models cannot have arbitrarily deep hierarchies, which appears to be a fundamental aspect of intelligence.
How to Create a Mind:
The lists of steps in my mind are organized in hierarchies. I follow a routine procedure before going to sleep. The first step is to brush my teeth. But this action is in turn broken into a smaller series of steps, the first of which is to put toothpaste on the toothbrush. That step in turn is made up of yet smaller steps, such as finding the toothpaste, removing the cap, and so on. The step of finding the toothpaste also has steps, the first of which is to open the bathroom cabinet. That step in turn requires steps, the first of which is to grab the outside of the cabinet door. This nesting actually continues down to a very fine grain of movements, so that there are literally thousands of little actions constituting my nighttime routine. Although I may have difficulty remembering details of a walk I took just a few hours ago, I have no difficulty recalling all of these many steps in preparing for bed - so much so that I am able to think about other things while I go through these procedures. It is important to point out that this list is not stored as one long list of thousands of steps - rather, each of our routine procedures is remembered as an elaborate hierarchy of nested activities.
Human Compatible: TODO get exact quote. It was something along: life goal: save world from hunger. Subgoal: apply for some grant. Sub-sub-goal: eat, sleep, take shower. Sub-sub-sub-goal: move muscles to get me to table and open a can.
Related to Leela Zero, a Go engine
mlcommons.org/en/ Their homepage is not amazingly organized, but it does the job.
Benchmark focused on deep learning. It has two parts:Furthermore, a specific network model is specified for each benchmark in the closed category: so it goes beyond just specifying the dataset.
Results can be seen e.g. at:
- training: mlcommons.org/en/training-normal-21/
- inference: mlcommons.org/en/inference-datacenter-21/
And there are also separate repositories for each:
E.g. on mlcommons.org/en/training-normal-21/ we can see what the the benchmarks are:
| Dataset | Model |
|---|---|
| ImageNet | ResNet |
| KiTS19 | 3D U-Net |
| OpenImages | RetinaNet |
| COCO dataset | Mask R-CNN |
| LibriSpeech | RNN-T |
| Wikipedia | BERT |
| 1TB Clickthrough | DLRM |
| Go | MiniGo |
Version of TensorFlow with a Cirq backend that can run in either quantum computers or classical computer simulations, with the goal of potentially speeding up deep learning applications on a quantum computer some day.