Weird Science. The Elixir Of Youth, Superwood & CGI Surveillance

™s our first weird science of: 2023

Matt Armitage:

• I think most are probably just inane.

• They achieve inanity and aspire to inanity.

• That sounds like something you’d read in an almanack.

• Like many of us, I took a bit of a break from the newsfeeds this year.

• And it’s often the case that it’s a bit quiet this. No really weird stuff.

• But genuinely, once I went back to my feeds there’s no shortage of freaky stuff.

• We’ll start with something that’s not really freaky.

• This is a site that was recommended by a friend but because we took a break last week, it’s everywhere already.

• ChatGPT. The solution to everyone’s problem for everything.

• From recipes to contracts to code.

• As I’ve mentioned before these are assistive tools, not solutions.

• If you want information, go to google.

• If you want it to write something for you – research the facts and give it the information.

• Don’t rely on it to contextualise for you, let alone search for correct or supporting information.

• We reported that OpenAi is looking at building patterns into the generated text to make it easier to detect as machine-generated.

• In the meantime, a new tool called GPTZero has been launched.

• I think it’s in public beta atm.

• It’s aimed primarily at educators who want to check whether their students are writing essays or cheating.

• By having the machine write it for them.

• And it works. I generated some text and then plugged it into GPTzero and it correctly identified it as machine created.

• I put in something that I’d written myself and it identified it as human-created.

Richard Bradbury: Do you count that as a win or a loss?

Matt Armitage:

• I’m genuinely unsure. It certainly knocks my credentials as the first human-AI.

• Or maybe it doesn’t. Maybe I am the man-machine.

• And this is a vindication.

• But my concern I think would be how this tool is used in much the same way that I’m concerned about the ways that ChatGPT is being used.

• GPTzero isn’t the solution to the problem of people using ChatGPT as a solution.

• Now, I know that sounds like the kind of gibberish that an AI might produce, so bear with me.

• GPTZero, to my mind is no more a solution that ChatGPT is.

• Yes, it can tell you if the text has been machine generated.

• I’m less sure about the degree to which it will tell you whether the user has used AI to help structure the writing and has sourced the arguments and the facts themselves.

• This kind of goes back to conversations decades ago about the role of Microsoft Word and that it would make secretaries and personal assistants irrelevant.

• Another friend who reminded me of that in a conversation last weekend.

• And I think that’s one of the bigger issues with technology like this.

• In a few years, it will be as commonplace as word processing software.

• What is needed now are the guideline for using these tools.

• For example, in the case of a school or college essay – how much help is permissible?

• In-built grammar and spelling tools, as well as extensions like Grammarly, are already helping us to write more correctly.

• Grammarly especially offers targeted help with phrasing.

• Obviously, using an AI to source your facts and your argument is wrong.

• But we forget that structuring an argument and presenting your information is almost as important as doing the research and finding the facts.

• An essay isn’t a list. It’s an answer to a question or a problem.

• So how much do we allow a machine to help structure those discussions?

• These are the discussions that I imagine a lot of educators and editors are having right now.

• Letting those tools help ideas breathe for people whose writing skills are not great or average.

• While stopping people from using them as a shortcut to the argument itself.

Richard Bradbury: Are we seeing anyone using it in a more commercial or industrial capacity?

Matt Armitage:

• Interestingly, it’s come out that the Tech website CNET has used it – quietly – to create about 75 pieces.

• Mostly on the CNET Money desk.

• In a statement on the site, the editor Connie Guglielmo said that they were trialling the tool for a number of reasons.

• Partly because they saw it as part of their role as a tech news provider to use the tool and accept its limits.

• And also to see how it could supplement the work of the writers.

• For example, if a writer was writing a piece about derivatives or interviewing a derivatives expert.

• Would it be possible to use tools like ChatGPT to provide a tailored explainer as an accompaniment to the main article.

• Ideally, you want another human writer to do that but in competitive news marketplaces, where your advantage on a story is measured in minutes.

• There often isn’t time. How often – especially when it comes to anything Elon – is what we talk about on this show slightly out of date.

• Simply because it’s recorded a day before broadcast.

• So I can see how this tool isn’t just a way to save on staff costs.

• It’s about exploring its potential to extend or complementing your reporting.

Richard Bradbury: The fact that we’re talking about it suggests that it didn’t go well.

Matt Armitage:

• I think firstly CNET deciding to do it on the quiet – I understand why.

• They want to see if their readers can tell the difference,

• That’s the kind of thing I would do.

• Which is why no one would let me be editor of a big tech site.

• So a number of rival sites ‘discovered’ in inverted commas.

• That CNET was doing this. It was less than obvious rather than hidden if we’re being charitable.

• And those websites discovered a number of inaccuracies.

• And that’s where we come back to that same – tools not solutions argument.

• Sorry to keep boring people with this.

• But this shows why it’s so important.

• Stories generated by machines need much heavier editing and fact-checking.

• Especially when the machines actively reinforce the incorrect statements to make them sound more plausible.

• And this is what seems to have prompted that statement by CNET’s editor.

• Ultimately, it calls into question the efficacy of the method.

• If this is supposed to be a time-saving tool - it places a lot more pressure on those editors and checkers.

• Possibly more so than having a well-informed writer create it in the first place.

• To quote Daft Punk we assume that technology is always harder, better, faster, stronger.

• We have to have those discussions now because these technologies aren’t from the future.

• They’re here. And they’re already influencing and impacting upon our world.

• Would you like a story about a smelly robot?

Richard Bradbury: [replies]

Matt Armitage:

• I think this one is from engadget.

• We’ve done a few stories on electronic noses over the years.

• We have all this incredible technology.

• Satellites and drones that can watch the world with superpowerful lenses.

• Microphones and sensors that can pick up the tiniest sounds and vibrations.

• All kinds of machines can be programmed to deliver the correct level of touch to an object its handling.

• I say programmed deliberately. We’re only recently getting to the point where cameras and other feedback devices can create machine arms or hands capable of varying the pressure according to the object.

• But we’ve found it much harder to replicate taste and smell.

• Well, a team at Tel Aviv University have come up with a biological robot that uses AI to detect scents.

Richard Bradbury: Yeah, so this is where Matt uses neutral sounding words to describe something horrific. A biological robot is a cyborg, isn’t it?

Matt Armitage:

• I guess you could call it that.

• In this case it’s really just a roller skate with some locust antennae.

Richard Bradbury: I know that’s your idea of cute. But calling something a roller skate with insect parts isn’t reassuring.

Matt Armitage:

• Potato/potato.

• Grasshopper/locust.

• In this case mostly locust.

• The research team used antennae from a locust.

• Presumably, because they’re very good at detecting scents at a distance.

• Rather than because they were going for a particular aesthetic.

• Coupled to an ML system, the system can learn to imprint and then detect certain smells.

• The first trials are quite limited – I think engadget mentioned 8 odours including lemon, geranium and marzipan.

• The idea is that as it expands, the device can be used to detect explosives and bombs.

• Or drugs. Or even spoiled foods.

• So a variety of public and private uses.

Richard Bradbury: What are we going into the break with?

Matt Armitage:

• It’s the shock revelation by scientists that 5 year old children are smarter than chimps.

• That might surprise the owners of some 5 years olds. But that’s probably because they haven’t also owned a chimp for a direct comparison.

• Interestingly, the same study showed that chimps are nearlyas smart as your 4 year old.

• And definitely smarter than humans below that age.

• If you can hear a slight rustling noise, that’s the sound of Richard shaking his head in a morose fashion.

• Either it’s because this isn’t where he thought his career would take him.

• Or because he faces all the backlash from parents because I don’t really associate with any.

• But yes, scientists at the university of st Andrews in the UK found that 5 year olds are better at switching their attention for different tasks than chimps.

Richard Bradbury: [replies]

Matt Armitage:

• Now, I initially had this vision that the lab contained a mixture of kids aged 5 and below.

• And a mixture of young, adolescent and adult chimps.

• And then I realised what a health and safety risk that would be.

• It would really endanger the chimps.

Richard Bradbury: Not helping…

Matt Armitage:

• Essentially they were given similar tasks with different rules.

• This is from NS btw.

• 2 sets of shelves, one green, one blue.

• Each had 4 cups on it. On the green shelves there were treats inside a green cup.

• On the blue shelves the treats were in a pink cup.

• The test was to see how they could switch between the two and remember the rules associated with them.

• The chimps did pretty well, their accuracy was around 52%.

• Quite similar to four-year-olds at 59%.

• And outperforming 3 year olds who only managed 50%.

• So there you are – that’s my public service bit – helping to give you the advice you need to choose between young humans and primates.

Richard Bradbury: On which everlasting nonsense we’ll take a break…

BREAK

Richard Bradbury: AI, locusts and children. Who are we going to annoy in this part of the show?

Matt Armitage:

• Hey, I just follow the science. You can’t blame me for where it takes us.

• Just be thankful I didn’t talk about the Australian spiny anteater that blows snot bubbles to keep itself cool.

• Although that story would probably work with the 5-year-olds.

• You can google that one for yourselves. They’re very cute.

• Until you think about the snot bubbles.

• It’s the new year, so let’s have some anti-ageing stuff.

• I’m getting to the point in my life where I have to start considering senescence.

• The point at which some of my cells don’t split and rejuvenate anymore.

• They just sit there getting lazier and lazier.

• Content to sit on the sofa eating pizza instead of heading out with their roller skates and antennae to keep themselves in shape.

• The big problem with that is that laziness spreads.

• Senescence can have a cascading effect, essentially behaviourally infecting neighbouring cells.

• And toxins can accumulate in the cells creating swelling effects like arthritis.

Richard Bradbury: Little ball of new year joy, aren’t you?

Matt Armitage:

• Well, I’m not doom-mongering. Again this is a NS story.

• A lot of treatments for age related diseases have taken the approach of eliminating the senescent cells.

• Which makes sense.

• But the more direct approach is to prevent cells reaching that dormant state.

• That’s what a team at the university of Texas medical branch is looking into.

• And they’re using a technology that may already be familiar to people in the cosmetic surgery world:

• Low frequency ultrasound.

• The Texas team has managed to rejuvenate old mice as well as restarting cell division in human cells.

• The human cells they used – foreskin cells, in case anyone is interested, presumably because there’s a constant supply in the US.

• The cells usually reach senescence after around 15 divisions.

• When subjected to ultrasound at around 100hz they have reached 24 subdivisions with no apparent abnormalities.

• Using marker dyes in mice, they found that the ultrasound treatment reduced senescent cells in the kidneys and pancreas.

• Treated mice seemed to reverse the effects of conditions like arthritis.

Richard Bradbury: Do we know why this works?

Matt Armitage:

• Not really. They know it works but not exactly why.

• The team’s lead thinks it may have something to do with the ultrasound waves mimicking the effects of exercise.

• Stimulating the cells in that way. And they know that maybe it’s too big a puzzle to solve on their own.

• They’re considering selling their machines to other research teams.

• And that way have a picture built up by a much larger group of scientific minds and approaches.

• One of the issues with this kind of research in the past is that different researchers have used different machines and frequencies.

• So it’s hard to compare one study to another.

Richard Bradbury: Could this be the elixir of youth?

Matt Armitage:

• I think there’s a lot of work to be done before we see any therapies on the market.

• I’d be a bit wary of anyone claiming that a quick zap of ultrasound will solve all ills.

• There’s the issue of delivery. Bones and organs can absorb a lot of the ultrasound waves.

• So figuring out what to target, how much power to use and how to actually reach the target are a ways off.

• And, of course, every time a cell divides there is a risk of cancer.

• So far, nothing in the research has suggested that it increases cancer risks.

• An initial trial with people suffering from osteoarthritis is being planned.

• One of the more surprising effects is that it seemed to reverse symptoms of dementia in mice.

• So the team is also interested in seeing whether it could be used for conditions like Alzheimer’s in humans.

Richard Bradbury: I believe we’re sticking with waves but moving on to WiFI.

Matt Armitage:

• Yes but no but. We’ll come back to waves in a minute.

• I want to move onto something that has been a huge favourite on Mattsplained over the years.

• And that’s superwood.

• I don’t know why. Because this isn’t a specifically environmentally focused show.

• But advances in materials science is something that we frequently cover.

• But there’s something about superwood that seems to grab people’s imagination.

• So this latest development is a treated wood called insulwood.

• As you might have guessed it’s a wood that can be used as an effective insulator.

• Now, in Malaysia, we don’t typically need much insulation.

• But it can also be used as a sound isolator.

• So any of you that live in concrete and steel boxes that transmit the sounds of your neighbours through the entire frame of the building in a thrum.

• Might appreciate this development.

• Also because it’s highly sustainable.

• A lot of insulating and isolating materials are made from foam, which as we know is made from oil.

• And it’s an additive – which I’ll get to in a minute.

Richard Bradbury: If we’re going to touch on sustainability, why is insulwood a good option?

Matt Armitage:

• It can be created from a lot of fast=growing woods.

• The research team at University of Maryland used a wood called paulownia which is commercially grown around the world.

• Apparently basswood, balsa and pine are also suitable.

• Typically wood is a poor insulator. What the Maryland team did was dip the wood in a solution of sodium hydroxide.

• It sounds awful but is common in the paper industry. It removes lignin and hemicellulose.

• The treated wood is then dried for 7 hours.

• And the result is much less dense, almost a third less.

• The removed fibres become tiny little air pockets, that increase the insulating properties to roughly the same as the equivalent EPA insulating foam.

• And should still be biodegradable at the end of its useful life.

• But the biggest benefit is that the wood remains almost as strong as it starts.

• Currently you inject insulating foam between two brock or concrete layers.

• With insulwood, you have a structural material that also insulates.

• It’s no longer an additive, it forms part of the physical strength and structure of the building.

• I know materials science can seem dry and boring.

• But advances like this, that help us to build better homes and buildings in a more sustainable manner.

• Can only be a benefit.

Richard Bradbury: So, how are you planning to wave goodbye today?

Matt Armitage:

• I see. A dad joke.

• To be fair – I made Richard say that. His actual jokes are much worse.

Richard Bradbury:

Matt Armitage:

• To end, this is kind of an update of a story we covered on Mattsplained in the years BR.

• Before Richard. Back when there was an idealistic young whippersnapper called Jeff chatting to me.

• And it was a story about what was essentially a denoising system that researchers at MIT came up with.

• That identified people moving around their home from the Wifi signals.

• Now, if that sounds like junk science, which, when I recounted it once, I was pretty much accused of presenting.

• Think of radio waves as a kind of invisible fuzz that surrounds you all the time.

• So when you move around your house you’re constantly interrupting that field of fuzz.

• The MIT team found a way to filter all that noise – because there are overlapping radio signals and other fluctuations that cut into those fields.

• And they were able to use an ML algorithm to turn the movements of people within those WiFi coverage fields into stick images on a screen.

• Which would portray you sitting, walking, bending etc. Your imagination can fill in the rest.

• One potential use would be home security.

• Rather than having cameras set up in every room, your WIFI coverage could tell you how many people are in a room or house or building.

• What rooms they’re in. etc etc.

• So positive benefits but also lots of potential for abuse.

• It would be relatively simply to sit outside a house with a radio receiver and see how many people were in a house and what they were doing.

• Or to snoop on family members.

• It was also hypothesised at the time that it could be used for low cost govt style surveillance.

• Can’t live without Wifi. System can be managed remotely by AI.

• And given the trends we’ve seen in certain countries since the start of the pandemic, there have been active attempts to extend surveillance of populations.

Richard Bradbury: That’s a long preamble…

Matt Armitage:

• I’m nothing if not long-winded. This story’s from Motherboard.

• The MIT tech has been improved upon by researchers at Carnegie Mellon.

• They took a system called DensePose, which was part developed by Facebook.

• It maps the pixels on human bodies in photos.

• The team developed a neural net that they applied to DensePose that would enable them to do the same thing with WIFI.

• Denoise the signal and turn the points of a person into pixels.

• The result is a 3D map of the humans in a given space, offering far more detail than the previous MIT breakthrough.

• If you want to try and visualise it – you’ve seen those 3D bodymap projections that are used to translate human actors into CGI.

• It looks like that.

• So people might be distinguishable from physical size, their gait or other physical characteristics.

• Of course, this is being presented as a way to enhance home security.

• It is. You can see intruders. You could also see what kids are up to inside bedrooms or other spaces.

• Without the intrusion of cameras.

• But you do wonder what kind of world this tech is opening us up to be.

• And what controls or restrictions, or even workarounds to prevent third parties spying on your network are going to be needed in the future.

• There you have it. The Elixir of Life. Superwood. And a CGI spying system.

• That’s Weird Science at its best.