Periodensystem der KI

Jeder kennt das »Periodensystem der Elemente« aus dem Chemieunterricht. Das Periodensystem ist ein intuitiver und schneller »Lego-Baukasten«, der uns unterstützt, komplizierte Zusammenhänge zwischen Bausteinen (Atomen) und Molekülen (Naturstoffe, Steine oder Metalle) intellektuell zu erfassen.

Der amerikanische Informatiker Kristian Hammond hat den Versuch unternommen, eine Lingua Franca für künstliche Intelligenz zu konzipieren. In Anlehnung an die Chemie bezeichnet er sie als »Periodensystem der Künstlichen Intelligenz«.

Das Periodensystem der Künstlichen Intelligenz unterstützt dabei, den Begriff KI auf Geschäftsprozesse abzubilden und ein Verständnis der Elemente aufzubauen – ähnlich wie im Periodensystem der chemischen Elemente. Der Ansatz hilft beim Verständnis und bei der Einschätzung von Marktreife, Aufwänden, benötigtem Maschinentraining sowie Wissen und Erfahrungen der Mitarbeiter.

generative art: flowers

It started with this tweet about someone called Ayliean apparently drawing a plant based upon set rules and rolling a dice.

And because generative art in itself is fascinating I am frequently pulled into such things. Like this dungeon generator or these city maps or generated audio or face generators or buildings and patterns

On the topic of flowers there’s another actual implementation of the above mentioned concept available:

TubeTime and BitSavers

I was pointing to BitSavers before. And I will do it again as it’s a never ending source of joy.

Now some old schematics had been spilled into my feeds that show how logic gates had been implemented with transformers only.

BitSaver brought it up:

And not only BitSaver is on this path of sharing knowledge, also TubeTime is such a nice account to follow and read.

Drawing Transit Maps

Almost exactly 1 year ago I wrote about transit maps. And it seems to be a recurring topic. And rightfully so – it’s an interesting topic.

Along the presentation of a redesigned Singapore transit map, there’s more content to gather on the “Transit Mapping Symposium” website.

The “Transit Mapping Symposium” will take place in Seoul / South-Korea on 20/21st of April 2020 with researchers and designers meeting up.

The Transit Mapping Symposium is a yearly international gathering of transport networks professionals, a unique opportunity to share achievements, challenges and vision.

Our participants and speakers include experts from all fields of the industry:

– Mapmakers
– Network Operators
– Transport Authorities
– Digital Platforms
– Designers

dangerously curious bitcoins

Some things you find on GitHub are more interesting and frightening than others.

This one is both and some more. What is it you ask?

R2 Bitcoin Arbitrager is an automatic arbitrage trading application targeting Bitcoin exchanges.

So it’s buying and selling Bitcoins. And it’s doing this on different markets.
On the topic of arbitrage Wikipedia has something to say:

In economics and finance, arbitrage is the practice of taking advantage of a price difference between two or more markets: striking a combination of matching deals that capitalize upon the imbalance, the profit being the difference between the market prices at which the unit is traded.

For example, an arbitrage opportunity is present when there is the opportunity to instantaneously buy something for a low price and sell it for a higher price.

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

Now this already is the second version of the tool and already 2 years old. See it as some sort of interesting archeological specimem. Please refrain to actually so something harmful with it.

I am writing this down here because apart from it’s obvious horrors this is a good starting point to understand how these computer-trading-systems do work in principle.

Given that an architectural drawing is also included it gives all sorts of starting points to thoughts.

Also. What could possibly go wrong if a tool to buy/sell on actual markets with actual bitcoins is confident enough to include the “maxTargetProfit” configuration option. Effectively setting the top-line of profit you’re going to make!!!111

25.5 kg force mobile power suite

It’s 3.8 kg and delivers 25.5 kg of force. Impressive! And it’s in stores (in Japan).

The “Every Muscle Suit” has a lot going for it. Weighing just 3.8 kilograms, the pneumatic artificial muscle suit is powerful enough to generate up to 25.5 kilogram-force and effectively relieves pressure on users’ backs when performing activities like heavy lifting.

Best of all, its streamlined design conceals an advanced air pressure system that doesn’t require electricity or batteries.

Japan Today

TESLA PowerWall 2 Security Shenanigans

EXECUTIVE SUMMARY

  • GUI wide open.
  • Default password on WiFi and management interface
  • Attacker can cause financial damage to consumer
  • Attacker can dump entire PW Load into the grid at once
  • Attacker can oscilate between CHARGING and DUMPING (microseconds, the poor sub-station!)
  • Attacker can change grid codes.

More here. At least somebody looked into the security and attack potential of these things.

about brains and silicon wafers

Please read this first paragraph and let it settle:

At the core of the BrainScaleS wafer-scale hardware system (see Figure 90) is an uncut wafer built from mixed-signal ASICs [1], named High Input Count Analog Neural Network chips (HICANNs), which provide a highly configurable substrate that physically emulates adaptively spiking neurons and dynamic synapses (Schemmel et al. (2010)Schemmel et al. (2008)).

I’ve highlighted in bold the portion that I want you to think about once more. We are not talking about chips, dies or cut-up wafers.

We are talking about real-size, huge, fully developed wafers filled with logic. For the sole purpose of brain scale neural network research and development…

The Neuromorphic Computing Platform allows neuroscientists and engineers to perform experiments with configurable neuromorphic computing systems. The platform provides two complementary, large-scale neuromorphic systems built in custom hardware at locations in Heidelberg, Germany (the “BrainScaleS” system, also known as the “physical model” or PM system) and Manchester, United Kingdom (the “SpiNNaker” system, also known as the “many core” or MC system). Both systems enable energy-efficient, large-scale neuronal network simulations with simplified spiking neuron models. The BrainScaleS system is based on physical (analogue) emulations of neuron models and offers highly accelerated operation (104 x real time). The SpiNNaker system is based on a digital many-core architecture and provides real-time operation.

https://electronicvisions.github.io/hbp-sp9-guidebook/index.html

time/space synchronization symbols, AGC training preamble, Viterbi detection/equalization, LDPC decoding and MIMO

Of course this post is talking about hard disks. The ones with spinning disks and read/write heads flying very close to the spinning disks surface.

There are several links to the source papers and works discussing the findings – take look into this nice rabbit hole:

Drag and drop ML with transparency

The machine-learning tooling is getting better. Take a look at Perceptilabs:

Fast modeling
With our drag and drop GUI we enable fast model development.

Increased transparency
The statistical dashboard increases the model’s transparency during training.
Get a better understanding of your model with instant feedback on the operations outputs.
We enable fast error debugging with our custom code editor.

Flexibility
Full flexible options for plugins and importing. Execute any custom Python code in our code editor.

Calling Bullshit

The world is awash in bullshit. Politicians are unconstrained by facts. Science is conducted by press release. Higher education rewards bullshit over analytic thought. Startup culture elevates bullshit to high art. Advertisers wink conspiratorially and invite us to join them in seeing through all the bullshit — and take advantage of our lowered guard to bombard us with bullshit of the second order. The majority of administrative activity, whether in private business or the public sphere, seems to be little more than a sophisticated exercise in the combinatorial reassembly of bullshit.

We’re sick of it. It’s time to do something, and as educators, one constructive thing we know how to do is to teach people. So, the aim of this course is to help students navigate the bullshit-rich modern environment by identifying bullshit, seeing through it, and combating it with effective analysis and argument.

https://www.callingbullshit.org/index.html

Shepard’s Pi

“Shepard’s Pi” is one continous song that lasts for 999,999,999 hours, or about 114 years.

Canton Becker’s music generating algorithm composed this music using the first one billion digits of pi (p). Each digit (3.1415…) determines four seconds of music, supplying the “turn signals” used to determine every musical expression.

Because the numbers in pi never repeat, each of the million hours of “Shepard’s Pi” music are in fact unique. By fast forwarding to some distant moment in the song, you are virtually guaranteed to find yourself listening to something that nobody else – including Canton himself – has ever heard before.

Shepard’s Pi

turn an Xbox 360 HD-DVD drive into an Fluorescent Scanning Thermal Microscope (FSTM)

Curtesy of Sam Zeloof I came around the fact that I’ve got a good part of a FSTM in a cupboard here.

Apparently my choice of purchasing the HD-DVD drive for the Xbox 360 will ultimately pay off!! As we all know Bluray won that format war back in the days.

But now it seems that this below would be useable for something:

Over the life of nuclear fuel, inhomogeneous structures develop, negatively impacting thermal properties. New fuels are under development, but require more accurate knowledge of how the properties change to model performance and determine safe operational conditions.

Measurement systems capable of small–scale, pointwise thermal property measurements and low cost are necessary to measure these properties and integrate into hot cells where electronics are likely to fail during fuel investigation. This project develops a cheaper, smaller, and easily replaceable Fluorescent Scanning Thermal Microscope (FSTM) using the blue laser and focusing circuitry from an Xbox HD-DVD player.

The Design, Construction, and Thermal Diffusivity Measurements of the Fluorescent Scanning Thermal Microscope (FSTM)

As mentioned, Sam Zeloof shows off the actual chip in more detail:

Xbox 360 HD DVD player photodiode chip reverse engineering, includes 49 bits of antifuse trimming from the factory

Making a RISC-V operating system using Rust

As RISC-V progressively challenges the existing ARM processor ecosystem it’s interesting to see more and more software projects popping up that aim that RISC-V architecture.

Here’s one project that aims to develop (and explain along the way) how to create an operating system from scratch. On top of the RISC-V specifics this tutorial also aims to teach how this all can be done in a programming language called Rust.

Keep in mind that all of this is done on a baremetal system. No other software is running.

RISC-V (“risk five”) and the Rust programming language both start with an R, so naturally they fit together. In this blog, we will write an operating system targeting the RISC-V architecture in Rust (mostly). If you have a sane development environment for RISC-V, you can skip the setup parts right to bootloading. Otherwise, it’ll be fairly difficult to get started.

This tutorial will progressively build an operating system from start to something that you can show your friends or parents — if they’re significantly young enough. Since I’m rather new at this I decided to make it a “feature” that each blog post will mature as time goes on. More details will be added and some will be clarified. I look forward to hearing from you!

The Adventures of OS

a 1980s style computer built today for fun

Can you display VGA and play audio on a Cortex-M4 in pure Rust? The short answer is yes, yes you can! Minus the hand-unrolled assembler loop for fixing the phase error in the RGB output. But we don’t talk about that in polite company.

Monotron project page

What currently is in place:

  • The Atari Joystick interface works, but two Joysticks would be more fun
  • The PS/2 Keyboard via the Atmega works, but the pinout was mirrored so you have to put the connector under the PCB :/
  • The RTC works
  • VGA Output works
  • The MIDI Out seems to work when looped to MIDI In, as does the MIDI Though.
  • The MIDI In seems to receive data when connected to my electronic drum kit
  • The Audio output seems to work quite nicely
  • The SD card works, but the power supply can’t handle hot-insertion of the SD card and it makes the TM4C reboot. More capacitors / some current limiting probably required.

I can load games and programs from the SD card into the 24 KiB of free Application RAM. You can interact with these games via the PS/2 Keyboard and Joystick. I can play simple games (like Snake) and play three channels of 8-bit wavetable audio simultaneously. I’ve even got a 6502 Emulator running a copy of 6502 Enhanced BASIC, if you want to go old school!

Hack-The-Planet Podcast: Episode 009

using AI to generate human faces from emojis and thumbnails

Back in March 2019 we’d already seen artificial people. Yawn.

Back then a Generative adversarial network (GAN) was used to produce random human faces from scratch. It synthesized human faces out of randomness.

Now take it a step further and input actual small images. Like thumbnails or emojis or else.

And what do you get?

Quite impressive, eh? There’s more after the jump.

Oh and they wrote a paper about it: Progressive Face Super-Resolution via Attention to Facial Landmark

more blacker

A month ago I wrote about a very black paint. This month brings me a papepr about an even blacker substance.

The synergistically incorporated CNT–metal hierarchical architectures offer record-high broadband optical absorption with excellent electrical and structural properties as well as industrial-scale producibility.

Paper: Breakdown of Native Oxide Enables Multifunctional, Free-Form Carbon Nanotube–Metal Hierarchical Architectures

DIY Lightboard

Usually when we visited lectures the notes and explanations where given on a chalk board or a projector. With the lecturer looking away from the audience most of the time.

This is where Light Boards come in handy. They allow the lecturer to face his audience and give explanations on a board…

Like so:

I was made aways by Ryan Heffernans tweet on the project he did together with his son. He built one of these light boards! The short clip above shows his son on their board.

My son and I built a lightboard. You write on it like a whiteboard, but you can face your audience and the writing is illuminated. Commercial versions cost around $10k, but we made ours for $400 in parts from Home Depot. Here’s how.

Ryan Heffernan Tweet

Of course there are commercial ready-made ones. But where’s the fun in that?

receiving data from deep space with 159 bytes per second

To remind you of the recently celebrated 42nd mission anniversary of the still active and data transmitting Voyager 2 space craft.

NASA’s Voyager 2 is the second spacecraft to enter interstellar space. On Dec. 10, 2018, the spacecraft joined its twin—Voyager 1—as the only human-made objects to enter the space between the stars

Voyager 2 Homepage

And what reminded me of this astonishing achievement.

Think of this: You are flying at >34k miles per hour. You are >18.5 billion miles away from earth. (It’ll take >16 hours at light speed one-way trip from earth to you). And on top, you are still able to send data back to earth at 159 bytes per second.

Mind. Blowing.

Wave Function Collapse

I’ve written on this topic before here. And as developers venture more into these generative algorithms it’s all that more fun to see even the intermediate results.

Oskar Stålberg writes about his little experiments and bigger libraries on Twitter. The above short demonstration was created by him.

Especially worth a look is the library he made available on GitHub: mxgmn/WaveFunctionCollapse.

Some more context, of questionable helpfulness:

In quantum mechanicswave function collapse occurs when a wave function—initially in a superposition of several eigenstates—reduces to a single eigenstate due to interaction with the external world. This interaction is called an “observation”. It is the essence of a measurement in quantum mechanics which connects the wave function with classical observables like position and momentum. Collapse is one of two processes by which quantum systems evolve in time; the other is the continuous evolution via the Schrödinger equation. Collapse is a black box for a thermodynamically irreversible interaction with a classical environment. Calculations of quantum decoherence predict apparent wave function collapse when a superposition forms between the quantum system’s states and the environment’s states. Significantly, the combined wave function of the system and environment continue to obey the Schrödinger equation.

Wikipedia: WFC

Right. Well. Told you. Here are some nice graphics of this applied to calm you:

tickling the dragons tail – the party game?

The demon core was a spherical 6.2-kilogram (14 lb) subcritical mass of plutonium 89 millimetres (3.5 in) in diameter, that was involved in two criticality accidents, on August 21, 1945 and May 21, 1946. 

Wikipedia: Demon core

Now you can have fun without the death-risk in the comfort of your home.

Meet the party-core:

If you’re interested in this topic I can recommend a book:

Atomic Accidents: A History of Nuclear Meltdowns and Disasters: From the Ozark Mountains to Fukushima

A “delightfully astute” and “entertaining” history of the mishaps and meltdowns that have marked the path of scientific progress (Kirkus Reviews, starred review).

Functional Threshold Power

I am cycling for fun and for the effect it has on my body and well-being. I do about 30km of cycling every day on average.

After my first stationary trainer broke I bought a new one with the capability to measure wattage and also to apply resistance measured by the watt.

After looking at my average speeds, heart-rates and times on the device I was able to build a quite detailed understanding of the broader picture. What effects my power output and what does not. The effects of nutrition and health to what the body will deliver while being asked the exact same power output curve than the last time.

In a nutshell the numbers tell me that I am usually at a mediocre wattage of 150W constant load doing about 40 km/h average. My reserves usually allow me to go for 1-2 hours without a break doing this.

So far so good. Now I’ve found out from more serious cyclers that there’s something like “Functional Threshold Power“. I do regular have tests at the doctors to check for any heart-rate issues.

Reading about this Functional Threshold Power my curiousity is sparked.

How much could I do? Should I even go for measuring it?