Manoj Nathwani

Building an Air Conditioning unit out of Lego

Thu, 20 Aug 2020 00:00:00 +0100

Like most people, this year I’ve been spending a lot of time indoors. The only catch is that it’s London summertime and no one has air conditioning in their homes. It’s been getting as high as 35° recently and I thought I’d try making myself a cheap air conditioning system to help me out. This project turned out to be pretty easy for once, and a lot of fun!

Hardware

1x 2M Soft Copper Tube Copper Pipe cut into 4x 50cm straight copper pipes used to transfer the temperature from the water to the air
4x Antec 120mm Case Fans For pushing room temperature air through the cold copper pipes
4x Micro Submersible Mini Water Pumps used to pump water through 4x channels of water tubes and through the copper pipes.
1x 400cm PVC Soft Water Pipe split into 8 equal pieces each 50m long and used to transfer water from the tank to the copper pipes and back to the tank again.
1x 4-Channel 5V Relay Module used to power the motors via an external power supply while being controlled by the Arduino.
A LOT of white Lego bricks

How it works

The electronics isn’t super complex, there’s just a lot of cabling going on! The way it works is that all 4x water pumps are held together by a Lego base I made which is then entirely submerged in a bucket of water. Each pump sucks in water and pumps it through the pipe which feeds the water through the copper pipes (thereby transferring the cold temperature to it) before looping back into the bucket of water where it’s mixed back with the original water source. The cooling happens manually by placing ice packs into the bucket to cool the water.

There technically isn’t any water cooling going on here. It’s just a system which pumps water in a loop between a bucket and copper pipes. As the copper rods cool, the fans push room temperature air through them which cool the air and act as air conditioning.

There’s also a 4-channel 5v relay to help add an external power source as the arduino isn’t powerful enough to power all 4x water pumps. It’s also worth noting that this project could be achieved via just one motor, one relay and a single channel of water instead of 4x however you’d need a much more powerful motor and probably a larger copper pipe to push enough water through.

Building the body

This was the funnest part as I didn’t really test or plan this out before building it. I just got a bucket of Lego and started trying to see how I could secure a PC fan using Lego to hold it up straight. Then combining 4x of those together to make a much wider air flow system. What worked out really great was that the copper pipe was just under 1x1x1 Lego block big which meant I could securely feed the copper rod through a Lego hole while maintaining a super sturdy design.

The PVC piping also fed nicely through the gaps in Lego and allowed me to build a frame above the water tank to evenly space out each pipe so that the water is input & output enely through the tank. Having the pipes distributed evenly makes sure warm water is always mixed with the cooler water before being pushed back around the pipes. The original plastic lid to the bucket also snaps nicely onto the Lego to help avoid dust getting into the water system.

Stopping it leaking

I’ve done plenty of hardware projects before, but none where i’m directly shooting water around which gets really messy when you’ve also got a lot of exposed electronics and cabling going on! It’s really worrying when you’re half way through soldering something to find a small puddle of water closeby too.

The big issue I had was that the PCV piping I was using was much larger than the copper tubes, so I had to find a way of securing them together so that water could pass through them without any leaking. It took some time but eventually I figured out a way of using electrical tape to thicken the girth of the copper pipe, feeding the PVC pipe around the copper pipe and then adding a zip tie to tightly maintain a seal so water wouldn’t leak. It took some trial and error but eventually it worked out and there was no leaking.

The final result

It not only works but it works perfectly! It’s able to consistently push out cool air, it’s quiet, cheap and low powered and of-course - it’s made of Lego and so looks badass.

I think the only upgrade or improvement would be to use higher powered fans as these were pretty weak in terms of airflow, I guess that’s what you get from fans specifically designed to quietly cool your computer but i’m still super happy with the end result.

Building a micro:bit gesture controlled lamp

Thu, 19 Sep 2019 00:00:00 +0100

Recently I attended the hacker camp Chaos Communication Camp which really got me motivated to get creative again with building side projects and specifically ones which GLOW because what’s the point of a hardware project if it doesn’t glow? I decided on building myself a lamp using the BBC micro:bit as it’s this super simple microcontroller made for children and comes jam packed with a bunch of sensors and best of all - can be programmed using MicroPython.

Hardware

1x BBC micro:bit used for super simple gesture controls. The programming of the micro:bit is unbelievably simple and has a whole bunch of sensors built in for future upgrades I could possibly make to this project.
1x ATMEGA328 5V 16MHz Arduino (basically a knockoff Arduino Pro Mini) used for controlling the neopixels as the micro:bit cannot run regular Arduino code with third party libraries so a simple arduino was needed to act as an interface between the micro:bit and Neopixels.
8x NeoPixel Sticks & 1x NeoPixel Ring for providing super high quality LEDs to the project. It’s worth noting that I started off using a knockoff NeoPixel Ring on the top and noticed a pretty big difference in the quality to the official Adafruit Pixel Sticks. I’d really recommend sticking to the official Adafruit LEDs for this project, it’s worth it!
1x 3D printed lamp from Thingiverse.

Building the main light bulb

Soldering the main light bulb together took SO MANY attempts to get right. The main issue was that the bulb would be made of 4x columns of LEDs however each column itself was built of 2x NeoPixel Sticks which meant I had to figure out a way of securely joining them together without them snapping in the middle. I eventually figured out that cutting the tips off some old header pins and using those to then solder both ends of the Neopixels together did the trick. I then daisy chained each strip together to act as a single addressable led strip rather than 9x individual components which made things significantly easier when it came to mounting the bulb to the base and driving the LEDs with an arduino.

It’s pretty clear some serious macgyvering was used to connect each strip when daisy chaining the tops and bottoms of bulb together, especially when adding in the Neopixel Ring on the top. The cabling worked out a little funky with 2x Neopixel Sticks going up one side which then connected to the Neopixel Ring at the top which then connected to the remaining 6x Neopixel Sticks grouped into 2x panels to make up the remaining 3x columns.

In the end the structure was super secure and I was very happy with the end result, it does look horribly hacked together however I plan on simply sealing up all the electronics using hot glue to make it a little shock proof from tips and falls.

Macgyvering the base together

Building the base was by far the most time consuming part of the project. This was also the first time I’ve really got my hands dirty wiring up small circuit boards and sharing power between multiple microcontrollers, so it was a lot of fun learning! What’s going on here is that the micro:bit controller is talking via serial (just RX) to the Arduino to tell it when gestures are fired, the Arduino then instructs the Neopixels accordingly.

Thanks to the use of sandwiching all the cabling between copper PCB boards and using header pins as load bearing columns the structure ended up being super stable and secure. I also added a fine layer of white foam paper between each board to avoid contacts wrongly touching while giving it something to pass heat off to when the microcontrollers become warm.

Circuit Diagram

Here’s a diagram showing how the entire thing is wired together. It seems pretty complex due to the funky daisy chaining going on between the NeoPixel Sticks where they zigzag up and down the bulb which cause every-other stick to be inverted. This was totally worth it though as it allowed the entire bulb to be driven by a single input pin on the Arduino which interacted with all LEDs as if they were part of a single LED strip.

3D printing the body

On Thingiverse I found this absolutely beautiful lamp made by brico3d which diffuses the light of LEDs perfectly. It took a whole 20hrs to get the final shell printed to a high enough quality but it was totally worth it. I’ve pinged brico3d €20 for this great design too because it’s by far the best lamp or vase design on Thingiverse.

Finishing off the hardware

Before sealing the electronics into the 3D printed case I made sure that both the micro:bit could still be programmed via USB and the Arduino too via FTDI so that I could make future upgrades easily. The Arduino’s FTDI input also allowed me to easily power the lamp via two header cables.

I also superglued Lego to the base of the lamp and to the micro:bit so that I could maintain a modular design and make changes in the future easily by snapping the electronics on/off the lamp. I also recycled an old USB cable and passed it through the hole of the base which gave the final product a really professional look.

Software

Put simply - the micro:bit is seriously cool. It has a bunch of sensors built in and supports MicroPython, Javascript and regular old Arduino C as languages to control it with. The Micropy API also has a bunch of common libraries built in for working on common projects such as Neopixels which would have worked great however I was having some latency issues with controlling my strip as well as the library being a little too simplistic for my needs. I ended up adding an Arduino in the middle to do the Neopixel heavy lifting while using the micro:bit for detecting the gestures.

Using the micro:bit to detect gestures and send information via the RX serial connector is quite simple:

from microbit import *

uart.init(baudrate=9600, tx = pin0, rx = pin1)

while True:
    if accelerometer.was_gesture("shake"):
        uart.write("1")
        sleep(2000)

The Arduino then picks up on incoming serial information and flips a binary switch deciding on whether to switch the Neopixels on or not. It’s really that simple at the moment. I plan on adding a bunch more logic for having configurable settings which allow the lamp to have multiple lighting modes such as fireplace, rainbow, white etc. You can find all the code for this project here.

The final result

The final product ended up looking awesome! The colours are super bright and vivid and the shell of the lamp really diffuses the Neopixels perfectly. I started off by testing a sample rainbowCycle test from Neopixel library which works great at showing it off. The shake gesture to turn the lamp/on worked out fantastically too thanks to the BBC micro:bit, I still can’t believe it only takes a few lines of code now to implement gestures into Arduino projects now! I also managed to find a nice Arduino sketch for imitating a flickering candle allowing the lamp to be used for subtle mood lighting. It’s a little hard to see on video but it gives my entire bedroom a nice warm glow.

The Neuroscience of Psychedelics

Mon, 18 Mar 2019 00:00:00 +0000

Recently I read an incredible book by Michael Pollan called How to Change Your Mind: The New Science of Psychedelics. It’s split up really nicely into individual chapters on the biology, neuroscience, cultural history and his own personal tale of going from someone who’s never been exposed strong substances to really giving them a shot with an open mind. There’s also a lot said about the pschotheraperic uses of psychedelics in helping patients with accepting death and tackling addiction & depression. I’d strongly recommend reading the book and personally found the chapter on neuroscience really mind blowing, so much so that I thought it’s worth writing notes on and sharing them with others. It’s also worth noting that these notes are 95% copied straight out the book and are not in my own words.

The Biology

The molecules Psilocin, LSD, and 5-MeO-DMT are all types of tryptamines which are found in plants, fungi, and animals where they typically act as signaling molecules between cells.

The most famous tryptamine in the human body is the neurotransmitter serotonin. It binds with a dozen or so different receptors, and these are found not only across many parts of the brain but throughout the body, with a substantial representation in the digestive tract. Depending on the type of receptor in question and its location, serotonin is liable to make very different things happen—sometimes exciting a neuron to fire, other times inhibiting it.

The group of tryptamines we call “the classical psychedelics” have a strong affinity with one particular type of serotonin receptor, called 5-HT2A. Psychedelics resemble serotonin closely enough that they can attach themselves to this receptor site in such a way as to activate it to do various things.

LSD has an even stronger affinity with the 5-HT2A receptor which is “stickier” than serotonin itself, making this an instance where the simulacrum is more convincing, chemically, than the original. The science of serotonin and LSD has been closely intertwined since the 1950s; in fact, it was the discovery that LSD affected consciousness at such infinitesimal doses that helped to advance the new field of neurochemistry in the 1950s, leading to the development of the SSRI antidepressants.

Trying to explain consciousness

What neuroscientists, philosophers and psychologists mean by consciousness is the unmistakable sense we have that we are, or possess, a self that has experiences. Sigmund Freud wrote that “there is nothing of which we are more certain than the feeling of our self, our own ego.” Yet it is difficult to be quite so certain that anyone else possesses consciousness, much less other creatures, because there is no outward physical evidence that consciousness as we experience it exists.

“The hard problem” or the “explanatory gap”: How do you explain mind—the subjective quality of experience—in terms of meat, that is, in terms of the physical structures or chemistry of the brain? The idea that psychedelic drugs might shed some light on the problems of consciousness makes a certain sense. A psychedelic drug is powerful enough to disrupt the system we call normal waking consciousness in ways that may force some of its fundamental properties into view. Someone on a psychedelic remains awake and able to report on what he or she is experiencing in real time.

The Default Mode Network

In 2001 neuroscientists discovered the DMN, the Default Mode Network. The network forms a critical and centrally located hub of brain activity that links parts of the cerebral cortex to deeper (and older) structures involved in memory and emotion.

Scientists noticed that several areas in the brain exhibited heightened activity precisely when their subjects were doing nothing mentally. This was the brain’s “default mode” the network of brain structures that light up with activity when there are no demands on our attention and we have no mental task to perform. Put another way, we had discovered the place where our minds go to wander—to daydream, ruminate, travel in time, reflect on ourselves, and worry.

The default network stands in a kind of seesaw relationship with the attentional networks that wake up whenever the outside world demands our attention; when one is active, the other goes quiet, and vice versa. The default mode is most active when we are engaged in higher-level “metacognitive” processes such as self-reflection, mental time travel, mental constructions (such as the self or ego), moral reasoning, and “theory of mind”—the ability to attribute mental states to others, as when we try to imagine “what it is like” to be someone else.

The default mode network is described as the brain’s “orchestra conductor,” “corporate executive,” or “capital city,” charged with managing and “holding the whole system together.” And with keeping the brain’s unrulier tendencies in check.

The brain consists of several different specialized systems—one for visual processing, another to control motor activity—each doing its own thing. Chaos is averted because all systems are not created equal. Electrical signaling from some brain areas takes precedence over others. At the top of this hierarchy resides the DMN, which acts as an uber-conductor to ensure that the cacophony of competing signals from one system do not interfere with those from another. The default mode network keeps order in a system so complex it might otherwise descend into the anarchy of mental illness.

The achievement of an individual self, a being with a unique past and a trajectory into the future, is one of the glories of human evolution, but it is not without its drawbacks and potential disorders. The price of the sense of an individual identity is a sense of separation from others and nature.

Turning off The Default Mode Network

In studies with psychedelics the steepest drops in default mode network activity correlated with volunteers’ subjective experience of ego dissolution. “I existed only as an idea or concept” one volunteer reported. Another recalled “I didn’t know where I ended and my surroundings began.” The more precipitous the drop-off in blood flow and oxygen consumption in the default network, the more likely a volunteer was to report the loss of a sense of self.

It appears that when activity in the default mode network falls off precipitously, the ego temporarily vanishes, and the usual boundaries we experience between self and world, subject and object, all melt away.

The mystical experience may just be what it feels like when you deactivate the brain’s default mode network. This can be achieved any number of ways: through psychedelics and meditation, but perhaps also by means of certain breathing exercises (like holotropic breathwork), sensory deprivation, fasting, prayer, overwhelming experiences of awe, extreme sports, near-death experiences, and so on.

It happens, taking this particular network offline may give us access to extraordinary states of consciousness - moments of oneness or ecstasy that are no less wondrous for having a physical cause. These and other centers of mental activity are “let off the leash” when the default mode leaves the stage and brain scans show an increase in activity (as reflected by increases in blood flow and oxygen consumption) in several other brain regions, including the limbic regions, under the influence of psychedelics. Material that is unavailable to us during normal waking consciousness now floats to the surface of our awareness, including emotions and memories and, sometimes, long-buried childhood traumas.

Everyday consciousness is a hallucination

The default mode network doesn’t only exert top-down control over material arising from within; it also helps regulate what is let into consciousness from the world outside. It operates as a kind of filter or reducing valve charged with admitting only that measly trickle of information required for us to get through the day. We are forever cutting to the chase, basically, and leaping to conclusions, relying on prior experience to inform current perception.

Our perceptions of the world offer us not a literal transcription of reality but rather a seamless illusion woven from both the data of our senses and the models in our memories. Normal waking consciousness feels perfectly transparent, and yet it is less a window on reality than the product of our imaginations—a kind of controlled hallucination.

How is normal waking consciousness any different from other, seemingly less faithful productions of our imagination - such as dreams or psychotic delusions or psychedelic trips? In fact, all these states of consciousness are imagined: they’re mental constructs that weave together some news of the world with priors of various kinds. In the case of normal waking consciousness, the handshake between the data of our senses and our preconceptions is especially firm because it is subject to a continual process of reality testing.

Your brain on psychedelics

With your DMN knocked offline for a few hours your eager brain tries to make sense of the data rushing in, the brain leaps to erroneous conclusions and sometimes what we call a hallucination results. (The paranoid does much the same thing, ferociously imposing a false narrative on the stream of incoming information.) But in other cases, the reducing valve opens wide to admit lots more information, unedited and sometimes welcome.

People who are colorblind report being able to see certain colors for the first time when on psychedelics and there is research to suggest that people hear music differently under the influence of these drugs. They process the timbre, or coloration, of music more acutely - a dimension of music that conveys emotion.

When confidence in our usual top-down concepts of reality collapses, opening the way for more bottom-up information to get through the filter. When all that sensory information threatens to overwhelm us, the mind furiously generates new concepts (crazy or brilliant, it hardly matters) to make sense of it all and so you might see faces coming out of the rain. That’s the brain doing what the brain does - that is, working to reduce uncertainty by, in effect, telling itself stories.

Suppressing entropy in an aging mind

By adulthood the brain has gotten very good at observing and testing reality and developing reliable predictions about it that optimize our investments of energy (mental and otherwise) and therefore our chances of survival. Uncertainty is a complex brain’s biggest challenge, and predictive coding evolved to help us reduce it. In general, the kind of precooked or conventionalized thinking this adaptation produces serves us well. But only up to a point.

Do we pay a price for the achievement of order and selfhood in the adult human mind? While suppressing entropy (in this context, a synonym for uncertainty) in the brain “serves to promote realism, foresight, careful reflection and an ability to recognize and overcome wishful and paranoid fantasies” at the same time this achievement tends to “constrain cognition” and exert “a limiting or narrowing influence on consciousness.” Research on entropy asks us to conceive of the mind as an uncertainty-reducing machine with a few serious bugs in it.

A better way to suppress uncertainty and entropy in the human brain emerged with the evolution of the default mode network, a brain-regulating system that is absent or undeveloped in lower animals and young children. Along with the default mode network, a coherent sense of self or ego emerges and with that, the human capacity for self-reflection and reason. This new highly evolved mode of cognition is called secondary consciousness. Secondary consciousness pays deference to reality and diligently seeks to represent the world as precisely as possible in order to minimize surprise and uncertainty (i.e. entropy).

At the high-entropy end of the spectrum, there are psychedelic states; infant consciousness; early psychosis; magical thinking; and divergent or creative thinking. At the low-entropy end of the spectrum, there’s narrow or rigid thinking; addiction; obsessive-compulsive disorder; depression; anesthesia; and, finally, coma. The psychological “disorders” at the low-entropy end of the spectrum are not the result of a lack of order in the brain but rather stem from an excess of order. This is perhaps most clearly evident in depression, when the ego turns on itself and uncontrollable introspection gradually shades out reality.

Lubricating cognition with psychedelics

People suffering from a whole range of disorders characterized by excessively rigid patterns of thought - including addiction, obsessions, and eating disorders as well as depression stand to benefit from “the ability of psychedelics to disrupt stereotyped patterns of thought and behavior by disintegrating the patterns of [neural] activity upon which they rest.”

So it may be that some brains could stand to have a little more entropy, not less. This is where psychedelics come in. By quieting the default mode network, these compounds can loosen the ego’s grip on the machinery of the mind, lubricating cognition where before it had been rusted stuck. The grip of an overbearing ego can enforce a rigidity in our thinking that is psychologically destructive. It may be socially and politically destructive too, in that it closes the mind to information and alternative points of view. When the influence of the DMN declines, so does our sense of separateness from our environment.

On Psychedelics the specialized neural networks of the brain such as the default mode network and the visual processing system each become disintegrated while the brain as a whole becomes more integrated as new connections spring up among regions that ordinarily kept mainly to themselves or were linked only via the central hub of the DMN. Put another way, the various networks of the brain became less specialized.

Distinct networks became less distinct under the drug as they communicate more openly with other brain networks. The brain operates with greater flexibility and interconnectedness under hallucinogens. When the memory and emotion centers are allowed to communicate directly with the visual processing centers, it’s possible our wishes and fears, prejudices and emotions, begin to inform what we see. One way to think about this blooming of mental states is that it temporarily boosts the sheer amount of diversity in our mental life.

The effects can be long term

Research on the psychedelic experience show long-term changes in the personality trait of openness raises the possibility that some kind of learning takes place while the brain is rewired and that it might in some way persist. Learning entails the establishment of new neural circuits; these get stronger the more exercise they get. The long-term fate of the novel connections formed during the psychedelic experience—whether they prove durable or evanescent might depend on whether we recall and, in effect, exercise them after the experience ends. This could be as simple as recollecting what we experienced, reinforcing it during the integration process, or using meditation to reenact the altered state of consciousness.

Psychedelic experiences may also facilitate neuroplasticity: it opens a window in which patterns of thought and behavior become more plastic and so easier to change. A high-dose psychedelic experience has the power to “shake the snow globe”, disrupting unhealthy patterns of thought and creating a space of flexibility entropy in which more salubrious patterns and narratives have an opportunity to coalesce as the snow slowly resettles.

One of the most interesting things about a psychedelic experience is that it sharpens one’s sensitivity to one’s own mental states, especially in the days immediately following. The usual seamlessness of consciousness is disturbed in such a way as to make any given state—mind wandering, focused attention, rumination—both more salient and somewhat easier to manipulate.

Psychedelics introduce more noise or entropy into the brain which may shake people out of their usual patterns of thought, lubricate cognition in ways that might enhance well-being, make us more open and boost creativity. Psychedelics could help adults achieve the kind of fluid thinking that is second nature to kids, expanding the space of creative possibility.

As for the unwell, the patients who stand to gain the most are probably those suffering from the kinds of mental disorders characterized by mental rigidity: addiction, depression, obsession. “You get stuck on the same thing, you can’t escape, you become obsessive, perhaps addicted. It seems plausible that the psychedelic experience could help us get out of those states, create an opportunity in which the old stories of who we are might be rewritten.”

The power of Psychedelics to temporarily quieten the default mode network and loosen the grip of the ego may just be simple reset a lot mind needs in an ever changing world.

My 2019 Charity Plendge

Fri, 28 Dec 2018 00:00:00 +0000

The recent GiveWell blog on staff member’s personal donations for 2018 got me thinking on how I really need to get-onboard with publicly stating how I plan on donating my time, money and life choices around making the world a better place. I think stating it and keeping track of it will really help in making sure my goals are met in 2019. So here it goes!

Goals

1) Offset co2 emissions of my lifestyle

2) Decrease animal suffering from my diet

3) Donate 50% of all gross income over £40k to charity

4) Continue working for effective charities

5) Volunteer more often

Why bother?

I’ve never really cared for sharing how much I give to charity however I think stating it publicly and sharing it with others will provide good utility as it’ll allow me to formally get my thoughts on paper and feel socially obligated to follow through with it all. The Life You Can Save by Peter Singer also states:

We need to get over our reluctance to speak openly about the good we do. Silent giving will not change a culture that deems it sensible to spend all your money on yourself and your family, rather than to help those in greater need – even though helping others is likely to bring more fulfilment in the long run.

Research shows that when people know that others are giving, they are themselves more likely to give. So publicly pledging to give will encourage others to give. This holds true for billionaires and for those of us who aren’t anywhere near that level of wealth. We can all make a difference, and play our part in making the world a better place.

1) Offset co2 emissions of my lifestyle

I’m world travelling asshole, flying over 50,000 miles per year via plane and about 200-ish miles via scooter, it’s a bit crazy and a horrible waste of resources. To offset this I will calculate my co2 emissions for every flight and pay a carbon offsetting charity to negate the damage my lifestyle inflicts on the environment. I’ll also track the amount of miles I drive with my scooter and offset those 2x as the fuel i’m using in Bali is presumably low quality and therefore more damaging. Carbon offsetting will be considered a tax and I will not be counting this as charitable money - this is simply an asshole tax i’m giving myself.

2) Decrease animal suffering from my diet

I was brought up vegetarian however i’ve been eating meat for about 8 years now, since i started travelling and wanted an easier time finding food. By spending half of 2018 in Bali it was not only easy, but preferable to be vegan as it was so delicious, so making sure i’m in a country which enables me to be vegan is definitely the way to go.

It’s hard to promise anything as my diet vastly changes based on wherever i’m travelling and so I think i’ll dedicate a large amount of time into reading/researching animal ethics to help motivate me away from meat and dairy. I’ll also pledge to figure out some kind way of financially disincentivizing myself from eating meat and dairy by fining myself every time i consume something i shouldn’t, just like the “asshole tax” for co2 emissions. I’m not sure how much i’ll give, but i’ll make sure to send that money directly to an organisation fighting for animal ethics and exclude that from my “charitable donations” pot when looking at my finances. That way a cheeseburger would cost £50 instead of £5 and it wouldn’t count as a charitable donation.

3) Donate 50% of all gross income over £40k

I tried giving to the most effective charities in 2018 however somewhat lost motivation as it didn’t provide me any warm fuzzies. I think a good motivator will be to simply track everything in a spreadsheet as making it all quantifiable will act as a good mechanism to provide me a clear feedback loop of the good i’m doing. In 2019 I intend on giving 50% of all income above £40k to charity, that’s all income I make, regardless of my costs or expenses. 80% will to go effective charities and the remaining 20% will go to charities or projects close to me where I have an emotional connection to it’s cause.

4) Continue working for effective charities

I’ve been working with rethink charity rebuilding their eahub.org website which has been hugely rewarding. They’ve expressed interest in working on other projects with me too and I love working with them as they provide a great community of like-minded altruistic co-workers. I’m also very well suited for the job, so feeling highly valued is a good motivator.

I’m charging $350/week which is about a 10th my usual price. I’ll stick with that for now and promise at-least 1 week each month. My main issue here it that I need to work on other less-motivating projects on the side to maintain my ability to work for rethink. Ideally I can get rethink to pay me $350+/day which should enable me to 2-3x how much time I dedicate towards them. We’ve only been working together for 2 months though, so i’ll continue to play this by ear.

5) Volunteer more often

I also want to volunteer at at-least one organisation every month or two in a physical way. This doesn’t need to be effective or even at a charity, but the goal here is to emotionally get a nice kick out of doing something good. Mostly for my own enjoyment, meeting new people and still having an altruistic impact on others. This is another point i won’t worry too much about quantifying or making sure i perfectly hit. Helping people is a fun activity, so let’s try doing that more often and see what happens.

Obligatory Altruism

Tue, 25 Sep 2018 00:00:00 +0100

This is a small essay i’m mostly writing for myself to formally put my thoughts on paper on how I feel about obligatory altruism. I also think it’ll hopefully help others in seeing my thought process to avoid falling for the efficiency paradox explained below.

Altruism is a moral obligation

Michael Lewis’s Princeton talk was instrumental to my views on affluence. He explained how successful people mostly owe their success to the random external events out of their control which put them in their position, and the obligation lucky people have to those less fortunate. This ingrained a certain sense of moral obligation which i’ve had for several years now.

Over the last few months i’ve become particularly interested in philosophy and Peter Singer’s views on Famine, Affluence, and Morality where he expresses how “If it is in our power to prevent something bad from happening, without thereby sacrificing anything of comparable moral importance, we ought, morally, to do it.” - which makes perfectly good sense.

The paradox of efficiency

I got trapped in the loop of constantly researching the most efficient thing I could be doing. This makes perfect sense as you would always want to make sure you were being as logical as possible. However the deeper I went, the less I valued my personal feelings, as it felt selfish to give them any weight at all.

Over time I found that the most effective ways I could be contributing didn’t align with the quality of life i’d still want to be living. Basic luxuries such as eating a meal at a restaurant seems unethical as you could have eaten at home and the money saved could have been spent helping to improve the lives of people suffering. Even when you attempt to do good by giving money to an education charity, you find that the money could have been better spent helping twice as many people alleviating suffering in another part of the world. Eventually this leads to a sense of hopelessness as whatever you do isn’t efficient enough.

Optimising on effectiveness makes logical sense. However, it seems as you increase efficiency there is a certain point where your emotional connection with the people you were helping quickly diminishes. I found that changing from helping people in my own country to more effectively helping those abroad didn’t have this effect. However, once I directed funds to social programs which didn’t have a clear way to quantify their impact on humans, my motivation to continue funding such programs quickly diminished.

It seems optimising for the most objective good by ignoring your subjective (somewhat selfish) needs reverses the effect on the very thing you were trying to optimise for.

Optimising for long term success

It’s also worth noting that the reason we must come to such an emotionally compatible sweet spot is that we want to optimise for the most amount of cumulative good we do over our entire lifetimes. This means making sure I continue to be motivated to contribute to given problem areas over the duration of my lifetime, rather than maximising effectiveness today and eventually burning out due to loss of interest. Peter Singer (who pretty much lead me down this entire path) has some good advice on this too: “Ultimately, I don’t think my indulgences can be justified. I know that I’m very far from being a saint. I should spend less on myself and give away more of what I earn. Of course, I give much more than most. But I know that that isn’t the right standard. As for deciding how much is enough, I just do a little better each year.”

Conclusion

I think taking time to consciously reflect on how you plan to have a positive effect on the world is a worthy cause. Being part of a community such as Effective Altruism helps.

Our goal in life should be to aim to optimise on doing the most cumulative good in our lifetimes by alleviating suffering for the most sentient beings.

Fortunate individuals such as myself have the moral obligation to help those who have been unlucky in life. Just as success usually comes from external events outside of one’s control, the same dynamic applies to unfortunate individuals who through no fault of their own suffer unnecessarily in life.

In an ideal world we would all act perfectly rational and utilitarian, however this isn’t the case as we are complex emotional animals.

In order to do the most good we must find the intersection between what we personally find motivating to work on and the most impactful work we can do.

Both these phenomena are constantly in flux and conscious effort should be made to regularly review this to optimise on lifelong impact.

Attaching a Raspberry Pi camera to a drone

Sat, 10 Sep 2016 00:00:00 +0100

Recently I got a Parrot MiniDrone which is a really great piece of kit. The drone already comes with a built in down-facing camera however i’ve been looking for an excuse to do a Raspberry Pi Zero camera project for a while now, so I thought this would be a great chance to put to 2 together. The project was a lot harder than I expected due to the very tight weight limit the drone could hold while still flying stable. This specific version of Parrot MiniDrone has a lego adapter on top which allows you to add up to 30 grams of lego bricks for fun. The goal was to attach a camera on top which would record video from an FPV perspective. Here’s how it went…

Mounting the camera module

I went through a LOT of iterations when trying to find the best way to mount the camera. The main issue was finding a light weight solution to sturdily attach it while having the cable attached to the Rasberry Pi without blocking the sensor on the bottom of the drone. The end solution ended up using some lego pieces with soldering wire holding the camera in position.

Positioning the Raspberry Pi

Positioning the Raspberry Pi took a very long time to get right as this seemed to be the deciding factor on weather the drone could successfully fly without losing balance. The issue was that it had to be center balanced as well as accessible to the camera module cable and lipo battery. I ended up supergluing a piece of lego to the bottom of the Pi and attaching it on-top of the drone.

Powering everything

Powering the Raspberry Pi was pretty straight forward thanks to the Zero LiPo which worked out great! This was the first time i’ve powered anything with lipo batteries and there were a few iterations of figuring out where to attach it and which batteries I could use to not add too much weight to the drone. I ended up using velcro to attach the lipo battery to the main drone battery so I could easily swap out the lipo battery with replacements. I ended up using some cheap 150mAh 20C lipo batteries which I soldered connectors to which lasted about 30mins when continuously recording video.

The final build

Like most projects, the simplest and most elegant solutions seem to take the longest to achieve however are completely worth it. The final solution was to use a single long piece of lego to attach the Pi and camera module to and have a simple power cable leading to the lipo battery on the back. It worked great and a nice unintended benefit of having all the electronics attached to that single piece of lego was that when the drone would crash, the lego would shatter and break gracefully, thereby protecting the electronics by not applying any pressure to them. If this wasn’t the case, crashing the drone would potentially damage the Pi or camera module, which is a lot worse than simply breaking some lego. An issue which was never solved and probably wouldn’t be possible, was the issue with accidentally cutting the camera module cable with the drone blades. I ended up breaking about 3-4 over the 2 weeks I worked on this in my spare time.

Programming the Pi

The code I ran on the Rasberry Pi was very dumb. All I wanted it to do was continuously record video while powered, however when killing the device, the video would corrupt, so I wrote a simple python script to continuously take 10 second videos. This way if I killed the device, worse-case I would lose 10 seconds worth of flight video, which isn’t terrible at all. For this project, I was a lot more interested in the hardware, so I even lazily recorded all video straight into the Pis limited home directory and USB FTP’ed the videos onto my laptop to process after each flight. You can see the code here.

The final result

The final result came out pretty great! What really surprised me was the framerate in which the Raspberry Pi camera module could record video in. I was also impressed with how well the drone could handle the extra weight of all the electronics which ended up weighing just over 35 grams.

Building a huge display of 576 ping-pong balls

Thu, 11 Aug 2016 00:00:00 +0100

As the end of 2019 slowly approaches I thought it’s better late than never to finish writing up about a project I worked on exactly three years ago in 2016 and never got around to publishing. I’ve had the draft of this post sitting on a trello card for three whole years under a list called “To Do (never happening)” so here’s me proving that wrong!

This summer I decided to do a project I’ve been wanting to do for years by building a massive LED grid for a festival installation. EMF Camp 2016 was soon approaching so it gave me the perfect opportunity to have a deadline to work towards and actually build something cool! What my brother and I ended up building was a huge 32x18 pixel display consisting of 576 individually addressable LEDs across 12x LED strips with each LED diffused by a ping-pong ball. It ended up looking pretty insane!

Installing all 576 ping-pong balls

We started off by splitting up our 32x18 display into two 16x8 halves to make it easier to transport and then drilled a hole in the middle of a 5cm/5cm square which would hold each ping-pong ball. After drilling the holes we cut an X shape into each ball before using a hot glue gun to attach it to the main board. Luckily we didn’t need to buy real pong-pong-worthy balls and found some much cheaper “amateur entertainment balls” being sold in bulk on Amazon so we went with those, I think due to the cheap thinner coating of plastic they use it actually diffuses the light from our LEDs much better than regular ping-pong balls, so it worked out better than planned.

Testing the LEDs

I purchased 12x WS2811 LED strips which were specifically waterproof to help protect our project from our famously rainy British summers seeing as the installation would be used outdoors on grass. The slight foggy coating attached to each LED bulb made our ping-pong balls really nicely glow in a gentle and elegant way in the dark too.

Wiring things up

Wiring it all up was great fun. We had to feed each LED into the designated hole for it, use a hot glue gun to glue the LED in place and then hold it for a little moment for it to cool down. It was a pretty quick process and pretty therapeutic. Each strip’s 5V and GROUND inputs were fed onto a copper PCB board so they could all share the same 5V and Ground inputs from a 50W Regulated PSU which would feed the entire grid power. The data inputs and outputs were all daisy chained together which sounded acceptable at the time until everything went wrong…

IT WORKS!

We hooked everything up and ran some basic Neopixel rainbow tests through a NodeMCU and it looked absolutely beautiful! The LEDs were perfectly diffused by the ping-pong balls which did a great job of displaying colours in a really nice elegant way. It looked super super nice.

Things were looking good so we decided to leave it there and pick it back up at the festival where we’d combine both halves together to make the full 32x18 display. Just like every project, testing things at home in a safe environment means it’ll work in the wild for sure.. right?

No wait, it’s entirely broken!

We arrived at the festival and everything seemed to function until we tried running more complex animations on the display and the display would just go wild with colours. It really looked like there was some kind of noise or interference with the data connection. What’s interesting is that we could run a simple colorWipe command through the Neopixel library and it would get about 200-ish LEDs in before the entire grid would lose control and become all fuzzy. Our first thought was that the NodeMCU wasn’t capable of driving so much data through a single port so we swapped it out for a regular Arduino Uno which also seemed to have the exact same issue. We also spent a LOT of hours rewiring every 5V & GROUND connection to each strip trying to make sure nothing was missed and no cables were wrongly crossed.

Things eventually became a lot more difficult to work on as night fell and we were trying to use a soldering iron while sitting in the grass at a dimly lit festival during the build-up process so we decided to call it quits and just roll with that we had. To our surprise everyone still seemed to love it! People were so taken aback by how beautiful it looked that no-one seemed to question what it was actually displaying. In-fact people just assumed it was some kind of data visualisation so we ended up satirically calling it a “Hard Disk Defrag Visualiser” and everyone who came by legitimately thought that’s exactly what it was doing. Little did they know it was just our broken LED matrix displaying random values on each LED because we we never figured out how to fix it!

So what went wrong?

What went wrong is pretty straight forward. We got everything perfectly configured apart from that single green cable which runs through every LED strip. As each LED strip was nearly 4 meters long and we had 12x strips all daisy changed there’s no way the data cable was able to transmit the signal along all 48 meters. That explains why it would work perfectly for the first few strips however once it exceeded about 200 LEDs the entire grid would start lagging and displaying random values. Usually this wouldn’t be an issue if you had higher quality official Neopixel strips however as I was using cheaper LEDs as well as having about 5cm of spacing between each light, it really made the total distance we were sending a signal through much longer through low quality cables.

As it turns out there’s a simple solution to this problem. Either adding a new data cable to the Arduino every few strips or purchasing a specialised LED strip driver which solves this exact problem like the FadeCandy by Adafruit.

Conclusion

Setting aside the issues around not being able to fully program the matrix to display what we wanted this was still a super fun project to work on! There’s something really enjoyable about building a large physical installation for a festival and sharing your creations with others who really take time to appreciate what you’ve built. There’s also something therapeutic about drilling 576 holes in a wooden board, carefully cutting an X onto a ping-pong ball and then feeling an LED into it piece-by-piece. The end result was also really pretty to look at, especially outdoors at night and it really came out much better than we originally imagined. At some point in the future I’ll likely take out the LEDs strips again and get all 48 meters of it to correctly work!

Building a wearable camera

Mon, 09 May 2016 00:00:00 +0100

Recently I worked on a side project where I built a wearable camera out of a Rasperry Pi. I had previously bought a Narrative Clip and was a little disappointed by quality and knew hacking together my own solution could be possible. I was also about to go on a holiday where I would be traveling to 4 countries in 5 days, so the timing was perfect. I started off by punching a hole through the strap on my backpack and sowing the Raspberry Pi Camera Module onto it. The ribbon cable was then wheaved up the strap and fed into the top backpack for the Raspberry Pi. The main reason for this was to protect the naked electronics from rain as well as keeping it tightly attached to my bag so I didn’t have to keep repositioning it like I was having to do with the Narrative Clip.

This ended up working fantastically and it solved both the alignment problems as well as keeping all electronics perfectly waterproof. The image on the right shows my bag completely drenched from heavy rain in Bratislava.

Adding GPS

As the primary use for this camera was to take pictures on holiday, I thought it would be great to add a LinkIt ONE which would stream GPS data to the PI and save it as Exif data to every image it takes. An interesting thing I learnt about GPS hacking was that it’s either 100% absolutely accurate or throws you somewhere in the middle of Morocco. The accuracy really depends on how many satellites the device detects and I eventually decided not to take it abroad as not much time was spent testing it enough to trust it, but i’m pretty sure it would have been fine.

Building the mobile app

One frustration I had with the Narrative Clip was that I had zero control over when it was taking pictures; it would simply take pictures providing it was charged, continuously, without an off-switch or any insight into if the pictures are coming out acceptable.

I solved this by taking advantage of the Raspberry Pi 3’s onboard wifi chip in AP (access point) mode where I hosted a wifi network with served a flask app which showed me the latest few images the device had taken. The simple interface allowed me to see what was going on, plus giving me some basic controls on deleting or renaming images.

The web-app worked out great and the access point stayed online the entire holiday! This was probably one of the biggest successes of the project as I didn’t expect it to really work. As a backup I also had Apache symlinked to the image folder, which meant if the flask app would die - Apache would simply serve a basic folder explorer for me instead.

RTC issues

An interesting problem which I hit with the project with RTC (real time clock) issues. I knew I would face some problems seeing as the Raserry Pi didn’t have a RTC chip, however it really was show stopper.

To keep an eye on the issue I added a Server timestamp and Javascript timestamp so that I could compare what time the device thought the time was and compare it with the javascript one which my phone would figure out when it loaded the page. After a day of the camera taking pictures, if I turned it off and turned it back on the next morning, it would reset back to the last time it was turned on, which was usually around 24 hours ago. This means it would then start overwriting images it previously took during the holiday, as the images naming convention was based on timestamp.

The solution for this was to try to never turn the device off however the power charger I was using could only last about 30hrs which meant it inevitably died several times during the holiday. Luckily Eastern Europe is still full of internet cafes, so I popped in every day or two to transfer the images to a temporary folder.

A pretty horrible problem for something which could have been fixed by buying a £6 RTC chip!

The final timelapse

The final timelapse video came out great! It was a little time consuming removing all the error images however overall the images came out significantly better than the Narrative Clip. YouTube also made an enhanced version of the video which is really trippy and worth checking out here.

Doing some OpenCV magic

After my holiday I played around with a little bit of OpenCV hacking to try and take full advantage of the images I took. From the image above you can see that you can make some pretty significant improvements with some simple Histogram Equalization work. I didn’t really bother adding enhanced images to the timelapse video though.

IoThenticate

Fri, 13 Feb 2015 00:00:00 +0000

IoThenticate was a project Eric, Uros and I built at Startupbootcamp FinTechathon where we researched and prototyped a solution to help banks better secure their customers using modern technology. We spoke to the folks running the hackathon who explained how banks were pretty confident in how their own technology runs however a problem they’ve never managed to mitigate is their customers being manipulated through social hacking to giving away their account details to hackers.

They explained how it was incredibly hard to educate such large amounts of diverse customers and how they needed help in automating the checks and balances which go into place to make sure that the person logging into a bank account online is actually the correct customer and not a hacker. The CTO of LLoyds Bank explained how they really needed young fast paced technologists who were who were familiar with modern technology such as IoT, machine learning, AI, VR etc as banks felt hopeless in staying up to date with those kinds of things. Trying to not only understand new technology but also build solutions which utilise it for their customers and then deploying them globally something LLoyds has struggled to do for many years now and feels left behind.

What we ended up building was a mobile authentication system which LLoyds could deploy within their own banking app where a simple AI model to run predicting the likelihood that the customer logging into a bank account was the correct person by using a mesh of different data points. We mostly relied on bluetooth devices which the mobile would be connected to such as speakers or watches would help us confidently predict that the customer was within their own home. We also took in typical security features such as IP address but extended it to taking a fingerprint of other devices in your local area network too. This means we could automatically detect suspicious devices and display an additional security question for added safety.

The most interesting part of this app was that we could build a loose fingerprint of all bluetooth devices in your area regardless if you were connected to them and the same goes for wifi networks too. Using all of this we could confidently predict patterns over time around the customers behaviour and help flag when something didn’t feel right and we could take extra precautions before allowing the customer to access their information. This is something banks have been doing for many years using IP addresses and cookies however by taking in this new fingerprint of bluetooth devices & wifi networks along with our AI model we could significantly upgrade the amount of data points banks have to allowing access to private information of customers.

In the end our project was a great success and we ended up winning the prize for the Best Technical Solution and Pitch at the hackathon. What’s even better was that we were later invited back by LLoyds Bank to their London offices to explain our project to management there who could look into taking our ideas further to securing their customers from hackers.

View photo album

Seymour. See-more

Sat, 07 Feb 2015 00:00:00 +0000

This article has been lazilly migrated from my old blog, so there isn't much information to show here.

View photo album