MATTSPLAINED [] MSP130 [] Science is Slick #8: Pencils, Ultrasonics and Water Cooled Windows
Could pencil and paper be the future of medical diagnosis sensors and facial recognition software? Can water filled windows cool your house? And are ultrasonic waves really the next jump in screen technology? Science is slick…
Photo by Unsplash. Glitching by Kulturpop.
Hosts: Matt Armitage & Jeff Sandhu
Produced: Jeff Sandhu for BFM89.9
TRANSCRIPT
There’s a smell in the air today. And that smell is knowledge. That’s right, MSP’s Matt Armitage is back with another round of deep tech stories as he announces that science is slick.
Where are we starting today?
Accessibility.
Something we’ve highlighted in a bunch of shows recently.
The coronavirus has made us all much more dependent on technology to maintain any kind of semblance of normality in our daily lives.
The Internet has become our tool for reaching the outside world - whereas only a few months it was the opposite.
A place for people to retreat from reality.
There are apps for everything.
And that leaves us with the problem of the people who don’t have access to the technology or for whom the tech doesn’t work.
We mentioned virtual touchscreens, I think…
Yes. Those Minority report style screens that enable you to navigate your hands without touching.
And we mentioned things like sonic pressure being used in those devices to create the haptics - the sensation of pressing a key or a screen.
A team of researchers at the University of Bayreuth in Germany has gone a step further.
They’ve created a haptic system based on ultrasound waves that can recreate braille characters in the air for the visually impaired.
The system can be used as an alternate display on ATMs or other public terminals.
How does it work? Using speakers?
Essentially. There would be a speaker array on the ATM, for example.
Those speakers in a 16x16 grid emits ultrasonic waves converge to create the dots that make up braille characters.
The user holds their hand up around 20cm away from the speakers, where the speakers create a virtual sonic image of the braille character that the user actually feels on their palm.
A bit like one of those air driers?
Not so violent, according to the researchers, more like a gentle breeze.
They tested various methods for displaying the text, and most users found that the method where the characters were spelled out dot by dot was the most accurate.
What’s also interesting about this is that it’s also a great way to access private or sensitive material in public places.
Privacy is an issue for many of us.
We worry that cameras in public spaces - whether nefarious or benign - could be capturing our screens as we type in passwords or access financial or personal information.
Obviously, most people are not going to go to the trouble of learning braille.
But this is one of those situations where a technology designed to level the playing field and increase accessibility for the visually impaired could also have wider applications for securing our own privacy.
I believe you want to stick with sound…
Yeah. Most of us are familiar with noise cancelling headphones.
Essentially the way that noise cancellers work is by broadcasting anti sounds.
If you have a noisy office background, the phones limit it by transmitting a frequency that cancels it out.
Where the waves of sound have peaks, the anti-sound has a trough.
So the overall wave is smoothed out and the noise is reduced.
In most instances, the range of the headphones is pre-determined.
Often tuned specifically for loud, low level hum like aircraft engines.
So you wanted to tell us all about a 30 year old technology?
No. I wanted to tell you about its evolution.
As I said, most of us are familiar with the headphones.
A team at the Nanyang Technological University in Singapore has created a noice cancelling system that works for windows.
We’ve mentioned on the show before that noise pollution can have significant psychological impacts.
Often, in modern office blocks it isn’t too much of a problem.
They’re climate controlled and have super thick double or triple glazed windows.
But that doesn’t help people at home who may be subject to the noise of traffic flyovers, trains, aircraft passing over.
Do you want to know how it works?
I feel that the show will probably grind to a halt if I say no…
Good man. I like the way they’ve done it because they’ve used a very South East Asia specific model.
At home most of us have window grilles, right?
So the system consists of 24 little loudspeakers around 4.5cm in diameter that can be mounted on the grille.
The system is flexible in that you can adjust the spacing to match the frequencies of the sound you’re trying to block.
And a sensor pointing outwards determines the frequencies that the speakers need to broadcast to cancel out the noise.
One of the nicest parts of the system is that it’s just as effective with the windows open.
Adjusting for the damping of the window itself…
Obviously, but because it’s anti-noise, having the window open won’t disrupt the efficiency of the sound blocking.
Current tests have shown that the system is most effective at cancelling out traffic noise and passing planes.
It isn’t attenuated for higher frequencies like human voices.
I’m not sure of the rationale - but it might be as simple as safety - you don’t want to cancel out the person telling you the building is on fire.
Similarly, blocking out low end noises would require larger speakers to displace more air and those might also block your view.
When can we buy them?
It’s just a prototype at the moment.
But with a bit of luck it will be real soon.
Tests with the prototype have shown that it can reduce traffic noise by up to 50%.
Which is huge, especially as our cities are getting noisier and more densely packed.
So the next stage is to test the prototype in more real world situations.
And hopefully, within the next year or two we might all have a system that can give us a more peaceful night’s sleep, or to enjoy an evening at home without having to put the TV or music on at ear-splitting volume to block out the world.
It’s hard to believe but we actually have more window technology…
I know. How exciting is this show?
I just mentioned double and triple glazing, right?
Triple glazed windows usually layers of argon gas to prevent heat transfer.
In cold climates this prevents heat leaking out of the home and cold leaking in.
Theoretically the reverse is the case, that they should keep the same area cool in hotter weather, though in my experience, they just turn into airless sweatboxes.
So, what are better gases for cooling your home?
Hydrogen? Oxygen?
Bang on. H2O. Our old friend water.
A team at the University of Kaiserslautern, Germany have built prototypes in Hungary and Taiwan that sandwich water between the panes.
Won’t that just create steam?
Yeah - one’s of water’s key properties is its ability to absorb heat.
Which is why it can be used in this instance for cooling.
We use heated water for drinking, eating, washing.
What is really interesting about this is that the water can be pumped out and replaced.
So you’re essentially using your windows and the sun to heat your water for a shower, for example.
Or even, for countries that have winter, to heat radiators or under floor heating.
So you get a double saving. You cut down on the energy required to cool - fans and ac - plus you’re using solar energy to heat your water.
Won’t the water distort the view?
The researchers claim no. They say that the water flow is slow enough that it’s practically invisible.
And the energy savings from computer simulations are pretty impressive.
The calculated savings of up to 72% for buildings with traditional double glazing. And up to 61% for triple glazed buildings.
There are still buts. Water windows are tricky to manufacture, and it’s not certain that the technology will scale or be cost effective.
And the simulations were mostly for buildings that have glass facades, so energy savings in homes with relatively small windows may be a lot more modest.
The inventors plan to set up a business selling water-filled conservatory glass systems next year.
Which makes sense. Conservatories are usually unbearably hot in the summer and unbearably cold in the winter.
They’re usable for about 30 mins a day, depending on the season.
I honestly don’t know why they’re so popular in Europe.
How radical is it?
The conservatory? Pretty established tech tbh.
If you mean water cooled windows - Well, as many countries place pressure on buildings to limit their emissions and essentially trends towards carbon neutrality,
This could be an important driver.
Because you start to look at the building as more of an enclosed system.
Your windows and your heating and cooling are now part of one system instead of 3 separate units.
So, it’s interesting from an architectural standpoint as well.
Because it potentially changes the way we look at the way buildings are structured and the kind of energy efficiency gains we can expect.
Window technologies. Where else would you be except MSP.
BREAK
There was quite a lot about different forms of energy in the first part of the show. I believe you want to talk about electricity next?
I do. I want to talk about electrodes.
Usually science is slick is about the good news stories.
Or at least, as close as I’m ever likely to get to them.
So, admittedly, this next story doesn’t really fit into that category.
Does it fit into the ‘we all know about’ category?
Yes. We all know about tasers.
Electrical stun guns that shoot out little darts and allow the operator, hopefully someone in law enforcement, to deliver shocks to the victim, rendering them incapacitated.
They fall under the rather dubious category of less lethal weapons.
Which has always struck me as a little chilling.
A bit like describing people as being less murderous.
I struggle for the context - chain guns are less lethal than atomic bombs.
But anyway…
Yes. We all know about tasers.
There’s now a kind of super taser called the Specter being trialled by the US military.
As usual, it seems like they’ve given it a cool name and worked backwards to make the tech fit it.
It stands for Small arms Pulsed Electronic Tetanization at Extended Range.
Being developed for US Dept of Defense by Harkind Dynamics in Colorado.
Unlike conventional electric pacifiers, this one doesn’t require a wire.
It can be shot from any traditional 12-gauge shotgun and can hit a target at up to 100m.
At that kind of range, wouldn’t the projectile itself be pretty lethal. Isn’t the wire still better?
Most tasers are only good up to about 8 metres.
Longer range devices have been developed and experimented with but the longer the conducting wire, the more risk of dangerous rather than debilitating shocks.
The Specter - i hate that this thing is cool - fires a smart projectile.
So yes, it could potentially kill if it hit someone outright.
However, sensors can detect when it’s about to impact and actually deploy a parachute that slows it down.
And it can also be programmed to detect further movement - so if the subject starts moving - it will automatically deliver additional shocks.
Also, the operator will be able to remotely trigger further shocks without approaching the person they shot.
Is this still a prototype?
The manufacturer is scheduled to deliver 100 rounds to the US Marine Corps for further testing this summer.
Bodies like Amnesty International are asking that it be independently tested and verified as safe.
Because it seems that if the tests go well, it’s destined for police departments and law enforcement agencies in the US at a time when police brutality is under the spotlight.
And we have more electricity?
Yes, I want to pivot from bad electrodes to good electrodes.
A researcher called Zheng Yan at the University of Missouri has come up with an inexpensive way to create sensors from pencil line drawings on paper,
The sensors can then be used to detect your heart rate, your skin temperature and even compounds in your sweat.
As well as the expense there’s also the waste - many similar medical sensors require the use of materials like copper and silicon.
Whereas these sensors, essentially all carbon, can be recycled.
What’s the catch?
No catch really. Other than a bit of artistic talent.
The sketches have to be geometric and conform to the parts of the body that are being sensed.
But there’s no trickery. The Missouri team used standard copy paper, a normal pencil and connected them to recording devices using a conductive cable.
The only thing they added was adhesive spray to help it stick to the skin a bit more securely.
They wouldn’t need that for me.
Plenty of sweat to create that papier mache bonding.
How does this sorcery work?
Well, as many listeners are probably mouthing to themselves right now, the graphite in the pencils is conductive.
So the sketched themselves carry the information from the electrical impulses in our skin.
And the research has also shown that in terms of ECG signals and the other things they’re supposed to do, they’re comparable to the silicon and copper sensors.
Do the patterns matter?
Very much so.
Serpentine - or to give them their nursery school terminology - squiggly, drawings work best, because the circuits isn’t interrupted if the subject bends or stretches.
Other than that it’s really about configuring them for purpose.
The electrical resistance of the sensor is determined by the length and cross sectional area of the traces, according to their inventor, Zheng Yan.
Do they have any further plans?
Well, as I mentioned you could also use the sensors to analyse sweat - for example to calculate glucose levels or uric acid levels.
So they would be useful for people with kidney issues, diabetes, gout. That kind of thing.
I don’t think the team is planning on making a fortune from this.
Most of us have a pencil and copy paper at home.
More impressively they’ve experimented with using the system to harvest electricity from humidity in the air.
And the hope is that they can harness those abilities to make the electrodes fucmtnion wirelessly.
Which really is my favourite kind of story.
Old, existing technology repurposed and effectively open source.
Although the idea that their kids drawings may one day be wireless and self-powered may be a little frightening to parents.
You’re linking these stories pretty impressively. The next story is also about line drawings…
I’m on a thematic role today.
We talked a while ago about a site that generated AI images on demand.
That way, if you’re a media owner say, you don’t have to worry about model releases, copyrights etc.
Or whether the story you’re attaching the image to in some way infringes on the subject’s rights.
Websites like Generated Photos have signed up a variety of media companies like the BBC, Fast Company and Vice.
You can go and play with them at generated.photos.
There are limited use licenses for free accounts.
Where do the line drawings come in?
That was just the set up.
We’re used to sites that create virtual people from scratch.
We’re also used to he growing number of deep fake apps like deep fake video apps like Zao and Doublicat.
Or the ability to age or switch genders thanks to the algorithms of Faceapp and similar services.
A team of developers at the Chinese Academy of Sciences in Beijing has found a way to reverse engineer faces from line drawings using AI.
The team used a photo set of around 17,000 celebrities and used image processing software to render each image as a simple line drawing.
The Ai was trained using the sketched pairs?
Yes, the original and the line drawing.
So it taught itself to recreate the faces by recognising five key components.
left eye, right eye, nose, mouth and then the rest of the face.
Once it creates the framework it then fills in those photorealistic details.
Can you customize the image?
No. Currently it’s entirely automatic.
So things like eye colour can’t be changed.
And the data set - surprise surprise - was mostly white people.
So skin colour and genetic diversity is still to be addressed.
One of the applications it can be used for is to replace police sketch artists for crime witnesses,
Now the witness can supply some really basic details and then the algorithm will get to work processing and creating potential people the sketch may resemble.
The witness can then select the generation with the closest likeness.
But it could also be used in film making and video games.
To save time but also as an additional creative input.
If you want a big square jawed bearded heroine, you can simply sketch the basics and the AI fills in the rest for you.
Can it be used for animals or objects?
Not yet. But that’s the plan for the future.
Which might be even more exciting than the human stuff.
I can’t draw. That’s not false modesty. I have no sense of perspective.
So I’ve never been able to sketch ideas.
This kind of tool would enable me to translate that image in my head into something real.
Did you ever think that Darwin prevented you from having that ability for a reason?
It’s not all torture devices.
Some of my ideas involve no applications of physical harm at all.
At a basic level you could imagine it being used on e-retail stores.
Sometimes you know what you want but the search term is so generic you get thousands of results.
And if you had a picture of the thing you wouldn’t be searching in the first place.
So a sketch might help you to narrow that search more efficiently.
I’m guessing that our final story is on AI too?
Bang on. I’m closing the circle as well.
We started with fairness and accessibility and we’re ending with fairness too.
We’ve talked about biased data sets.
In fact, that last story was a good example.
Fantastic results - if you’re like me, essentially devoid of colour.
The other 6bn people in the world, not so useful for them.
So we’ve talked about historical bias in datasets plenty of times in the past.
For example, a company that hires mostly people with blond hair, is likely to recommend hiring job candidates with blond hair because statistically the company has done better with blondes.
But the flip side is that a more balanced data source is less accurate…
That’s the prevailing wisdom. Your data set might be fairer but it may be less accurate or precise.
So you’re replacing one set of error-prone information with another.
However, Sanghamitra Dutta, a PHD candidate at Carnegie Mellon University, Pennsylvania thinks she’s come up with a solution.
This is an area of maths known as information theory that I know absolutely nothing about.
But the idea is to use its principles to create an idea data set that an AI can then be trained on.
In the case of the blondes, you might use the information in original data set to create brown haired candidates with similar attributes to the blondes that the company normally hires.
Then the AI won’t automatically weed out candidates with other hair colours and the brown haired candidates will be accurate representations of the people the company is looking for.
You know that most companies don’t hire according to hair colour?
I’m using it as a handy and non-controversial shortcut for race, gender, disabilities.
Also, because the blonde kids always got picked first for everything when I was at school.
So I’m addressing my own historical bias.
Dutta worked with colleagues at IBM and the system can be used to ensure all kinds of data sets, not just hiring information, is fairer and more balanced.
It can also be used as a control, to check whether existing data sets are biased and operating both accurately and fairly.
The research is due to be presented at International Conference on Machine Learning, actually on the day we record this show, so we wish Ms Dutta well and hope to see her ideas creating more balanced and representative data sets in the future.
Before we wrap up - all of today’s stories came from the New Scientist.
Some of you might think that’s just me being lazy. Looking at a single source.
I mention that partially to give the site the credit its due but also because the reason I do these science is slick episodes is…
…because of all the incredible breakthroughs and ideas we see on a day to day basis that sometimes get lost in the torrent of bad and depressing news.
The fact is, I got all these stories with just a a few clicks on one website.
This is just skimming the surface of our ingenuity, so don’t forget that, yes, there’s a lot going on right now.
But we’re still inventing and creating all these amazing, fantastical things.
Episode Sources: