Welcome to STEMology – Show Notes

Season 1, Episode 4

Patents, Rabbits, Wine and Bushfires, AI’s in the lab and Teeth

In this episode Dr Sophie Calabretto and Dr David Farmer talk about…

Humans only for Patents

It was literally developed to have ideas that could be patented

Rabbits doing handstands

They identified a single gene mutation that changes gate in all kinds of animals

Bushfires impacting wine

It works by shining a beam of white light through a sample of grape juice or wine

AI’s taking over the lab

Start from the virtual world, connect it to the real world, and then it turns out that it can actually do this incredibly efficiently, much better than people.

Cold Sensing Teeth

Apparently before this, we didn’t know exactly how teeth sense the cold


This is a “kind of, sort of, vaguely close” copy of the words that David & Sophie speak in this episode.

IT IS NOT 100% accurate.  We are very sorry if we have spelt something completely incorrectly.  If it means a lot to you to have it corrected, email us at stemology@ramaley.media

s1e4 Transcript

INTRO [00:00:00] Sophie: [00:00:00] Welcome to episode four of STEMology,

[00:00:02] a podcast that is your one-stop podcast shop for the interesting fun, and sometimes just patently bizarre news in science, technology, engineering or maths,

[00:00:13]  David: [00:00:15] Your hosts are Dr. Sophie Calabretto and Dr. David Farmer

[00:00:18] Sophie: [00:00:18] on today’s episode, we’re going to be talking about the need to be human to get a patent,

[00:00:23]David: [00:00:23] how some rabbits walk on their hands

[00:00:25]Sophie: [00:00:25] Bushfires and wine,

[00:00:27] artificial intelligence taking over the lab

[00:00:30]David: [00:00:30] and the cold sensors in your teeth.

Being human to get a patent

[00:00:31]  So Sophie.. I’m a human,

[00:00:37]Sophie: [00:00:37] correct, from my understanding.

[00:00:39] David: [00:00:39] We together are humans. and that means something that I didn’t know, I’m reading this week, which is that we can hold patents.

[00:00:47]Sophie: [00:00:47] Do you know something about the word patent that I really struggle with. Just saying it, so I’m really sorry, but I’m going to have to pause before I say the word patent every time, because in my head, I want to say paint or something similar. So just a heads up. Yeah. Pet patent. I just really have to think about it.

[00:01:04] Yeah. So we can hold

[00:01:05] David: [00:01:05] What about patents with

[00:01:07] Sophie: [00:01:07] night. Just doesn’t. I just feel like that should be two t’s in patent. Pretend that’s true. Anyway, we can hold patents. So if we came up with some super good idea, we could go to the patent office and we could say, Hey, I’ve come up with a super good idea. I’m claiming it as my own.

[00:01:23] And no one else can manufacture this thing.

[00:01:25] David: [00:01:25] That’s right. And our being human is the sole reason that we are able to do this because earlier this year, Intellectual Property Australia rejected a patient application from an AI called DABUS citing the fact that it was not human, it was artificial intelligence and that therefore only the owner, dr. Steven Thaler could be the patent holder. But he says that’s not necessarily appropriate either. Well, in fact, no, he didn’t say this, uh, professor Abbott, who’s a professor of law and health sciences at the university of Surrey school of law, said that’s not always appropriate because you could have a team of thousands of programmers who make an AI and the people who apply the AI and all of those could lead to patentable ideas, but it’s not necessarily legally correct to attribute the patent to all of those people who are involved either.

[00:02:10] So it’s not okay. You don’t want to give it to the AI, but you can’t give it to the people either. So what do we do?

[00:02:15] Sophie: [00:02:15] I mean, yeah, this is a real problem. Right. So I looked up DABUS. Do you know what DABUS stands for Dave?

[00:02:20] David: [00:02:20] No. Tell me, please.

[00:02:21] Sophie: [00:02:21] It is a Device for the Autonomous Bootstrapping of Unified Sentience. And from what I could tell it is just AI that it was created to invent things. So it sort of, it combines memories of various learned elements into potential inventions.

[00:02:35] And then it kind of assesses those and if they’re good or bad, it makes decisions based on that. And then it invents things.

[00:02:42] David: [00:02:42] It’s not even incidentally, not allowed to hold patents cause it happens to have a good idea. It was literally developed to have ideas that could be patented and it’s not legally allowed.

[00:02:50] Sophie: [00:02:50] Yeah, from what I can tell from the internet, there’s not a huge amount of information about DABUS, but the two patents that they were trying to claim, one was for a flashlight and one was for a food container. And obviously Dr. Steven Thaler thought that these were good enough that he needed to patent them, but then obviously he’s not allowed to patent them cause he didn’t create them.

[00:03:10] But then DABUS, can’t either, as you said, because DABUS is not a human. And so this is a little bit problematic and there’s actually a quote from Dr. Steven Thaler. And I like it a little bit because it insults academics just slightly. But he says that “AI is getting to the point where it can produce mediocre work”

[00:03:29] David: [00:03:29] Yeah, I love that.

[00:03:31] Sophie: [00:03:31] Right. “…for things like short journal articles, people can’t tell the difference between things written by an AI and things written by a person.” So sorry academics, your short journal articles, are mediocre work. But I mean, he brings up a good point. So there are more and more things that rely on AI.

[00:03:46] And if AI ends up inventing something, either on purpose or accidentally, who owns that intellectual property and who can make money out of it? And so I think, I mean, really, they just kind of need to update the laws to reflect the things that we can do now with computers.

[00:04:02] David: [00:04:02] Well, it also seems like, so this is a difficulty with patent laws and with the Australian patent office, but they have apparently not always been at the top of their game, anyway. Are you familiar with Ignoble prize?

[00:04:12] Sophie: [00:04:12] I am Dave.

[00:04:13] David: [00:04:13] So for the listeners, the Ignoble prize is a kind of satirical science award for research that makes you laugh and then makes you think.

[00:04:21] And one year in recent memory, it was awarded to John Keke, an Australian lawyer of Melbourne. It was awarded jointly to him and to the Australian patent office. It was awarded to him for patenting the wheel. And to  the Australian patent office for awarding that patent, he submitted a patent application for a circular transportation facilitation device. And it was in fact awarded. And so it’s, it’s, we’ve got thorny issues to do with AI here, but it kind of seems like maybe there’s a little bit of rubber-stamping going on there anyway.

[00:04:55] Sophie: [00:04:55] well, and this problem is not going to go away because I also looked up Dr. Thaler and he, from his website, he claims that he has, uh, owns 26 patents from all across the world. So US, Japan, Australia, Canada, India, UK, and Europe, are where  filed these things. So if anything, this problem’s going to get ,worse than better.

[00:05:12] And I think maybe the way that we stamp things needs to catch up. But also what we do with artificial intelligence, when they invent things in a way that we can’t

Hand standing rabbits

[00:05:32] Dave. I have a question for you. I know you have cats, but I’m interested in how your cats walk. Can they walk on their hands or do they tend to sort of just walk on all of their legs at once?

[00:05:44] David: [00:05:44] Well, I guess it, well, I mean, I guess it depends on whether you think they have hands and feet or just all feet, but usually they use all four of their feet at the same time. And then maybe sometimes just the back ones, if they’re like maybe in attack mode.

[00:05:59] Sophie: [00:05:59] Oh, like a kangaroo?

[00:06:01] David: [00:06:01] Yeah.

[00:06:01]Sophie: [00:06:01] Well,  the reason I bring this up Dave is cause we read something very interesting this week, and it’s about bunny rabbits. So it’s a certain breed of domesticated rabbit that it can’t hop.  It can walk around normally like a rabbit and you may confuse it for a normal rabbit when you just see it walking.

[00:06:16] But when it goes to hop, it doesn’t hop. It walks on its front paws with its hind legs in the air. Sort of like when you watch a CrossFit documentary and people have to walk on their hands from one end of the thing. Like to the other before they have to lift something very heavy. That’s generally where I see a lot of adults walking on their hands these days.

[00:06:33]David: [00:06:33] Whereas the CrossFit people are, for some reason, choosing to engage in this behavior, this rabbit, which is called the  rabbit, which is French for set that foot, which is French for Sutter of Al forte.

[00:06:48] Sophie: [00:06:48] And then, Oh, the funny thing about that is sorta actually means like hopper or jumper, which is an interesting way to name this particular rabbit that can’t hop or jump properly.

[00:06:58] David: [00:06:58] Okay. So Al’s hopper basically is what we’re talking about here. but so, whereas the CrossFit people will engage in this behavior willfully for some reason, this rabbit, it seems it does this because it’s got a point mutation and its genes. They identified a single gene mutation that changes gate in all kinds of animals.

[00:07:18] So they basically took some of these rabbits that exhibit the strange gate, and they mated them with rabbits whose genome had been sequenced in its entirety. And then they looked at the offspring of those rabbits. And specifically, they looked at the ones who showed the weird gate and the one who didn’t use the weird gate.

[00:07:36] And then by comparing the genomes, they were able to work out that a gene called ROR-B was the one that if you had it, a mutated version of it, then you did this weird walk where you stick your ass in the air.

[00:07:47]Sophie: [00:07:47] Yeah. And they found that when they bread days, I think they had 52 grandchildren and 23% of those grandchildren -so 12 of these baby -bunnies, or I dunno, maybe they let them grow up. We’re non hoppers and, yeah. And it’s the same so that I know that they did some studies a little bit. They’re still quite recent.

[00:08:04] I think it was 2017 where they did a similar thing with mice and I’m going to call this the ROR-B mutation because it sounds more fun than our RB and Rob and it’s quicker to say. And what they found was just like these bunnies that can’t hop these rodents. And I quote, would waddle around on their front paws, like a duck with their tails and hind legs sticking up in the air.

[00:08:26] So this is, and it was exactly the same mutation. No, I mean, I thought, and especially comparing it to a duck of all things!

[00:08:33] David: [00:08:33] One of the astonishing things about this seems to be. The fact that it’s a single gene mutation and that walking is so complicated. So walking comes about, because there’s a set of nerve cells in the spinal cord, which we call a central pattern generator, which is usually what you’re a scientists call a group of sales that do something that they don’t know how it works.

[00:08:51] It’s like the central PAX, your generator is like the black pixie dust containing pattern generator.

[00:08:57] Sophie: [00:08:57] Oh, it’s like a black box. Something

[00:08:59] David: [00:08:59] it’s a black box

[00:09:00] Sophie: [00:09:00] and a result comes out.

[00:09:01] David: [00:09:01] Exactly. It’s a central pattern generator, which means it makes a rhythm somehow. And in this case, it’s the rhythm of walking. It makes a pattern. And this happens in the spinal cord.

[00:09:10] And we know this happens in the spinal cord cause there was quite an amazing study a few years ago where they showed that even in rats where the spinal cord have been completely severed, if you put them on a treadmill, they would walk with their hind legs. Yeah, so the brain is not very involved in this.

[00:09:23] So walking is very, very complicated and happens in the spinal cord. And also they seem to have been able to nail this down to a single gene, which given how complicated it is, seems really weird.

[00:09:35] Sophie: [00:09:35] It does seem really weird. And what I liked, and even, you know, even the author, this is authors of this paper talked about the impact, but it’s the fact that they’ve combined developmental genetic and behavioral studies to sort of pinpoint this thing. And you think if we look at  all these different diseases and human motor deficits, this is potentially quite big.

[00:09:54] I mean, I don’t know that much about our genes compared to rabbits or mice, but it seems like we’re getting closer to being able to deal with

[00:10:02] David: [00:10:02] Well, I mean, it seems unlikely. Like if you’re an animal that’s evolved the need to get around by walking on legs, it seems unlikely that there’s going to be hugely different ways of doing that. Like, there’s gotta be some commonality in the programming for how we walk around because we all need to walk around.

[00:10:19]So I think it’s, it’s probably got some wide application and they mentioned a human motor deficit called Charcot Marie tooth disease, which is a nervous system disease characterized by muscle weakness might specifically benefit. But this is also just some really kind of cool general work that highlights how genetics influence something, which is really quite  complicated in a neuroscientific way.

[00:10:40] Sophie: [00:10:40] Yeah. And for a little bit of fun, everyone should now go and Google sorta Delpha or you could just say bunny handstand, walking, probably you Google search terms and there’s some really cute videos. So that’s our gift to you.

[00:10:54] David: [00:10:54] Enjoy.

Bushfires effect wine

[00:10:55]  Sophie, I enjoy wine and I suspect that you’ve enjoy wine.

[00:11:08] Sophie: [00:11:08] Yeah. I very much enjoy wine.

[00:11:10] David: [00:11:10] So I don’t know about you, but I get very upset when wine is ruined

[00:11:15]Sophie: [00:11:15] Yeah. In general, I hate it when wine, well, to be honest, I don’t think I leave one long enough for it to ruin via oxidation, but , I hear there are other ways that you can ruin wine that I was not even aware about until this week

[00:11:27] David: [00:11:27] I was not, I was not accusing you of committing blasphemy by opening a bottle of wine, and then leaving it to go bad. , but sometimes Sophie, it’s not in our hands because sometimes wine gets ruined even before it’s finished being made. And this is some, yes, this is some work of the university of Adelaide by chemical and pharmaceutical engineer, Dr. SueAnne Law. And she is leading research into how to identify smoke taint, which sounds like it might be rude, but isn’t, in wine. smoke taint specifically occurring when phenols produced by burning wood, For example, in bushfires, are absorbed by the growing grapes. And these phenols can bind to the grape sugars, which don’t taste or smell smoky, but it causes, during the winemaking process, they break apart.

[00:12:09] And then those smoky flavors come out in the wine and apparently people don’t want smoky wine. And therefore it’s not able to be sold.

[00:12:15]Sophie: [00:12:15] Yeah, I hear a lot of people spend lots of money on wine that specifically doesn’t taste smoky, but can we just very quickly get to phenols because you know what another word for phenol is? Dave?

[00:12:25] David: [00:12:25] what is it?

[00:12:26] Sophie: [00:12:26] Carbolic acid. Isn’t that terrifying? So phenol this, this thing that is getting into our grapes, is sometimes used in things like paint stripper or my favorite application, which is what I learned when I went to quarantine station at manly, is if you came from overseas on a boat and they thought that maybe you had scabies or lice or whatever, they’d put you in the carbolic acid showers, and shower you with carbolic acid until the outermost layer of your skin just fell off.

[00:12:54] David: [00:12:54] Yeah. It’s like Victorian soap. Isn’t it?

[00:12:56] Sophie: [00:12:56] Yeah.

[00:12:57] David: [00:12:57] It’s old-school soap. So

[00:12:59] Sophie: [00:12:59] yeah, in the wine, giving it a terrible, terrible taste. I mean, I’m sure it’s probably not great for our bodies either. But there is a way to detect whether or not our grapes have been smoke-tainted. And that’s the, that’s the important part here, not the carbolic acid showers.

[00:13:17]David: [00:13:17] Okay. I’ll just go excited about it. Cause it involves nanofabrication, which is when you make nanoscale structures, which are about a thousand thousand thousand times smaller than a well, a metre. And it works by shining a beam of white light through a sample of grape juice or wine.

[00:13:32] So beam of light, and then measuring spectral shifts in the light spectrum, and the shifts indicate this presence of smoke taint molecules. So what I’m imagining, if any of everyone listening can just imagine the cover to the pink Floyd album, dark side of the moon. And instead of a prism, it’s a nano fabricated  wine sensor.

[00:13:51] Then you’ve basically got the gist of how  this taint sensor works.

[00:13:55]Sophie: [00:13:55] And apparently manufacturing these tiny nano things, these sensors are quite difficult and they use an, you know, I get very excited when I get to learn new words, Dave, they use a lot of scary words in the names of the materials.

[00:14:10] David: [00:14:10] They sure do buddy.

[00:14:12] Sophie: [00:14:12] these tools will use nanporous annodic aluminium, photonic crystals as optical trans juices. So all of those words in the beginning,are too much for me, but an optical transducer is basically something that just measures the quantity of light. And then the inner surface of the crystals will employ peptide Aptamil is to target smoke tank molecules. And from what I can tell a Pepto Optima, a small Combinatorial proteins that are selected to bind to specific sites on their target molecules.

[00:14:41] So I don’t really understand the science, but it sounds very cool. But having said this, this is not. This is not yet being created. I think from my understanding, it is still at the testing, the emotional testing phase.

[00:14:56] David: [00:14:56] Very much so, and I think that’s evidenced by the fact that they sound quite excited about what they’ve they’ve managed to do. So they talk about  this poreous Aluminium smoke tank tool. And that’s nanostructures and because it’s holes, it’s poorest, which means holes. So they’re just like, we actually fabricated nano holes.

[00:15:13] Law says these nano holes, they have their own extraordinary optical transmission band. Again, nanoholes, sounds rude, don’t think it is, but I think this gets into what engineers call meta materials, which is basically, so usually if you have a material like a metal or plastic or something, Then the way light hits it is determined by the physical properties of the object.

[00:15:34] But with metamaterials, you have these tiny structures and the tiny structures actually have more influence on how the light behaves when it hits it, then the nature of the structure of the material itself. So I’m wondering if this is actually metamaterials and that’s, what’s responsible for this weird optical transmission band that’s letting us detect wine taints with nano holes.

[00:15:55] Sophie: [00:15:55] Yeah, it must be because otherwise. I could not find a huge amount of information. The closest, it sounds like to me is sort of absorption spectroscopy, but it’s not because it’s nano stuff. So I’m pretty sure I had to bet money, Dave. I would say that it works in the way that you just said it did.

[00:16:10] David: [00:16:10] Yes. And so, and this is really good and it’s really, really good because it turns out in the Bush fires, the big Bush fires we had recently, but 3% of the national crush, which is, I believe the amount of wine made were, were lost, rejected, or substantially downgraded as a relative smoke. And one of the hardest hit regions was Adelaide Hills.

[00:16:27] So it seems like where people are affected is quite localized. So the ability to test  locally is quite important to people’s ability to know whether their wine is going to come out smokey before they go to all the hassle. So it’s very good.

[00:16:41] Sophie: [00:16:41] So we can yeah. Interject at the, the right step and not waste all your time and go and test them better grapes that aren’t smoke tainted.

[00:16:48] David: [00:16:48] And we can all drink some bloody wine and not let it go too bad.

[00:16:52] Sophie: [00:16:52] Cheers to that Dave.

Sensitive Teeth

[00:17:03] David: [00:17:03] How are your teeth, Sophie?

[00:17:05] Sophie: [00:17:05] I believe my teeth. They’re quite good. I go to the dentist every year.

[00:17:10] David: [00:17:10] Specifically like thermally, how are your teeth?

[00:17:14] Sophie: [00:17:14] I I think  thermally their fine. I don’t think I have overly sensitive teeth. I do have an issue where I grind my teeth horrifically, which means that I wear down the enamel and I should really expose them to that kind of cold sensitivity, but not so much. So I’d like to say that my teeth are as strong as an ox,

[00:17:29] David: [00:17:29] Okay, so neither too warm, neither too cold. Just right. If Goldilocks was asked to pick teeth, she would pick yours.

[00:17:35] Sophie: [00:17:35] Exactly.

[00:17:36] David: [00:17:36] Okay. So , this week, we’re learning about research from the Howard Hughes medical Institute into tooth pain, and specifically why drinking cold things or being exposed to cold things when you have tooth decay is quite so sore. ‘it’s a unique kind of pain”, says David Clapp and vice president and chief scientific officer of the Howard Hughes medical Institute. It’s just excruciating. I feel like this guy has some strong, personal motivation to get involved in this research.

[00:18:03] Do you get that?

[00:18:03] Sophie: [00:18:03] I would, I would say that whenever he drinks something very cold or eats an ice cream, he is in a distinct kind of pain.

[00:18:12] David: [00:18:12] maybe an excruciating kind of pain.

[00:18:14]So these people worked out that tooth cells called  odontoblasts contain cold sensitive proteins that detect temperature drops. Right. and this was apparently controversial, apparently before this, we didn’t know exactly how teeth sense the cold.

[00:18:28] Because about 15 years ago, some post-docs in this lab  worked out that there’s an ion channel. An ion channel is something on the surface of a cell that lets molecules in or out.. And it’s important for the functioning of the cell. And this ion channel called tripsy five was very, very sensitive to cold.

[00:18:44] So when you made these cells cold, this ion channel would open and these cells would become active. But when they knocked out this ion channel and some knockout mice, or when they blocked this channel with a chemical compound that blocked them, they found that these mice could still sense the cold. So they had this cold sensitive channel, but they couldn’t work out where the hell it was sensing the cold, they hit a dead end.

[00:19:07] So 15 years later, they worked out at lunch one day. Apparently there was sitting at lunch, discussing the problem when they work too well, what other tissues in the body sends the cold? Teeth. So maybe it’s teeth. And that was a particular interest because not only do we sense cold, but the teeth, cold hurts our teeth when we have a cavity.

[00:19:26]Sophie: [00:19:26] Yeah. And so, and they found that it was in these odontoblast cells,

[00:19:30]David: [00:19:30] Yes. And that’s in between the pulp, which is the, the gross middle bit and the dentin, which is the hard bit, I think.

[00:19:37]Sophie: [00:19:37] and the dentin is the bit that’s under the enamel. Is that the structure of a tooth?

[00:19:41] David: [00:19:41] I’m guessing.

[00:19:42] Sophie: [00:19:42] I looked at a picture, I think that’s right. And so basically the location of the specific cell type , that was the thing that we didn’t know. So we knew that. So it was 15 years ago. They discovered this iron channel, but it was, they had no idea about the cells. And this is this new thing that we have now worked out and they did it. Like, I, I know that everyone knows this, but as someone who has never been an experimental scientist in terms of biology, they just use mice for everything.

[00:20:09] David: [00:20:09] Yeah, that’s right. Well they reproduce very quickly , and we know their genome and that makes them very useful. And that was how ultimately how they showed this. So in order to identify that it was the odontoblasts that were doing this, they couldn’t just look at the odontoblasts in a dish.

[00:20:21] What they had to do was to look at the system as an intact whole. So what they did was actually placed the teeth of these mice into some cold fluid, and then record the nerve activity. And they could see that when these odontoblast were cooled down, that the nerve activity was increased.

[00:20:38] So they must be the ones that do it. And what they’ve also shown, what they also say is that not only does this tripsy five channel show up on the odontoblast, if you have tooth decay, if you have cavities, then you have even more of this channel, which is what changes it into an extreme  noxious, ouch stimulus.

[00:20:55]Sophie: [00:20:55] so I think this is interesting, cause obviously, you know, in, as you said, they had a chemical that blocked this ion channel and they found that  there was no cold in these mice, but if this is a warning that you have a cavity, I don’t think we should be blocking these things.

[00:21:08] I think it might be the body telling us that your teeth are decaying and you need to do something about it.

[00:21:13] David: [00:21:13] No, but I’ve got, I’ve got mates who’ve had tooth decay and sometimes it’s a few days between like getting the tooth pain, like getting the tooth, the hearty tooth and getting the dental care and then it would be quite good if you could give people some compound that

[00:21:26] Sophie: [00:21:26] that blocks it in the meantime. Okay. So we’re being responsible adults. We’ve worked out that we have a cavity we’ve assessed that we need to get it fixed. And this is sort of as just a tie us over. So we don’t have to drink too much smoke, tainted wine,  to alleviate the pain.

AI in the lab

[00:21:46] David: [00:21:46] So I, I used to be a lab scientist, Sophie, and that meant the thing about being a lab scientist was you had to be in the lab, all the time. So I’m quite excited about this next story, because this is a story out of the us department of energy, office of science user facility located at DOE’s Brookhaven national laboratory.

[00:22:06] And I’m assuming how many words it has in the title is directly proportional to how cool and exciting the research is. But basically this is an ultra bright x-ray facility and scientists are so keen to use this in order to advance their research in areas such as battery development, micro electronics, and drug development.

[00:22:26] That demand for this place is absolutely three times as high as  it can handle in one day.

[00:22:33] Sophie: [00:22:33] Well, it’s interesting. I’m just going to jump in Dave, cause I was not a lab scientist, but supercomputers and this idea of queuing to use a resource to do research, it hurts me too. Because there will be times that you will submit a job to a supercomputing facility, but it’s a big one and a lot of people use it and then you just watch the status of your job as it is.

[00:22:53] queues. And it queues for maybe some hours or some days or some weeks. So I understand the pain of an oversubscribed facility very much.

[00:23:02] David: [00:23:02] So as a lab scientist, who had to once go into the lab, I’m excited about this work because basically they’re using an AI that used to play video games to run their experiments so that while they’re doing this lab science, they don’t actually have to be in the lab.

[00:23:18] Sophie: [00:23:18] Yeah. So I love this because one, I didn’t know that AI could play games, but apparently go and super Mario are things that they’ve been able to train AI to play very well, to the point that they’re masters. And so what they’ve done using the same approach, it’s the train, the AI to undertake experiments.

[00:23:36]So the way that they do this is they don’t even need to give the AI a bunch of rules to follow. They basically train it, using what they call reinforcement learning. And if you look up reinforcement learning,  if anyone has a dog, it’s just all of that sort of positive association stuff.

[00:23:51] So basically when a dog does something, correct, you reward it with a treat and when it does something incorrect, you just ignore it. You don’t punish. And that seems to be basically how this works, whereas you’re not giving the AI a treat. They understand that the more rewards you collect is like your key to mastering something and yeah. So as long as you have it, understand what a reward and what isn’t a reward, then it will train itself.

[00:24:19] David: [00:24:19] So, yeah, so the direct analogy would be if you’re super Mario collecting coins, or if you are Sonic the hedgehog collecting rings, depending on your allegiance to the 1990s console Wars, then you have a very simple thing. You can give the computer the number of those things it does multiple runs through.

[00:24:38] And then if it gets a higher number than it that’s the reinforcement, it gets a higher number and that’s the reinforcement. So it likes that path that it took better than the previous ones that got lower numbers.

[00:24:47] Sophie: [00:24:47] Yeah, and that’s how it learns. And so what they did is because they tested out their AI doing science on the pear distribution function beamline, which is just a facility in this facility. And what they did is they built up a virtual version, so they could train that AI on the virtual version.

[00:25:07] And then once it got trained, they unleashed it on some unknown samples that, you know, when people would deal with this often they don’t know the best way to assess these unknown samples. So they tried the AI on the unknown samples. They found out  that the AI had twice the efficiency of humans under  strongly constraints, circumstances such as limited measurement time.

[00:25:29] So basically the AI did this thing better than humans. And then it was just a matter of actually getting the AI to carry out the experiment. So, because it’s been living in this little virtual beamline world, what they did is they developed a program called blue sky adaptive, which was just an interface.

[00:25:44] Between the AI tools and then blue sky, which is just the software that  runs all of the beamlines in this facility. And so it was  start from the virtual world, connect it to the real world, and then it turns out that it can actually do this incredibly efficiently, much better than people.

[00:25:59]David: [00:25:59] And that’s not least because it doesn’t need to eat and sleep. Like speaking of scientists who are very like strongly personally motivated to do their work. They talk about how the bottleneck is the human beings themselves. We can’t just watch the measurement all the time because we also need to eat, sleep and do more than just run the experiment. I need a life, is what they’re saying. The AI doesn’t need a life and that’s why it’s more efficient. And then that’s quite on top of the fact that it doesn’t need to  have a life

[00:26:25]Sophie: [00:26:25] Yeah. So it’s sort of science coming out of a cry for help from scientists.


David: [00:26:41] And thank you for listening to another fun episode of STEMology. Be sure to check out all the links to these great stories on our show notes.

[00:26:47] Go visit www.stemology.com.au.

[00:26:50]Sophie: [00:26:50] If you have any news that you think is STEM ology worthy, drop us an email stemology@ramaley.media. We would love to give you a mention, if you give us some good ideas,

[00:26:59]David: [00:26:59] Your hosts have been Dr. Sophie calabretto and Dr. David Farmer.

[00:27:03]Sophie: [00:27:03] This is a podcast from Ramaley Media.

[00:27:06] Our executive producer is Melanie De Gioia.

[00:27:08]Our Music is from Elizabeth Maniscalco.

[00:27:11]David: [00:27:11] Be sure to hit subscribe on your favorite listening app. So you never miss our episodes.

[00:27:14] We look forward to sharing the latest in all things, science, technology, engineering, and maths with you next week, and be sure to bring your friends.