Dr. Alan Mackworth — professor emeritus with UBC Computer Science — has been studying AI since the 1970s. He’s served as president of the Association for the Advancement of Artificial Intelligence and held a Canada Research Chair in Artificial Intelligence from 2001 to 2014. In this conversation Dr. Mackworth discusses the challenges of regulating artificial intelligence, if age-based bans on AI make sense, and how chatbots are trained to flag harmful actors.

By Geoff Gilliard

You’re a pioneer in the development of AI. Is it being developed and used in the way you hoped?

I knew there was enormous potential in machine learning, but it’s gone beyond what I expected, and it’s come faster than I expected. Of course, there were the true believers in deep neural nets who thought that they were the answer, and they were right about how effective those models turned out to be. But I think the majority of AI researchers, including me, were quite surprised by the speed and effectiveness of those techniques.

I always believed that AI would eventually be successful. Sixty years ago, when I started in the field, I was ridiculed. People said that’ll never happen: “Humans will always be wiser and smarter than machines.” Many philosophers said that it’s just not going to work. Computers will never play chess. They’ll never be able to fool people into thinking that they’re human. Now they do all of that and much more. So it’s dangerous to predict that AI will never be able to do “x” — whatever “x” is.

One focus of your research is constraint-based agents. What are those?

An agent does something in the world. It senses the world. It may reason a bit and then it acts according to a plan it constructs. Making those plans satisfy various constraints is important.

For example, we built the world’s first robot soccer players, and one of the tasks that soccer players have to do is to intersect with the ball at the right time, at the right place, in the right direction, in order to redirect the ball into the net. You can think of that as being a dynamic constraint, because it’s changing over time. You have to make sure that you’re solving that constraint in real time.

It also means that you take into account safety constraints like Asimov’s Three Laws of Robotics: You should never allow the robot to harm a human being, the robot should always obey a human being, but not if it tries to get you to harm another human being, and a robot must protect its own existence except where such protection conflicts with the first or second law.

Is a chatbot an agent?

Yes, they are agents. They take inputs — speech or text. They do a little bit of reasoning. What they’re typically trying to do is to predict the next word of the output. They’re not acting in the physical sense, in the real world, but speech is an act.

How are chatbots monitored?

When tech companies originally train these agents, they often say the most outrageous things, and those responses are then filtered by human observers. Tech companies are paying people — often minimal wages in the third world — to monitor the results of these agents. They’re looking for threats of violence, sexual innuendo, and other responses that are flagged as nasty. It’s called reinforcement learning through human feedback and that’s supposed to produce safe bots.

There are billions and billions of messages going out every day, so you can’t have humans looking at all of them. But you can use AI to help solve this problem by doing a triage of messages saying this one is definitely harmful, this one’s definitely safe, and this one we’re not quite sure of, you should take a look at it.

There is an attempt to monitor chatbots with other agents, looking for and flagging any conversation with a human that might talk about self-harm or threats of violence. But not always successfully. As we know, many things get through. OpenAI (builders of ChatGPT) admitted that, in the case of the Tumbler Ridge shooting, their safeguards were too slack and that they have tightened them up. So they were voluntarily changing the safeguards. But who knows how long they will keep those safeguards? Also, reports indicate they failed to detect further activity after they blocked the account of the shooter before the tragedy.

A study we published argued that companies require independent audits, they can’t monitor themselves. It’s just not acceptable. And it’s not just OpenAI. There are many, many other companies like Google and X and Meta that are doing the same thing. We have to have uniform, public, enforceable safety standards. That’s why it’s essential to have government regulation of these activities.

Should there be an age limit for using chat bots?

We have to be very careful not to throw at the baby out with the bathwater. I don’t think we should be banning all access to AI systems for kids because they need to learn how to use them wisely, just as I don’t believe in banning them in universities or anywhere else. But what we should do is tell the tech companies that we’re considering an age-related ban if they don’t agree to proper public regulation. We have to ask what methods they’re using to regulate age-appropriate access to their system and monitor its effectiveness through a public safety auditor. If those methods are proven to be effective then we would permit the companies to allow appropriate access and leave it primarily up to parents to decide when, and how, to let their child use AI and social media.

Now the companies are saying, well, it’s always up to the parents, but the parents can’t monitor all of this stuff. It’s not good enough to say that we have to have public regulations, public monitoring and appeals. We have to make sure that there’s redress if harm is done, and make sure that we do as much in terms of public education and the wise use of these resources.

It seems tech companies have adopted a similar approach to regulations as that taken by auto industries’ opposition in the 1960s to putting seat belts in cars. Isn’t it in their best interest of tech companies to make chatbots more trustworthy?

Yes, indeed. There are two kinds of AI safety. One is the shorter-term concerns, like the ones we’re seeing — online harms, addictive use, encouraging people to kill themselves, deep fakes, IP theft, hallucinations and all the rest of it. Then there are longer-term safety concerns, like the worry that there’s an existential risk that AI might conspire against us and eventually kill all the humans.

Tech companies talk about the latter a lot in public. I think, in a way, it’s marketing, because they’re saying that the agents they’re building are so powerful that they could conspire against us and kill all humans. So they’re claiming we have a vested interest in developing the most powerful agents before anyone else does. It’s also a way to deflect attention away from the immediate shorter-term concerns.

But Big Tech originally was on board with regulation when Joe Biden’s administration was in power. There was a big move towards regulating artificial intelligence through the Kid’s Online Safety Act.

But then, the same tech companies donated large amounts to Trump’s presidential campaign and have now lobbied for very little or no regulation of AI. In fact, the tech companies are even trying to stop individual U.S. states from regulating them. In Canada, we’ve been trying to bring in the Online Harms Act, and it’s failed for several reasons. Over the last seven years there’s been virtually no progress, partly, I think, because of lobbying.

Can Canada regulate an industry that’s based in another nation and has a global reach?

It’s a problem, and it’s certainly one that the U.S. is resisting, saying that Canada cannot regulate these companies. So there’s a real problem of sovereignty here. But we could do it. The tech companies are operating in this country; any company operating in Canada has to follow Canadian law. It’s been done in other countries. The European Union passed the AI Act in 2024. In the United Kingdom there’s an Online Safety Act, which served as the model for the Online Harms Bill that the Canadian government has been working on.

The Online Harms Bill proposed a Digital Safety Commission of Canada with an enforcement body with the power to audit platforms, issue compliance orders and lay massive fines up to six per cent of a company’s gross global revenue. The bill called for a Digital Safety Ombudsman, a resource for the public to support victims and advocate for user interest. And there would be a Digital Safety Office of Canada, which would support both the Commission and the Ombudsperson. We can create these obligations for platforms. There will be pushback. There will be lobbying, but we have to assert our sovereignty. The safety of our citizens is at risk. There is some recent progress in the establishment of the Canadian AI Safety Institute.

At the moment there’s pushback against artificial intelligence by the public because of some of the negative effects that chatbots are having. AI is being widely used, more and more so every day, but people are very wary of it, both for these immediate, short-term threats and longer-term threats about job loss and lack of safety. There’s now sufficient support among the public, whereas there may not have been in the past. We’re realizing how real and urgent these dangers are. They’re not artificial dangers, and there is support for passing the bill.

Can Canada constrain AI? was originally published in Focus on Medium, where people are continuing the conversation by highlighting and responding to this story.

UBC outreach is shoring up science education on Hornby Island — and having ripple effects

By Geoff Gilliard

On a gloriously sunny day 11 schoolchildren scamper barefoot across a beach on Hornby Island searching for seaweed. Who knew school could be so much fun?

Dr. Bridgette Clarkston did.

“Seaweeds are a gateway into learning and caring about the ocean and coastal biodiversity,” the UBC botany professor says. “They’re so easy to access and it’s such a great excuse to go down to the beach and explore a place where most of us don’t spend a lot of time — even if we live near the seashore.”

Throughout the 2025 school year Dr. Clarkston taught a Hornby Island Community School Grade 7 and 8 class about seaweeds long distance via Zoom, leading up to an in-person visit in May, when she and four of her UBC biology students headed to Hornby for some hands-on learning. The visit was funded by a grant from the UBC Centre for Community Engaged Learning.

“I do a lot of work with my undergraduate students to share their learning beyond our classroom with elementary students,” says Dr. Clarkston, a professor of teaching in UBC’s Biology program. “They could see the immediate impact in sharing their knowledge with new learners. And they loved it.”

Early in the school year, the Hornby Island kids were assigned “seaweed buddies” — specific seaweed species that they learn to recognize over the course of the school year. On beach day they collected and prepared 24 seaweed specimens that are now scientific records in UBC’s Beaty Biodiversity Museum.

There is no high school on Hornby Island. Once the children get to Grade 8 they’re bussed to the nearest high school in Courtenay on Vancouver Island, a daily two-hour, two-ferry commute each way. It’s quite a slog. Some families choose to leave the island when it’s time for their children to go to high school. Others home-school or make billeting arrangements for the kids to stay with friends or family closer to the school during weeknights.

“My class is an extension of the elementary school so we don’t have access to a science lab or anything like that,” says teacher Yakotsenhtha Heather Royal, who teaches the split Grade 7, 8 and 9 class. “When things in their education are missing, it’s really hard for parents and students to opt to stay on Hornby Island until, say, Grade 10.”

Forests of the sea

Seaweeds are a subset of algae. The many thousands of species worldwide are classified into three branches of the seaweed family tree of life — red, brown and green. Seaweeds photosynthesize and are foundational to coastal marine ecosystems, but the three branches evolved independently. Red and green seaweeds are grouped together with land plants like mosses and vascular plants — brown seaweeds are related to microscopic algae called diatoms.

“We’re very lucky in this part of the world because there’s an overlap where two iconic kelp species live — bull kelp, Nereocystis luetkeana and giant kelp, Macrocystis tenuifolia,” says Dr. Clarkston, who specializes in seaweed biodiversity. “Think of these two ‘canopy-forming’ kelp as the redwoods or Douglas-fir trees of the underwater forest, reaching from the seafloor to the water’s surface, creating a three-dimensional underwater habitat.”

Coastal B.C. is a diversity hot spot hosting over 600 species of seaweeds. But climate change is driving drastic declines. Summer sea surface temperatures of the inner Salish Sea are hitting 20 or 21 degrees Celsius. That’s about six degrees too hot for some kelp to grow.

“We’re seeing declines in some kelp species in British Columbia for a number of reasons. In some areas, in particular the inner Salish Sea between Vancouver Island and the mainland, warming waters may be one of the biggest drivers of this loss,” Dr. Clarkston says. “Bull kelp used to be quite abundant in the waters near Hornby Island and my hometown of Comox. Now they’re gone.”

The day after last May’s beach walk, Hornby Island students hosted a public shoreline walk with family members and other islanders.

“Some of the community members who’ve lived on the island for years remembered when there used to be kelp forests,” Dr. Clarkston says. “Other folks who are involved in the seaweed restoration efforts came to learn more about what other species are found on the island.”

Seaweed School serendipity

For the past 10 years, Dr. Clarkston has been running public workshops through the Raincoast Education Society in Ucluelet, B.C. on the outer coast of Vancouver Island. Entrepreneur and seaweed advocate Elizabeth Sheehan, a volunteer steward with the Hornby Island Natural History Centre, met Dr. Clarkston at a workshop and invited her to bring seaweed education to Hornby Island.

“I just loved how Bridgette approached her teaching, and I felt it was really compelling,” Ms. Sheehan says. “Here was this underappreciated member of our intertidal coastal life, and Bridgette brought so much spark and curiosity to it. I thought, we need to learn how to care for our local seaweeds — especially at this time of ecological instability and climate change — and do that in a local way.”

Since it opened its doors in 2009, the Natural History Centre has built a strong relationship with the Hornby Island Community School which is right next door.

“The Natural History Centre came to me as a new teacher because a pilot program had just been launched to keep Grade 8 students on the island for school for one year longer,” Ms. Royal says. “We were discussing ways that we could work together and they came up with the idea of seaweed focused curriculum.”

“Seaweed School became a win-win,” says Ms. Royal. “It helps Bridgette with community outreach, the Natural History Centre gets to make connections with the community, and the students get to shine within their community. And as a teacher, I get to see the wealth of this knowledge that weaves into our curriculum, but surpasses things that you simply can’t do within the confines of the classroom.”

Museum pieces: a pressing issue

Accompanying Dr. Clarkston and her UBC Science students to Hornby Island for beach day was Keelie Taylor, a curatorial assistant at the Beaty Museum. Taylor, who tends the museum’s algal herbarium, set up a mobile pressing station to preserve museum-quality specimens.

“The Beaty Museum really pushes community outreach because so much of our collection comes from citizen scientists,” says Ms. Taylor. “We work with researchers and academics, but we also have a lot of people that come in as volunteers and work on our collection.”

On beach day, Ms. Taylor and the UBC Science students kept a spreadsheet for the collection data: where a specimen is from, who collected it when, and who identified it. Without consistent year-after-year baseline data it can be hard to know if species have disappeared, or if invasive species have been introduced. All of the seaweed specimens collected that day have the students’ names on the labels and serve as snapshots of the biodiversity of Hornby Island in May 2025.

Pressing seaweed is similar to pressing a flower, but uses thicker, more absorbent layers because of the seawater specimens contain. Special archival-quality paper is used that won’t degrade the specimen or yellow as regular printer paper might, since it’s not acid free.

“We want specimens that will still be of good quality 100 years from now,” says Ms. Taylor. “You want the features of interest to be visible. You want it to dry quickly and evenly, and you want it pressed so that it’s not going to catch or come off the page.”

The pressed specimens are attached to the paper with herbarium glue which can be reactivated with water, making it easier to lift the seaweed from the paper so it can be studied under a microscope or have DNA samples extracted.

As well as teaching Hornby Island students about seaweeds from a Western scientific perspective, Seaweed School adds Indigenous ways of knowing to the lessons.

“I’m from the Bay of Quinte Mohawks in Ontario, part of the Haudenosaunee and the Confederacy of Six Nations,” Ms. Royal says. “We’re not surrounded by ocean so I’ve had to do some research in terms of Indigenous knowledge. The Haudenosaunee look at all of creation as relatives — so here’s a seaweed relative that we haven’t paid much attention to.”

One of Ms. Royal’s goals going into the project was to learn alongside the kids.

“Generally, teachers like to be the authority on something. What I know from my Indigenous elders is that one of the beautiful places to be is in a place of humility. Just say: I don’t know. Let’s go on a journey together.”

Now in its second year, Seaweed School’s current theme is “the secret lives of seaweeds.” Dr. Clarkston, though, has plans to ensure the importance of B.C.’s seaweeds isn’t a secret. She has a dream to work with curriculum designers in UBC’s Faculty of Education to get seaweed education into B.C.’s K-12 classrooms in a more formal way.

“To me, it makes sense for everyone in British Columbia to be aware of seaweeds just like they learn about salmon. I want all kids in this province to learn about the roles seaweeds play in our coastal marine ecosystems and in our lives, from creating habitat for economically and culturally important fish species to contributing to the oxygen we breathe.”

Gardening with dirty undies

Alongside Western science and traditional ecological knowledge, community involvement makes up the third strand of Seaweed School’s “braided lens” approach.

Seaweed fertilizer does what common farming practices no longer do — help rebuild and regenerate soil through enhancing its microbiological activity and adding essential nutrients. It’s a common method used at Hornby Island’s community garden where clinical herbalist and Seaweed School collaborator Angela Willard guided the students through a project called Dirty Undies.

“The dirty underwear test is a way to gauge the microbiological activity of your soil, and demonstrate the connection between the health of soil and seaweed,” explains Ms. Willard. “You bury 100 per cent cotton underwear in your garden and the faster the underwear rots, the healthier the soil — because it’s biodegrading the cotton. Usually the underwear is buried for a 60 day window. We chose to do 90 days because we took into account the winter months are cooler, so the microbial activity is a little bit slower.”

Late last winter, three plots were set aside in the Hornby Island community garden to run the Dirty Undies test. On the first plot only leaf mulch was applied — the control treatment. On a second plot, large amounts of seaweed were stacked in piles on the soil in the traditional way. The third plot was dosed with seaweed tea, concocted by adding seaweed to a water base and letting it “steep”. As the mixture ferments, seaweed nutrients are extracted and enhanced in the tea.

The goal of Dirty Undies is to see which pair of underwear buried in the three plots breaks down fastest. Seaweed tea requires much less seaweed than mulch. If the tea plot wins out over the other two, it could reduce the amount of biomass removed by gardeners from Hornby Island beaches, allowing the seaweed to remain and replenish the habitats.

At the end of the experiment, there was a bit of cotton left in the control treatment, and a tiny bit left in the stacked seaweed treatment. No cotton remained in the seaweed tea treated undies.

“Everyone in our school wrote down which test plot they thought would decompose undies the fastest and why,” says Ms. Royal. “It created natural enthusiasm in the community garden to engage with the scientific method. It’s one of the ripple effects Seaweed School has — now we’re rippling into the community gardeners.”

“I’m the luckiest teacher in the world to have these experiences.”

Seaweed School is in session was originally published in Focus on Medium, where people are continuing the conversation by highlighting and responding to this story.

By Matthew Ramsey

In his short life, Kip the coyote with one bad leg became an unofficial mascot on UBC’s Vancouver campus.

He was a frequent sight, sleeping in the open, basking in the sun and hunting down squirrels. Kip’s casual limping strolls across the campus, when he seemed oblivious to throngs of people around him, were legendary.

Being hit by a car sometime after he was born in early 2020 caused the limp (first observed on a trail camera in November 2020), and Kip was ultimately struck and killed by a car near the Botanical Garden in October of 2024. As tragic as his death was, it provided UBC researchers a unique opportunity to learn more about his life.

Third-year PhD candidate Iris Li is focussing on the ecology and physiology of urban coyotes. The zoologist performed a necropsy on Kip at Beaty Biodiversity Museum after his body was retrieved. Kip’s remains are being respectfully preserved and he will stay in the museum so that future generations can continue to learn about and from him.

The results showed a not-wild-enough wild animal, likely with a bleak future. While massive internal injuries from the vehicle killed Kip, his body and observed behaviours prior to his death showed he quite possibly would have continued having interactions with people that were too close for comfort.

“At 16 kilograms, he was one of the heaviest coyotes I’ve ever seen,” said Li, who has performed 80 coyote necropsies in her research to date. “With the other coyotes I necropsied the average weight is 10 to 12 kilograms.”

Kip’s fat reservoir suggested he was supplementing his natural diet with human food, Li said, and the later isotopic analysis proved that theory. To learn what a coyote’s diet consists of, a whisker is analyzed from either a living or dead animal to find its carbon signature, which reveals what foods the animal has eaten.

“Food differs in its isotropic signature,’ says Dr. Sarah Benson-Amram, an associate professor in the department of forest and conservation sciences and department of zoology. “A lot of human-based food is corn-based, and corn is C4 plant that has a different Carbon 13C isotope signature compared to natural prey. By looking at the different ratios of stable nitrogen and carbon isotopes of these tissues, we can tell how much food is coming from marine sources, how much food is coming from human-based sources, how much is coming from more natural prey.”

Li has collected whisker samples from coyotes collared at UBC and their diets show similar trends to Kip’s. That doesn’t mean they rely solely on human food though.

“It could be their prey, because we also see UBC squirrels feeding on pizza and trash all the time,” Li says. “Those squirrels are also showing a carbon signature similar to that of anthropogenic food, so if a coyote is consuming those squirrels, their carbon signature is also going to leans towards anthropogenic sources.”

Notably, both Li and Dr. Benson-Amram had witnessed Kip eating garbage and being fed by people on several occasions.

“He’d lost his fear of humans,” said Dr. Benson-Amram. “He was close to dangerous interactions with area pets and people, for example attacking leashed dogs. If those conflicts continued there’s a possibility he would have had to be removed. He was definitely showing too much habituation to humans.”

The reason we do this work is because we love animals

By taking tissue samples at the Beaty Biodiversity Museum and then performing various biochemical analyses of Kip and other deceased coyotes, Dr. Benson-Amram and Li are able to determine the dietary strategies, body conditions, and chronic stress physiology of urban coyotes. That data can be used to inform outreach and management efforts.

Dr. Benson-Amram runs the UBC Coyote Project which has received funding from the B.C. Ministry of Water, Land and Resource Stewardship and the Habitat Conservation Trust Foundation to better understand the behaviour and ecology of coyotes living and traveling within Metro Vancouver.

In coyote society when an adult male and female mate, their offspring stay with them. Sometimes a family consists of parents, pups from the previous year, and current offspring, depending on the resources available in a territory. When the young disperse they have to establish a new territory, which can take them some time.

One young male coyote with a GPS collar was tracked travelling from UBC across Metro Vancouver to Burnaby.

“It’s pretty typical of these younger guys as they’re figuring out where they’re settling,” Dr. Benson-Amram says. “They move a lot, but as to what their adult home range size will be, we don’t know that yet.”

Coexisting with coyotes

“Kip’s death was obviously very sad. You watch these animals for years, you get attached to them,” said Dr. Benson-Amram. “The reason we do this work is because we love animals and we want to reduce the need to euthanize any animals. We want to help coyotes, people and pets co-exist safely in cities.”

Back when Kip was first observed at UBC, he was skittish around people thanks to the campus being largely empty during COVID-19.

At that time, experts provided the same advice they provide today when it comes to ensuring urban and campus coyotes coexist safely with people. That advice is especially relevant with the estimated five coyotes that are living primarily on campus now:

· Do not feed them! Coyotes do not need human food to survive.

· An injured coyote does not mean it will be an aggressive coyote, but it’s best not to approach it in case it feels threatened.

· It’s best to keep pets on leash in greenspaces and forests to prevent predation. Off-leash dogs may also charge at coyotes, forcing the wild animals to defend themselves.

· Keep trash and compost contained so it doesn’t attract coyotes.

If you see a coyote on campus the best guidance is to “haze” it — essentially scaring it so it doesn’t become too accustomed to being around people as Kip eventually became. Hazing can mean making a lot of noise, raising your arms to look big, throwing rocks near them, opening and closing an umbrella or using another kind of noisemaker to scare them away.

What a cherished campus coyote’s death tells us about his life was originally published in Focus on Medium, where people are continuing the conversation by highlighting and responding to this story.

The future of medicine is unfolding at the nanoscale. New imaging technology from UBC captures genomic medicine delivery in action, opening up avenues for drug discovery and testing.

By Geoff Gilliard

We are moving into a new era of genomic medicines that will target cancer, viruses and disease by editing our DNA and RNA. But the path to developing these drugs is not yet clear because of the incredibly small scale drug developers are working with — our cells — and the complexity of the activities within them.

“It’s a very crowded, busy, hot, dynamic environment inside our cells,” says Dr. Sabrina Leslie, a biophysicist at UBC’s Michael Smith Laboratories and department of Physics and Astronomy. “To understand how cells grow, divide and are repaired we needed to look at the interactions of the molecules within them in detail. And then at how various drugs interact with our genome at the single-cell level.”

To study molecules in cell-like conditions Dr. Leslie invented a high-resolution imaging platform called Convex Lens-induced Confinement (CLiC). CLiC lets drug designers actually see the intricate workings of a cell and how they respond to nanomedicines’ drug cargoes. In September 2025, she and her team took on a CEO to lead a company called ScopeSys to commercialize CLiC.

“We’re working towards a vision where CLiC can be available in the pharmaceutical and biotech industry to help develop new medicines at the nanoscale, and also in every university in the world,” Dr. Leslie says.

mRNA and lipid nanoparticles

In each cell’s nucleus, chromosomes hold DNA (deoxyribonucleic acid) which carries the blueprints for how a cell should grow. RNA (ribonucleic acid) also stores genetic information. It’s the messenger, hence the “m” in mRNA. It carries the blueprints from the nucleus to a cell’s “building site” — specialized compartments teeming with a biomolecular ballet of proteins, carbohydrates, polymers, lipids and nucleic acids, as well as vitamins and hormones, all held together within a cell wall.

When a cell is attacked by the antigen of a virus or bacteria (aka a pathogen), it triggers an immune response as the cell builds antibodies to fight the infection. Traditional vaccines deliver a weakened version of a pathogen into our bodies to train cells to recognize threats and quickly create antibodies. On the other hand, mRNA vaccines combat pathogens by delivering blueprints engineered by drug developers to cells with instructions on how to trigger a specific immune response. These vaccines, a form of genomic medicine, can be easily and rapidly engineered for any pathogen as long as the antigen is known.

In January 2021, as new mRNA vaccines targeting COVID-19 were rolling out, Dr. Leslie packed up her lab at Montreal’s McGill University and moved to UBC, her alma matter. Since 2018, after meeting and collaborating with UBC biochemist Dr. Pieter Cullis, a pioneer in the design of lipid nanoparticle (LNP) systems, she’s focused her lab on genomic medicines. LNPs developed by Dr. Cullis’ company Acuitas Therapeutics served as the delivery system for Pfizer’s Comirnaty mRNA-based vaccine that allowed pandemic lockdowns to be lifted and saved over 14 million lives.

To pass through a cell wall, mRNA must travel within a LNP that serves as a protective bubble. Drs. Cullis and Leslie conducted early-stage research on imaging lipid RNA particles to understand how different LNP formulations — their composition, size and drug payload — impacted their properties at the most fundamental level — the single particle — and in connection with what is observed in cells and tissues.

“The lipid nanoparticles that we’re dealing with are one 100th the size of a cell, which is the size that gets taken up into cells most readily,” Dr. Culllis explains. “If you draw a line on a piece of paper, that’s about 10 cells wide, so you get an idea of how small these things are. CLiC picks up a signal from the particle and lets you visualize it. Being able to detect particles that are smaller than the wavelength — you shouldn’t be able to see and track them for long times — but the CLiC approach is really quite remarkable.”

Since the COVID pandemic, UBC researchers have helped to expand the promise of genomic medicines to not only protect against future pandemics, but also to target cancer tumours, hereditary genetic disorders, and neurological and infectious diseases.

With these gene therapies we’re essentially using the body’s own processes to fight disease

“Anywhere a specific protein is missing in a patient, you could insert it with the right mRNA cargo,” Dr. Leslie says. “There’s thousands of medicines at the research stage — there are ones that edit the genome, and there are ones that cleave and remove a pathological part of the RNA.”

“You can pretty much go after any disease you care to name using this approach,” adds Dr. Cullis. “If a child is born not making a particular protein in the liver, well, let’s make some messenger RNA to code for it, and then inject that into the child and cure it of this genetic disease. With these gene therapies we’re essentially using the body’s own processes to fight disease. It’s very powerful and very focused and not very toxic.”

Dr. Cullis is also designing “triggered release” delivery systems to deliver common chemotherapy drugs for cancer treatments. Patients can suffer severe side effects from chemotherapy drugs, resulting in healthy tissue damage. But if drugs could be triggered to release their cargo at the site of a tumour, the medication would be applied only where it’s needed while leaving healthy tissue untouched.

“We have a big program going on at UBC where we’re developing cancer vaccines and it looks like we’re going to be able to go after metastatic cancers,” says Dr. Cullis. “This has been shown now for pancreatic, lung and breast cancer. By expressing various proteins in the body, you can craft responses that turn on the immune system to eliminate the cancer.”

“This is going to grow exponentially into the future, providing a shift towards personal medicine where we can optimize medicines for each person’s genome,” says Dr. Leslie. “If medicine is made of the same molecules that we’re made of, we’d be able to interface with our own biology more predictably than, for example, small molecule drugs such as penicillin or aspirin. But first we need to find out how mRNA interacts with a cell’s molecules in order to develop improved therapeutics that need fewer and smaller doses, cause fewer side effects, and are easier to manufacture and distribute at a global scale.”

CLiC

Biomolecules move very quickly, so they must be held in place in order to be imaged with CLiC’s sensitive detection. CLiC holds molecules between two thin sheets of glass in a cartridge containing row upon row of glass microwells that resemble an egg carton flat. The glass sheets are then compressed with a convex microscope lens that pushes down on the glass, trapping molecules in the microwells and isolating them for analysis.

Unlike previous imaging platforms that tethered molecules in place to keep them in focus, CLiC can confine single molecules in live, cell-like conditions for study without having to bind them to hard surfaces. Tethering molecules can twist them or prevent them from interacting normally, altering the results of the observation. It’s akin to tying up the fish when studying the life in an aquarium — you won’t get a realistic picture of how the fish interact if they can’t swim freely.

A series of images is taken with millisecond exposure

“We needed a way to confine those molecules in a way we could watch them for a long time and follow their steps,” Dr. Leslie says. “And because biology is so variable, we needed to repeat these experiments thousands of times on many different copies of the molecule of interest. The ability to see single cells and molecules means that you can make predictions about those fundamental units of life.”

Above the CLiC platform, microscopes and ultra-sensitive cameras record images of the activities taking place in the wells below. Images are captured in either snapshot or narrative mode. In a snapshot, a series of images is taken with millisecond exposure. Molecule and drug interactions are recorded by comparing snapshots taken over time. Narrative mode records the same molecule or set of molecules over a longer time which allows for measurement of the changing conditions in a cell as drug cargoes work to silence or edit a gene.

“An event like a drug molecule binding with its single molecule target needs to be imaged with very sensitive detectors in snapshot mode,” says Dr. Leslie. “Then there are the bigger single lipid particles containing mRNA or other drugs. They’re more like 100 nanometers in size. With these we use CLiC’s narrative mode to obtain very long tracks of their motion.”

The microscope configuration used with CLiC varies depending on whether the subject of observation is a single molecule, particles, or an entire cell. A confocal/wide field configuration is used for both single molecules and interacting particles, and for particles as they fuse with or react to a drug. Scanning confocal configurations are used to image whole tissues and cells, illuminating the sample with a small beam of light that shines through the cell to prevent “cooking” it with a wide field of illumination.

Much of that technical innovation in Dr. Leslie’s lab has been aided by in-kind support from Nikon, the vendor for the microscopes.

“To be relevant in terms of the scale that drug developers need, we have to run hundreds of samples in a fairly short amount of time,” Dr. Leslie says. “Students working in the lab on our hand-controlled systems aren’t fast enough. Nikon collaborated with us to install a high-content robotic arm in conjunction with their automated Ti2 and AXR microscope platform, which allowed us to dramatically increase our throughput to hundreds of samples a week, hundreds of thousands of species of particles or cells, which will be compatible with industry needs.”

A vision for ScopeSys

CLiC and ScopeSys were born out of a fairly mundane necessity: “When I started making initial measurements, scientists around me wanted the data from our assays,” explains Dr. Leslie. “Then when I was building my lab, professors wanted access to this data and copies of our tools in their own labs.”

But the technology could have very real impacts — including the potential to create research jobs for UBC students, as well as impact drug discovery development and manufacturing control. In return, UBC created an ecosystem of talent that provided a safety net as CLiC moves from research lab to the marketplace — where many ideas die before they’re born.

The new company’s first commercial product is Scope Top, which lets labs attach their microscopes to the CLiC imaging platform, and is being developed by UBC graduates who now work for ScopeSys.

“Scope Top transforms a simple microscope to a high powered one enabling multi-parametric, quantitative data and valuable insights,” says Dr. Leslie. “There are hundreds of thousands of microscopes out there where this could be dropped. It can also be a great data-feeding engine for training AI and machine language algorithms for drug discovery, development and manufacturing. I think we play in a very exciting space.”

When it comes to mRNA, seeing is believing was originally published in Focus on Medium, where people are continuing the conversation by highlighting and responding to this story.

How the discovery of the first major lithium-bearing pegmatite in B.C. — and the geochemistry behind it — could help make Canada more self-sufficient in critical minerals resources.

By Geoff Gilliard

The small town of Revelstoke is nestled in the steep slopes of the Columbia Mountains in southeastern British Columbia. Moist air from the Pacific cools as it sweeps up the mountainsides, forming some of the heaviest snowfall in Canada and making Revelstoke a renowned ski destination.

But soon the region may also be known as a source of the infinitely recyclable element that’s powering the global transition to electric vehicles: lithium.

In 2021, about seven kilometres northwest of Revelstoke, University of British Columbia (UBC) mineralogist Dr. Catriona Breasley made the first significant discovery of lithium in B.C. Perhaps more importantly, the newly minted PhD from the department of Earth, Ocean and Atmospheric Sciences found a way to identify lithium deposits below ground by examining rock outcrops on the surface.

Lithium is mostly found trapped in a rare type of rock called pegmatite. Pegmatite is derived from cooling magma and can contain a variety of minerals, making it surprisingly challenging to evaluate. Some of those minerals contain lithium: petalite and elbaite, a lithium-bearing tourmaline.

“Pegmatites are probably one of the most difficult deposits in the world to investigate, because we don’t understand much about them compared to other ore systems,” says Dr. Breasley, now a principal consultant at Criterra.

“The pegmatites I look at are LCT pegmatites, or lithium, cesium, tantalum-bearing pegmatites. A lot of minerals within them are white and look very similar, so it does take quite a trained eye to pick them apart.”

Proof of lithium in Prof pegmatite

Dr. Breasley’s supervisor, Dr. Lee Groat, has been studying pegmatites on Mount Begbie south of Revelstoke since 2010. In 2017 he reviewed a mining company assessment of a five- by 70-metre outcrop called the Prof pegmatite rock formation above the town on Boulder mountain.

“The report cover photo was of a beautiful spray of pink tourmaline crystals,” recalls Dr. Groat. “Suspecting that the colour meant that these were lithium-bearing tourmalines, I decided that we needed to go there and look at the pegmatite. When we got there Catriona pointed to some clear crystals with pink rims and a distinctive lustre and said that she thought they were actually petalite that the report hadn’t identified. Petalite contains a lot of lithium.”

Dr. Breasley confirmed her hunch back at UBC using X-ray powder diffraction, a process that reveals a crystal’s “fingerprint”. Over the course of three field seasons she collected 103 tourmaline samples from the Prof pegmatite.

“B.C. is a geologist’s dream, especially up in the alpine where there’s perfect rock exposures. When you find petalite mineralization in one pegmatite, the likelihood you’re going to find much more is quite high. There are hundreds of pegmatites in the area, which are fairly difficult to reach, places that could contain much more significant lithium mineralisation than what I found at the Prof pegmatite.”

Global demand for lithium soaring

Dr. Breasley’s discovery of the first major lithium-bearing pegmatite in B.C. could help make Canada more resilient and self-sufficient in critical mineral resources. Lithium has a high energy density and a high electrochemical potential, making it ideal for use in rechargeable vehicle batteries. Electric vehicle batteries account for 87 per cent of the total global demand for lithium.

Worldwide lithium production jumped by 30 per cent between 2021 and 2022. According to the International Energy Agency, the world will need around two billion battery-operated electric vehicles to reach net zero carbon emissions by 2050. Each electric vehicle’s set of batteries requires about ten kilograms of lithium. Lithium batteries also power our cell phones, laptops, tablets and other devices, plus grid storage.

The soaring demand requires not only finding new deposits, but developing more effective ways of processing the ore from which lithium is extracted. Canada currently produces lithium ore from two mines, located in Manitoba and Quebec, but no chemical processing plants. The few chemical processing plants that exist are all in China, although one is being built in Canada.

Crystalizing research

The Prof pegmatite formed around 49 million years ago through a process called “magmatic crystallization.” During this process, fluids containing lithium and other rare elements are incorporated into gigantic pegmatite crystals.

Pegmatites in the Revelstoke region are called zoned pegmatites, meaning they contain distinct mineralogical and textural groups within the rock. Like the layers of an onion, they hold the least exotic minerals at the outer edges and the most exotic in the middle.

Dr. Breasley focussed on analyzing tourmaline, which captures snapshots of the pegmatite geochemistry as it evolves and becomes more enriched.

“I discovered that in the wall zones of the pegmatites — which we think of as having a quite boring mineralogy that doesn’t usually have a lot of rare element mineralization — completely clear and colourless secondary minerals formed which record elevated lithium concentrations.”

“We’re at a really exciting time where there’s so much potential for exploration in the region”

The discovery means that rather than relying solely on what is visible above ground, geologists can use tourmaline as a chemical roadmap in their search for lithium. Because pegmatites in a given region often share mineralogical traits, the geochemistry recorded in the Prof pegmatite tourmalines can serve as a guide for exploration across the wider area.

“We can see a hundred or more pegmatite intrusions in a 15-kilometre band running from Mount Begbie to Boulder Mountain,” says Dr. Groat. “However, very few show lithium minerals on the surface — and the Prof pegmatite is the only one that has petalite. What Catriona found was that the chemistry of the tourmaline can tell you if an intrusion is likely to contain lithium at depth. This is important because the surface expression of a pegmatite dike is often small.”

Geologists can now use Dr. Breasley’s techniques to improve lithium exploration and gauge a pegmatite’s profitability on the spot.

“We’re at a really exciting time where there’s so much potential for exploration in the region, and furthermore, there could be much more mineralization at depth as well, because no drilling has ever taken place,” Dr. Breasley says. “There could be much bigger deposits out there. We just haven’t found them yet.”

Will B.C. get a charge from lithium? was originally published in Focus on Medium, where people are continuing the conversation by highlighting and responding to this story.

By Dr. David Poole, Professor Emeritus, Computer Science, UBC

In 1948, the founder of information theory, Dr. Claude Shannon, proposed modelling language in terms of the probability of the next word in a sentence given the previous words. These types of probabilistic language models were largely derided, most famously by linguist Dr. Noam Chomsky: “The notion of ‘probability of a sentence’ is an entirely useless one.”

In 2022, 74 years after Dr. Shannon’s proposal, ChatGPT appeared, which caught the attention of the public, with some even suggesting it was a gateway to super-human intelligence. Going from that proposal to ChatGPT took so long because the amount of data and computing time used was unimaginable even a few years before.

ChatGPT is a large language model (LLM) learned from a huge corpus of text from the internet. It predicts the probability of the next word given the context: a prompt and the previously generated words.

ChatGPT uses this model to generate language by choosing the next word according to the probabilistic prediction. Think about drawing words from a hat, where the words predicted to have a higher probability have more copies in the hat. ChatGPT produces text that seems intelligent.

There is a lot of controversy about how these tools can help or hinder learning and practising creative writing. As a professor of computer science who has authored hundreds of works on artificial intelligence (AI), including AI textbooks that cover the social impact of large language models, I think understanding how the models work can help writers and educators consider the limitations and potential uses of AI for what might be called “creative” writing.

LLMs as parrots or plagiarists

It’s important to distinguish between creativity by the LLM and creativity by a human. For people who had low expectations of what a computer could generate, it’s been easy to assign creativity to the computer. Others were more skeptical. Cognitive scientist Dr. Douglas Hofstadter saw “a mind-boggling hollowness hidden just beneath its flashy surface.”

Linguist Dr. Emily Bender and colleagues described the language models as stochastic parrots, meaning they repeat what is in the data they were trained on with randomness. To understand this, consider why a particular word was generated. It’s because it has a relatively high probability, and it has a high probability because a lot of text in the training corpus used that word in similar contexts.

Selecting a word according to the probability distribution is like selecting text with a similar context and using its next word. Generating text from LLMs can be seen as plagiarism, one word at a time.

The creativity of a human

Consider the creativity of a human who has ideas they want to convey. With generative AI, they put their ideas into a prompt and the AI will produce text (or images or sounds). If someone doesn’t care what is generated, it doesn’t really matter what they use as a prompt. But what if they do care about what is generated?

An LLM tries to generate what a random person who had written the previous text would produce. Most creative writers do not want what a random person would write. They want to use their creativity, and may want a tool to produce what they would write if they had the time to produce it.

LLMs don’t typically have a large corpus of what a particular author has written to learn from. The author will undoubtedly want to produce something different. If the output is expected to be more detailed than the input, the LLM has to make up details. These may or may not be what the writer intended.

Some positive uses of LLMs for creative writing

Writing is like software development: Given an idea of what is wanted, software developers produce code (text in a computer language) analogously to how writers produce text in a natural language. LLMs treat writing code and writing natural language text the same way; the corpus each LLM is trained on contains both natural language and code. What’s produced depends on the context.

Writers can learn from the experience of software developers. LLMs are good for small projects that have been done previously by many other people, such as database queries or writing standard letters. They are also useful for parts of larger projects, such as a pop-up box in a graphical user interface.

If programmers want to use them for bigger projects, they need to be prepared to generate multiple outputs and edit the one that is closest to what is intended. The problem in software development has always been specifying exactly what is wanted; coding is the easy part.

Generating good prompts

How to generate good prompts has been advocated as an art form called prompt engineering. Proponents of prompt engineering have suggested multiple techniques that improve the output of current LLMs, such as asking for an outline and then asking for the text based on the original prompt augmented with the outline.

Another is to ask the LLM to show its reasoning steps, as in so-called chain of thought. The LLM outputs don’t just the answer a question, but explains the steps that could be taken to answer it. The LLM uses those steps as part of its prompt to get its final answer.

Such advice is bound to be ephemeral. If some prompt-engineering technique works, it will be incorporated into a future release of the LLM, so that the effect happens without the need for the explicit use of the technique. Recent models that claim to reason have incorporated such step-by-step prompts.

People want to believe

Computer scientist Dr. Joseph Weizenbaum, describing his ELIZA program written in 1966, said: “I was startled to see how quickly and how very deeply people conversing with (the program) became emotionally involved with the computer and how unequivocally they anthropomorphized it.” The tools have changed, but people still want to believe.

In this age of misinformation, it is important for everyone to have a way to judge the often self-serving hype.

There is no magic in generative AI, but there is lots of data from which to predict what someone could write. I hope that creativity is more than regurgitating what others have written.

Dr. David Poole is Professor Emeritus of computer science at the University of British Columbia

This article is republished from The Conversation under a Creative Commons license. Read the original article.

Why AI can’t take over creative writing was originally published in Focus on Medium, where people are continuing the conversation by highlighting and responding to this story.

By Geoff Gilliard

Finding a needle in a haystack is child’s play compared to discovering new molecules in space.

But last October, Dr. Ilsa Cooke and her team at the University of British Columbia helped do just that when they, along with collaborators, discovered cyanopyrene in the Taurus molecular cloud 430 light years from Earth.

“Pyrene is a polycyclic aromatic hydrocarbon (PAH) — we think PAHs may make up more than 20 per cent of the carbon in space — and life is made of carbon,” says Dr. Cooke, who heads UBC’s Astrochemistry Laboratory. “By learning more about how PAH molecules form and are transported through space, we learn more about our own solar system and the life within it.”

So far Dr. Cooke and other researchers have uncovered 10 PAHs in space — but the field moves fast. Carbon — the fourth most common element in the Universe — exists in stars, asteroids and comets, as well as the atmosphere of most planets. It’s also the second-most abundant element in our bodies by weight. When Joni Mitchell sang “We are stardust” she wasn’t just being poetic.

“There’s nitrogen, carbon and hydrogen in space, all of which are common in organic molecules on Earth,” Dr. Cooke says. “Since we now know that many organic molecules can form in space, the question now is: can the most prominent building blocks of life — amino acids, nucleobases and sugars — form there as well? Because if they’re in space, they could be on other planets, too.”

Space: The most extreme chemical laboratory

The Taurus molecular cloud (TMC-1) is the closest large star-forming region to Earth, although in this case close is relative. It’s about four quadrillion kilometres distant — four followed by 15 zeroes.

Molecular clouds began forming from simple mixtures of atoms and molecules — mainly hydrogen — a couple of billion years after the Big Bang. As atoms collided, they joined together to form more complex compounds including PAHs. Over eons, gas and ice-covered dust grains were drawn together by gravity, densifying as they accumulated into protostars, then evolved into stars and their surrounding protoplanetary discs. Eventually, the systems’ stars ran out of fuel and died.

“We expect that many of the atoms contained in the molecules that are found in molecular clouds were once within stars and were cast out by supernovae or other means,” says Reace Willis, a graduate student with the Cooke group. “They then populated dense regions of space, and now they’re coming together to form the chemistry that we’re witnessing. We want to know whether we can find the same sorts of complex organic molecules that we see in TMC-1 in other regions of the interstellar medium — the gas and dust in the space between stars — and we already have for many of these molecules.”

“Pyrene might form within the molecular clouds itself, or perhaps it is more resilient than we previously thought and can survive its journey through space.”

Astrochemists don’t yet fully understand all the forms PAHs take in the interstellar medium, how they form or how they’re transported to solar systems like our own.

“The fact that we see pyrene in the cold molecular cloud TMC-1 is surprising,” says Dr. Cooke. “Traditionally, PAHs were thought to form in hotter environments and then travel through space to reach the molecular clouds. But pyrene could not survive this journey, it would get blown up by ultraviolet light. This suggests that pyrene might form within the molecular clouds itself, or perhaps it is more resilient than we previously thought and can survive its journey through space.”

Part of that mystery is that while some PAHs that have been identified in space are building blocks of life, on Earth many PAHs are carcinogenic, formed in hot environments, both from natural sources (volcanoes, wildfires) and human activities (burning fossil fuels, cooking). In TMC-1, where pyrene has been found, the temperature is about 10 Kelvin (-263 degrees Celsius).

The cold, harsh conditions of space mean that molecules can exist there that would not survive on Earth because they would be too unstable to survive exposure to our atmosphere. Small molecules such as carbon monoxide or methane form fairly easily because only one or two collisions are needed, whereas several reactions are needed to form the larger molecule pyrene from atoms and smaller molecules. In TMC-1 these collisions are infrequent because of the large space between atoms and molecules.

“On Earth those molecules would be removed quickly by collisions with other molecules in our atmosphere,” says Dr. Thomas Speak, a post-doctoral researcher in Dr. Cooke’s astrochemistry lab. “Although TMC-1 is defined as a dense molecular cloud, its ‘density’ is relative — approximately one thousandth of a trillionth of Earth’s atmosphere at sea level. The ratio of the density of gas in TMC-1 and Earth’s atmosphere is equivalent to comparing one drop of water and 25,000 Olympic sized swimming pools.”

GOTHAM

Dr. Cooke is co-principal investigator of GOTHAM, a collaboration of astronomers and astrochemists who study the Universe using radioastronomy to listen for the radio signals from stars, planets and molecular clouds. GOTHAM is an acronym for Green Bank Telescope Observations of TMC-1: Hunting Aromatic Molecules. The telescope is so sensitive that its site in West Virginia sits in a federally mandated 34,000-square kilometre quiet zone where cell phones, wifi routers and iPads are banned so their signals don’t interfere with the whispers of the Universe.

Radioastronomy can detect individual molecules millions of light years away because a molecule can give off energy as it rotates around its centre of mass. That rotational energy is released as a radio signal when the molecule changes the way it rotates.

“It’s like a fingerprint. The exact frequency is specific to the molecule, so we can tell different molecules apart.”

“These rotational emissions are really low energy,” says Dr. Speak. “If you take all of the radio observations that have ever been carried out on Earth, and you summed up the energy of all of the photons in the radio waves that are being collected, it would be less energy than one snowflake hitting the ground.”

So far, each of the more than 300 molecules detected in space by radioastronomy has a unique electromagnetic signal depending on its mass and the bonds within the molecule.

“It’s like a fingerprint,” Dr. Cooke says. “We’re looking at rotations at a specific wavelength range. The exact frequency is specific to the molecule, so we can tell different molecules apart.”

GOTHAM began searching for PAHs in TMC-1 after the Japan Aerospace Exploration Agency collected samples from the near-Earth asteroid Ryugu that contained large amounts of pyrene.

“The pyrene in Ryugu was shown to be a sample of what’s made in the cold regions of space, which supports the notion that solar systems inherit carbon molecules that form long before their birth,” Dr. Cooke says.

Pyrene molecules, however, can’t be detected by a radio telescope.

“The four fused aromatic rings of pyrene molecules are symmetric so there is no overall change as it spins around its axis, meaning it doesn’t emit radio frequencies that we can detect,” Dr. Speak explains.

So instead, GOTHAM researchers listened for cyanopyrene — pyrene with a carbon and nitrogen atom attached, substituting one of its hydrogen atoms.

Carbon and nitrogen have different abilities to grab hold of the electrons being shared in that bond, with nitrogen atoms able to pull slightly more towards them. Adding the nitrile group (CN) to the parent pyrene molecule allows the cyanopyrene molecule to emit radio waves when it rotates.”

Creating space in a lab

To determine whether small PAHs can survive the journey across the Universe from a circumstellar envelope to molecular clouds, the UBC astrochemists are investigating whether PAHs form from the bottom up or the top down.

“Do the molecules start as small hydrocarbons that react with each other to form larger structures?” Dr. Cooke says. “The other hypothesis is top down — are large species of PAHs, or even fullerenes, blown apart, leaving just fragments?”

To try and answer that question, the lab is building instruments to investigate reactions that could occur in TMC-1, and explain or predict the presence of molecules such as pyrene. In one of these instruments, a flow cell apparatus, gas is flowed into a reactor to study the reaction of interest. Then a laser shines ultraviolet radiation into the chamber to start a chemical reaction. The researchers will use a mass spectrometer to probe the reactions’ reliance on physical conditions like temperature and pressure to understand if they possibly take place in the interstellar medium. The reactions will occur in a matter of microseconds or milliseconds, depending on the complexity of the molecules and the temperature, although the experiments will monitor these processes on the seconds or minutes timescale.

“The main things we’re interested in learning from the experiment are, does a reaction happen, if it happens how fast does it happen and at what temperature, what does the reaction make and in what proportions,” says Dr. Speak.

The findings will be entered into an astrochemical computer model containing thousands of reactions. Starting with very simple atoms, the reactions will be run through simulations to learn how the molecules evolve over time.

By understanding how the reactions occur over a range of temperatures and pressures, Dr. Cooke’s team will come one step closer to solving the mystery of PAH formation in the interstellar medium.

With files from Alex Walls, UBC Media Relations

How to listen for the building blocks of life in space was originally published in Focus on Medium, where people are continuing the conversation by highlighting and responding to this story.

‘Peripheral vision’ helped Dr. Julia Levy co-found one of UBC’s most successful spinouts. The renowned immunologist passed away in December.

By Chris Balma

By the early 1970s, Dr. Julia Levy had a well-established lab up and running at UBC, studying antibodies for treating various forms of cancer. Outside her lab, she was enjoying recently purchased property on Sonora Island with her family.

“The vegetation around the cabin included an invasive plant called cow parsnip,” Dr. Levy recalls in her 2020 memoir In Sight: My Life in Science and Biotech.

“I noticed that sometimes the children would develop redness in patches on their bodies … We figured these patches must be caused by some kind of plant sap, but there seemed to be no pattern of occurrence so we couldn’t draw any conclusions.”

Solving that mystery would draw Dr. Levy — who passed away last December — into new research and professional territory. She learned that a substance in cow parsnip could indeed attack skin cells — but only when exposed to sunlight.

That led to an interest in the potential of photo-activated drugs — a mechanism that might allow Dr. Levy and her colleagues to better deliver the cancer treatments they were developing to targeted tissues and blood vessels. Photodynamic therapy uses lasers to activate a therapeutic in proximity to abnormal cells and tissue in a specific area of the body.

While not enamoured with the prospect of a career in business, Dr. Levy quickly recognized the commercial potential presented by the combination of the therapeutics her team was working on and the light-activated delivery mechanism.

In 1981 Dr. Levy cofounded QLT, one of B.C.’s most impactful early biotech startups, in partnership with UBC chemist Dr. David Dolphin. As well as cancer treatments, their work would also save the vision of hundreds of thousands of patients diagnosed with macular degeneration, an age-related eye disease.

“[Julia] showed that as an academic, it is possible to successfully collaborate and partner with the pharmaceutical and biotechnology industry to bring new products to market,” said Dr. Helen Burt, former Associate Vice President of Research and Innovation at UBC. “She definitely inspired me to want to be an innovator in terms of developing controlled and targeted drug delivery systems.”

QLT developed the photodynamic therapy drug Photofrin for a range of cancers — bladder, esophageal, and non-small cell lung cancer. Later, Dr. Levy and QLT leveraged much the same delivery system to launch the drug Visudyne, at that time the only treatment for age-related macular degeneration.

Visudyne took a decade to develop, but was approved by the FDA in 2000. It was an immediate success — the largest ophthalmic product launch on record with sales of $150 million in the first 12 months.

Much like the inspiration behind harnessing Photofrin, Visudyne had a personal connection. Dr. Levy’s mother, who enjoyed embroidery, reading and television, was diagnosed with macular degeneration in the 1980s.

“She had the wet form of age-related macular degeneration,” Levy recalls in Insite. “She would lose her central vision, and there was no treatment other than a laser treatment that usually worsened your vision and didn’t always halt the continued deterioration of the eye’s macula. She was stoic, as she always was, but I knew she was devastated. As was I.”

Dr. Levy left QLT in 2002, after 21 years with the company. Visudyne had placed the company on a solid footing, and royalties from the company also provided financial returns to support teaching and research at UBC. Eventually QLT would merge with the U.S.-based Atrix Laboratories in an effort to diversify.

In her remarkable career Dr. Levy left an indelible mark as a researcher, mentor and reluctant (she called it “accidental”) entrepreneur. Despite her success in research and business, she often cited mentorship as one of the most enjoyable aspects of her career.

“She was an original and to see what she had achieved in that era, it was amazing,” said Dr. Stephen Yip, a former PhD student of Dr. Levy, and a clinician-scientist at UBC, Vancouver General Hospital, and BC Cancer. “She imparted that spirit to all of her trainees. She was supportive of my entry to the combined MD-PhD program, which at that time was still very much a novelty.”

Modestus Obochi, a former PhD student of Dr. Levy and chief business officer of the biotechnology company Sernova, described her immense impact.

“As a mentor, Julia was incredibly generous with her time, helping me approach challenges from every angle and think through solutions. But she was more than just a mentor she was a guiding light, a role model, and someone I cherished like a mother and a best friend. She was my go-to person for every major career decision I made, and I cannot begin to imagine who I would have become without her profound influence.”

After retirement, Dr. Levy remained an active advisor to the biotechnology community and provided guidance to new life science startups through UBC’s Sauder School of Business.

In 1981 she was appointed to the Royal Society of Canada, and in 2001 to the Order of Canada. The Chemical Institute of Canada’s Julia Levy Award for successful commercialization of innovation in the field of biomedical science and engineering was launched in her honour in 2006.

With files from Sarah Anderson, UBC Microbiology and Immunology.

Visionary: Celebrating the legacy of a B.C. biotech pioneer was originally published in Focus on Medium, where people are continuing the conversation by highlighting and responding to this story.

Computer scientists and linguists at UBC are using AI to identify disparities across translations of biographies of LGBT-identifying public figures

By Geoff Gilliard

Today — for better or worse — the world finds its facts on Wikipedia.

The self-described “largest and most-read reference work in history” receives more than 1.5 billion visits a month. It contains 64 million articles in over 300 languages — with the English version leading the way at over seven million articles.

But with content created and edited by volunteers, the possibility of varied versions of the truth — whether intentional or not — looms large. And now that ChatGPT and other language models are being trained on Wikipedia, its reliability matters more than ever.

Enter InfoGap — an AI tool developed at UBC, Johns Hopkins and the University of Washington to help Wikipedia editors identify biases across languages. The research team used InfoGap to analyze 2,700 biography pages of lesbian, gay, bisexual, transgender (LGBT) people on English, Russian and French Wikipedia.

InfoGap Findings

· Articles on the same topic vary substantially across language versions of Wikipedia

· On average, 34 per cent of the content in Russian Wikipedia pages about public figures is not present in the corresponding English page

· 50 per cent of the negative facts contained in Russian versions of LGBT biographies are shared with their English versions — only 38 percent of negative facts in non-LGBT are shared with English versions

· Non-English language versions have a significant amount of unique information — a “local heroes” effect where biographies of people whose nationality matches the language of the article tend to have greater depth

We spoke with UBC computer scientist and linguist Farhan Samir about how advances in AI made InfoGap possible, the nature of discrepancies in Wikipedia, and how the InfoGap team is working with the Wikimedia Foundation to improve content accuracy.

How did AI make your research possible?

Studies from 2012 said detailed cross lingual comparisons weren’t possible because they were too complex. But skip ahead 10 years and our AI models have gotten so good at handling multilingual text that they can tell us the similarity between a piece of an English sentence and a Russian sentence. We can find exact facts that were in one article but not the other.

For example, in Brittney Griner’s biography the first sentence covers when she was born, that she was a prolific WNBA player as well as a college basketball player. All in one sentence. We deconstruct the sentence into those individual predicates: she was born on October 18, 1980, she was an accomplished college basketball player, and she played for a WNBA team. Those are three separate facts.

InfoGap deconstructs this into its constituent predicates, and gives you that list of individual facts. Then you can ask, is this fact available in Russian? Is the next fact available in French? And is the next one? You couldn’t do that with the whole sentence because part of the sentence may be different. It’s hard to get that level of granularity.

Why did you choose to compare French, Russian and English language articles?

We thought there’d be discrepancies between English and Russian based on previous research that found differences in the average positivity of sentiment in Wikipedia biographies of LGBT-identifying public figures. We wanted to do a more detailed comparison. Also, we had English, French and Russian speakers on our team which made it easier to validate that InfoGap was working correctly.

Were the Wikipedia biographies flagged by InfoGap translated by humans or AI?

I think this is a surprise for a lot of people, but Wikipedia is pretty vigilant about using articles automatically translated from the English version into another language. They explicitly say in their guidelines for translation that they would prefer to have no content at all as opposed to a poor translation. They’re quite stringent about human supported curation. It’s a combination of translated content and editors editing the language version that’s most familiar to them.

There’s a pretty big correlation between the language version and the region where the editor is from. So you get a lot of contextualized edits. It’s not a universal translation between different language versions of articles.

Did you choose Brittney Griner as an example of InfoGap’s findings because she had been imprisoned in Russia and because of her sexual orientation?

Not from the start. I was looking through my data at which pages had the biggest disparities. Her page percolated to the top and was a really clear example of how politically polarized national variants could show up in Wikipedia language versions.

In a previous paper my collaborators used the example of Alan Turing, who is maybe the most famous computer scientist ever. Turing was chemically castrated when he was found guilty of homosexual acts in 1952. The disparities between different language versions of Wikipedia describe different levels of agency he had. The English page portrayed him with of a lack of agency, whereas the Russian page portrayed him with much greater agency and navigation of his persecution.

How does InfoGap measure sentiment differences in a biography?

We focused on the omission of content with positive implied sentiment from one language version. For example, Chelsea Manning’s French page describes praise for her whistleblowing during the Afghanistan war. The French page also discusses her whistleblowing on conditions in Abu Ghraib prison in Iraq. Conspicuously, both events are omitted from the English page, despite the English page being longer. American perception of this instance of whistleblowing skewed negative, which may have played a role in the disparities between English and French pages.

I imagine in some countries, there could be political pressure for the editors to toe the government line.

That’s entirely possible and our work could help facilitate analyzing that. We had some interesting examples in our paper of possible politically oriented motivations. The Russian Wikipedia biography of Apple CEO Tim Cook mentions his association with the Russia-Ukraine war when he told Apple employees that he would triple any donations they made to support Ukraine. That was only in the Russian Wikipedia language version, not in the English one. There might have been some intention to keep his English Wikipedia page uncontroversial due to the partisan divide on U.S. involvement in that war. It may not have been included in English to maintain a veneer of neutrality.

Have you had any feedback from Wikipedia?

We’re in touch with the Wikimedia Foundation. A scientist there who looks specifically at data gaps in different language versions. We’re collaborating on approaches to facilitate information sharing across language versions, so that readers around the world can get the most complete set of facts about the topics that matter to them.

When the discrepancies in the language text are identified and Wikimedia is notified, have you found that they are remediated?

Our aim was just to identify differences, because they’re hard to notice. A comparative analysis takes quite a lot of time if you were to do it manually.

I think it should be up to the editors of the different language versions to decide whether they want to remediate or not. We didn’t have any prescriptive norms for what should and should not be there. Each language community for Wikipedia has some level of autonomy to set the edit norms for that language version. There’s autonomy for community level norms as well as for individual editors to figure out what they want to add and what they want to remove.

Where else could InfoGap be useful?

Comparing the framing of geopolitical conflicts in news articles. You could look at the media coverage on CNN and compare that to RT News (the Russian 24/7 English-language news channel) and see how they’re portraying the conflict, what sort of events they’re focusing on, or how they’re highlighting the agents.

What’s next for InfoGap?

The web also has an Anglocentric design to it. A lot of the text on the web comes from a particular demographic that tends to be well educated, male and white. Wikipedia reflects that demographic. It’s certainly a concern.

We’re thinking about how to enrich other Wikipedia language versions, especially those that don’t have as many contributors. How can we use InfoGap to point out exactly what the data gaps are, and then contact editors likely to speak the language and contribute high-quality edits?

When I go on Google and search for the former French Prime Minister Gabriel Atal, one of the first search results is a link to his English Wikipedia page. You also get a Google sidebar pulled from English Wikipedia. InfoGap can highlight regional figures of national prominence — Gabriel Atal is a French national — and that there’s more information in his French bio.

We want to point out to readers who are reading about a topic on English Wikipedia, that they might be interested in French Wikipedia because it contains information that’s not available on the English page. We’re charting out what that looks like with the Wikimedia Foundation. It’s an interface design problem as well as a computational AI problem.

Mind the InfoGap: Uncovering cultural bias in Wikipedia was originally published in Focus on Medium, where people are continuing the conversation by highlighting and responding to this story.

By Dr. David R Boyd, Claire Ewing, Dr. Amanda Giang, and Nivretta Thatra

Millions of kilograms of toxic pollutants, over 17,000 deaths annually and environmental laws that aren’t being diligently enforced. This is the troubling picture that emerged when we, a group of environmental researchers, investigated trends in air pollution enforcement in Canada.

Federal and provincial governments share responsibility for regulating air pollution. However, environmental laws and regulations are only useful if they’re properly enforced. Our research shows Canada needs to take greater action in enforcing the widely endorsed “polluter pays” principle for air pollution. According to this principle, those who produce pollution should pay for cleaning up any environmental damage.

We built a publicly-available dataset in Canada of air pollution enforcement actions. We scoured all available sources, creating a database of more than 2,200 enforcement actions that took place between 2000 and 2020 from eight provinces as well as the federal government. This helped us identify patterns in the way air pollution laws were being enforced.

Broken rules

One of the disappointing patterns we saw is that the majority of enforcement actions in our dataset—63 per cent—were against individuals for offences such as illegal campfires. Meanwhile, only one-third of enforcement actions were brought against companies—including those that had dumped vast volumes of toxic substances into Canada’s air, or caused catastrophic emissions offences (such as the Toronto Sunrise Propane explosion).

Even in the uncommon cases where rules were enforced against large corporations, the penalties imposed were extremely lenient. These penalties amounted to barely a slap on the wrist for repeat industrial polluters.

For example, the mining corporation Rio Tinto was fined $150,000 in Québec for breaking air pollution laws in 2013. This fine equated to only 0.00023 per cent of the company’s annual revenue. To put this into perspective, if a Canadian family earning the average income of $62,900 after taxes was given a 0.00023 per cent fine, this would equate to $14.47.

It’s not surprising, then, that this company would go on to violate air pollution laws again less than one month later. They also violated these laws again in 2016 and 2019.

Even the relatively small fine of $150,000 is well above the median fine for industrial air polluters. According to our study, fines ranged from $2,500 to $10,000 for most types of offences—including excess emissions or violating an environmental standard. This is less than some people would be fined for driving with a suspended licence. These fines are less than one per cent of the maximum penalties permitted by law for environmental offences—which range up to $12 million.

Another concerning pattern our study revealed is that some large industrial polluters are repeat offenders. While government policies indicate there should be increasingly strict enforcement applied in these cases, this doesn’t appear to be the general practice. Instead, chronic law-breakers tended to receive multiple warning letters—not increasingly large fines or prosecutions.

For instance, over the last five years, four provincial orders were reportedly issued against INEOS—one of the world’s largest chemical production companies. These orders were issued so the company would address its benzene emissions. This toxic chemical is linked with cancer.

Following federal and provincial orders to reduce benzene emissions in 2024, INEOS decided to close the offending facility. The company was never fined for its toxic pollution.

Enforcement actions don’t seem to be taking into account the way human and environmental health are jeopardized by industrial air pollution. Vulnerable or marginalized groups who live near large industrial facilities are particularly at risk of harm.

We found that on average, businesses were generally fined less for committing an actual pollution violation—such as illegally dumping large quantities of contaminants into the air—than they were for failing to notify an enforcement agency that they’d committed a violation.

Improving environmental enforcement

Canadian enforcement agencies are failing to properly hold high-risk offenders and repeat offenders to account.

But positive change is possible. Going forward, there are three key actions enforcement agencies should take:

• Increased penalties: Polluters should pay for their pollution. The consequences of breaking the law should be proportional to the risks to public health and the environment. Substantial mandatory minimum fines should replace the current practice of warning letters and grossly inadequate tickets or fines.
• Transparency: The public should have access to environmental information through standardized data on air pollution violations. This information should include who committed the violation, the details and location of what occurred and what was done about it. Ideally, the federal government would co-ordinate and publish all environmental enforcement data from across provinces the same way it publishes a national inventory of industrial pollutant releases.
• Focus on high-risk offenses: The focus of enforcement actions should be on high-risk offenders, such as super-polluters (businesses that produce disproportionate volumes of air pollution relative to their competitors), repeat offenders and inter-provincial offenders.

By properly enforcing environmental regulations, Canada can protect the public from the perils of poor air quality. This would also be a vital step towards realizing everyone’s right to a healthy environment—a right that was recently recognized in an amendment to the Canadian Environmental Protection Act.

Coauthor credits: Claire Ewing, Senior Climate Action Specialist, completed her Master’s of Science in Resources, Environment and Sustainability (IRES) at UBC in 2021. Nivretta Thatra is the Communications Manager with IRES.

This article is republished from The Conversation under a Creative Commons license. Read the original article.

Big corporations are getting away with catastrophic air pollution—putting Canadians at risk was originally published in Focus on Medium, where people are continuing the conversation by highlighting and responding to this story.