This is a machine transcription and therefore it may contain inaccuracies, errors, or mispronunciations. Notice an error you think needs changing? Please contact the Bitesize Bio team using this form: https://bit.ly/bsbtranscriptions

Welcome to the Microscopists,

a bite-sized bio podcast hosted by Peter Oto,

sponsored by Zeiss Microscopy. Today on the Microscopists.

Today on the Microscopist,

I'm joined by Beth Chimney from the Broad Institute of m i t and Harvard.

And we discuss why harsh peer reviews

Leads to user-friendly software tools,

Because we want tools to go out there in a state where someone can use them,

where they're user-friendly,

The difficulties of developing popular

Biology software tools in academia.

So when I started, um, in the lab,

cell profiler did not actually have any dedicated funding at all.

It was basically funded by running a contract image analysis fee for service

thing,

and then hoping that that made enough money on the side that you could hire a

software engineer to do cell profiler maintenance.

And what makes people

Mad at scientists on Twitter?

The most mad I'd ever made someone on Twitter, um,

was I had made a giant thing of pasta sauce to make the lasagna,

and I had a very full pot of pasta sauce. And someone was like,

you don't deserve to run a lab if you keep food containers that full on the

stove. What is like, what is wrong with you?

Like, why is that? It was like, oh, well, you're a computationalist,

so that's why I guess we won't take your live away, or something like that.

And I was like,

that was a very strange reaction to someone posting a picture of food.

Oh, in this episode of the Microscopist,

Hi, welcome to the Microscopist. I'm Peter Oton from University of York,

and today I'm joined by Beth Sim from the Broad Institute of M I T and Harvard.

Beth, how are you today?

I'm doing great things. How are you?

I'm really good, thank you. I I Where to start? Actually,

I'm going to start on a slightly different note, not personal. 'cause uh,

we recently published, so this isn't a plug for my publication in Nature Ations,

by the way.

It's okay. Finish.

Beth actually declared herself as one of the reviewers. Now,

we weren't aware of that when it's being reviewed, but afterwards I,

I noticed Beth's thing on, on the publication. Beth, how, how, how,

why declare yourself to start with?

Um, I think, so one of the reasons, um, that I've done that I don't def,

I definitely don't do that on every paper.

I've had some people who do it on every paper,

and they've reviewed my papers a few times and I see the same name over and over

again. Um, part of the reason on, on this paper that I,

that I really wanted to declare myself was just, um,

I felt like I had asked you guys for a lot and I felt like you guys had done a

lot, and I wanted to sort of like,

show my admiration and sort of come out publicly that I thought that with all

the improvements you guys have made to the paper,

that it was really amazing and that I wanted to sort of be on record as saying

that you guys had put together something really cool

That, that, that, so that's really kind. Thank you. I guess say,

so for the list is, uh,

publication reviewing as a someone who's pub publishing

and you get the reviewer's comments, you look at them, you go, oh,

I've got a duet. Hi Zed. Mm-hmm. Uh, I've gotta say, I,

I think that we had three reviewers. One was quite positive to start with and,

and all slick. Mm-hmm. Two others were quite critical.

So maybe one of yourself there. Mm-hmm. The paper is so much better mm-hmm.

Because of those criticisms. And actually when I look back, it wasn't ready.

Mm-hmm. It wasn't ready. We, we, I,

I think we rushed it out or we didn't know we were rushing it out,

didn't feel like we were out.

The feedback that came back and then the end result that's now out,

there's just so much better. Mm-hmm.

And we were so much more confident that it can get used. Mm-hmm. It,

it's an app for analyzing data that actually, I,

I was hugely thankful of the reviewers. Mm-hmm.

But it's a difficult process to, to see it. And they come back and,

and it's really nice that people do decline. I,

there's been times when during the reviewing process,

I wish people knew that it was me reviewing.

'cause actually you can just enter a dialogue even mm-hmm. And address them and,

and help mm-hmm. Which reviewers, in your case were doing,

you are helping mm-hmm. Forward, uh,

to as best data as we could in the PhD's time. So I would say it's not me. Yeah.

Wiggins who, and, and Judy Wilson put loads into this far more. Mm-hmm. Uh,

but thank you though.

No. Um, I mean, I get asked to, since we make a lot of tools,

I get asked to review a lot of tool papers and, um, I think I,

I give those papers a really high bar because, uh, one thing that, um,

n Carpenter who I work really close with, and I,

who used to be my boss for a long time, um, uh, say is, um,

hard on authors is kind to users. Um,

because we want tools to go out there in a state where someone can use them,

where they're user friendly. Um,

'cause there's a lot of tools that people make something that works really great

for them. Um, and it works really great for them and not a lot of other people.

And people try to use it.

And there's so many hours spent in this ecosystem with people doing, using,

trying to use tools that someone absolutely put their best effort in.

But user friendly software design is really hard.

And it's not something that you're taught as a scientist, certainly of like,

how do you make user friendly software? And so when I'm reviewing tool papers,

what I see my role as is to, you know, make sure it's scientifically sound,

which basically it always is.

But to also make sure that for what the paper said,

these are the people we want to use our tool.

That it's sort of accessible to the people that they say are their stated

audience. Um, and I think if we do that, then again,

sort of all of science gets a little bit better. It's hard 'cause you don't,

you don't wanna make somebody, you know,

make something that has 20 years of polish with a software engineer before they

publish it. But you do wanna make sure that, um,

it's accessible to the people who wanna use it. Ah,

As was definitely not in that case at the start,

because I think Laura would use three different, uh, software codes mm-hmm.

To get there. Because I think during the process of,

and this is something actually to consider for anyone listening mm-hmm. Now,

each step, we, we, we had an end goal for the project,

which was to analyze some data, but to analyze the data. We had segmentation,

segmentation issues. We had the machine learning side, we had the, uh,

uh, algorithm to, to the principle components analysis type outputs of it.

Mm-hmm. And actually,

there's three different programs that work better for each component.

And essentially we chose, uh, not just me, you know,

collectively the easiest package to, to solve each of those three points,

but at the end of it, they were linked together.

But it means you have to do one bit in one package, move to another,

move to another, which is so un user-friendly. Mm-hmm.

So the development to selfie in the end, thanks to the feedback mm-hmm.

Gave enough push to make it better. Mm-hmm.

And it's where needed to be to start with. So I think if we'd have,

in hindsight, we'd have designed it knowing we had to do all these steps,

but as, as a group that's never really done much in this area. Mm-hmm. We,

we were, we,

we just boded our way there and then you had to

harmonize it all at the end. And of course that bit is, you know, top,

but it's hard on the author kind to the user is exactly right.

Yeah. No, and, and, um, you just, you start with making the thing that works,

that's always, you know, whether you're designing for yourself or someone else,

you always start with just sort of getting something that functions.

And then it's just a question of how, how much more work do you put in there?

And again, it comes out to,

if you think you're the only person who's ever gonna use this code,

like use the code that works for you. But, um,

if you think your code is sort of worthy of sharing or you think it will help

other people, then um,

the likelihood that that will happen definitely goes up if you, if you take the,

take some extra time, um,

and get advice and feedback from naive users is often the best way to do it. Um,

and figure out how to get them to use it. Um,

I will say we fixed more bugs than cell profiler in response to having go, uh,

just sort of gone and taught tutorials with our tools.

And then just seeing where people struggle, um, and saying, alright,

I've explained this to people five times in a tutorial.

That means that we did something wrong in the software.

It's something where we haven't set things up correctly to help people succeed.

And so that's honestly where we fix a lot of our bugs.

Well, I'm gonna give a shout out to Grant Calder,

who's our Guinea pig in the end, once we got the app in a more policy mm-hmm.

He still found bugs even with the, uh,

even with the data that uploaded for sampling was anyway. Mm-hmm. That, that's,

but thank you.

I think it's really cool people to be brave enough to declare yourselves.

I think most people, most authors are actually very grateful to reviewers,

you know? Mm-hmm. The, the criticisms seldom wrong. Mm-hmm. Uh,

and it, and if they are wrong, then they've misunderstood something and there's,

and there's a perspective change,

which means a paper clear enough to start with arguably now. And Okay.

There's always going to be the odd reviewer that is, uh, somewhere out there.

But

Yeah.

It's not that often you encounter that reviewer. I don't think, uh,

Yeah, no, I, I, I think if you assume that someone,

someone's heart is in the right place, and again,

like mine is always just around sort of,

let's make these things user friendly and reviews hurt even when they're good,

good reviews. If they're like, this isn't perfect, you're like, no,

this is my baby. I made this. How, how do you mean? It's not perfect? But yeah,

if, if the process goes the way it's supposed to go, which it doesn't always,

um,

then what you end up with at the end is something better than where you started.

Yeah. No. So thank you very much Beth. And we, we ought to, so actually,

if you don't know Beth very well,

I apologize 'cause I haven't really introduced you fully. Uh, you,

you are now really motoring in the, the way of cell profiler. Mm-hmm.

Lots of data analysis, the algorithms around data analysis,

all the processing of data and images. Mm-hmm.

But you weren't a computer scientist where to start with?

No, absolutely not. Um, my, I, I am a molecular biologist by training. Um,

and in graduate school I wanted to do a very fiddly project that revolved

around, um,

co localizing and measuring relative levels of a couple different proteins at

hundreds of spots per cell times hundreds of cells. Um,

and it turns out that's not trivial. Um, and I had, I didn't know how to code,

I'd taken a sort of two day MATLAB bootcamp and I asked around in my lab and I

asked my friends, and everyone's like, yeah, I got nothing for you.

So I was like, front, I'm gonna need to learn to code. Um,

and then I was introduced by, uh, by Nico Sterman,

who was then in Ronell's Lab to sell Profiler, which, uh,

Anne Carpenter had put out a few years before the first version. I was like, oh,

this thing's really amazing. Like this is gonna solve all my problems. And it,

it did solve a lot of my problems. Um, but then at the end of it,

I ended up with still a giant folder of spreadsheets with all of my

measurements. And I was like, oh no, I still don't know what to do. So I,

I took a nine week Python class and I learned to code and I realized that, um,

working with Cell profile or in helping people do analysis,

that was way easier than what I had done,

which was a lot of counting things by hand and circling stuff by hand.

I was like, this is great. I wanna make this easier for everyone. So, um,

when a position opened up on the, the self profiler team with Ann Carpenter, um,

I was like, oh my gosh, this is a job helping people self profiler.

I've been doing this for free. I wanna get paid to do this. Um, and from there,

yeah, I've been at Broad about seven years.

I, I, I've got two questions. Mm-hmm. How much do,

do you ever get in the wet lab anymore?

No, I don't. We don't have a wet lab, so, uh,

I haven't touched I pet in a very long time.

Long time. Do do, do you miss that side of it?

I don't think I miss most of the bench stuff. Um, like cloning tetra culture,

uh, I do miss doing microscopy. So, um,

I was lucky enough to just recently be down visiting the,

the quantitative imaging course at Cold Spring Harbor and sort of looking at all

of the microscopes that they had as the students were going through and doing it

was like, oh man, those are really nice. And, and even just in, in seven years,

like some of the technology has come so far. Um, so if,

if we had a wet lab in our space, the broad,

I would probably be okay doing some microscopy. But the rest of it, um,

I'm happier coding.

That's fair enough. What was the first microscope you ever used?

The first microscope I ever used was in undergrad. Um,

there was a laser scan and confocal that I did for a lot of tissue section

looking for, um, I h c of a couple of molecules. Um, I mean,

if you're not counting the little sort of plastic one that I got when it was

four, um, which there's a whole movie of somewhere that I, I wanna find someday.

Um, but yeah, um,

laser scan and confocal doing I h c 4 88, 5 68.

And I was just like, whoa, my data is beautiful. Um,

and that was really the thing that drew me to microscopy at first,

is I was like, oh my gosh, it's data, but it's also really pretty.

And if I'm gonna be staring at this for however many hours a week,

like at least it can be beautiful also. That's,

That's cool. So you went to and Carpenter's lab and you,

you sent a picture of yourself and Anne mm-hmm. And Anne's a previous guest on,

on Flow Stars itself, because Anne herself is also a star.

But it's amazing that now, and you, you've spun out now mm-hmm. By yourself,

out Ann's lab and now your own star.

How does it feel to now be your own star,

you know, outta Ann's wings and actually being mm-hmm. You?

Um, I mean,

Ann and I still work super closely together and I'm incredibly grateful to her

for giving me a chance in her lab in the first place. 'cause I,

I had a very difficult grad school experience. I didn't publish.

It was just not great. And she took a chance on me, um,

that a lot of other people wouldn't have taken. Um, and of course,

the major tool that we're making Cell Profiler is not a tool that I made.

It is a tool that she made. Um,

but I'm so glad for her that she gets to focus on the thing that she has come to

really love, which is all of this, uh,

morphological profiling sort of downstream bioinformatic stuff.

And I get to focus on the parts that I love,

which are the sort of making tools that make people's science better and make

people's lives easier. Um, and so, um,

it's really fantastic to sort of be doing that.

It's still a little strange to me 'cause I still,

I've been in sort of bio analysis as my main focus seven years now,

and it still kind of feels like I'm new on the scene.

'cause there's some great people, um,

who've been doing this a lot longer than I have. But, um,

I'm really excited to be now more involved in the community and sort of helping

out on projects like Quare and, uh, the reboot of New Bias. And, um,

so all of that is really exciting.

Right. It's great.

We'll come back to new bias because I know it's great that they are mm-hmm.

Rebooted. Mm-hmm. Or, or mm-hmm. I, I,

I don't think they ever really faded away completely.

It's just obviously they're dropped and they, they're, they're now back,

which is awesome. So we'll come back to new bias, uh, in a bit. Mm-hmm. So, I,

I always ask guests who their inspirations would be. Mm-hmm.

I'm gonna ask you now, please just see, I, I'm presuming Anna's one of,

of your inspirations.

Yeah, absolutely. Um, again, she sort of saw a need in the community.

She went around and asked a lot of people and was like, this is a need. Right.

I'm not just crazy. And there were people were like, yeah,

we don't have a thing that does, uh,

high content or sort of reproducible analysis without coding. And she,

she found someone who could help her make it, and she made it. Um,

and she's built, I should say,

like an incredibly positive lab full of kind people and,

um, just sort of really good science. Um, Shanus, who's her co-pi now,

between the two of them, they're the two nicest and smartest people that I know.

Um, and it's rare that you find that combination, um,

let alone when you find that combination in the two PIs of the lab. Um,

definitely also my mom, um, my mom is not a scientist at all.

She was a librarian, but she taught me how to,

that the most important thing to in life is to, or in, in doing your science,

your professional life, is to know how to find more information.

And I think that's one thing that I try and teach folks in my lab or folks with

who are trying to get started in image analysis,

is just sort of knowing where to find the information you need.

Not everything needs to live in your head.

There's so much stuff that I look up and, um, when new,

new computational biologists who come from biology join my lab,

they're always like, oh, but I have to Google everything. So I'm not really a,

a software engineer or coder. I'm like, no, that's what we all do. Um,

the most important thing is just learning how to find the information that you

wanna find. Um, and my mom really taught me that. And she's also, again,

just sort of a great and really smart and really kind person.

And that's who I try to be.

Do do your parents realize how successful you are, um,

and the early stage of your career as well?

Yeah. My, my mom's really excited. Um,

unfortunately my dad passed about 10 years ago, but, um, yeah,

everyone el no one else in my family's a scientist,

so they're just sort of like, okay,

that's off doing her own sort of strange thing. It seems like she's really busy,

so I guess it's going well. Um, but yeah,

I think my mom would be really proud of me no matter what I was doing.

That's super cool. So I'm, I'm gonna take you back then,

when you were 10 years of age, around that age, 10, 12, around that,

what did you see yourself doing? Um,

what was your dream career when you were a young child?

I wanted to be either a singer or an actress. Um, I wanted to be famous.

And then it turns out you have to be good at those things,

which I'm not particularly, so when I got to about high school,

the sort of realization of ohoh,

like I'm not actually good at singing or acting. I love those things,

but I'm not good at them. Um, I did, you know,

choir and acting lessons and all of that stuff,

and I'm just not very good at it. So I was like, okay, what,

what do I also love but am good at?

And I took an AP bio class in high school and I was like, oh, wow,

this is really cool. This is all really fascinating.

So I decided to go into biology and, um, from there, again,

never saw myself as a computational person,

never took a sort of college computer science class or anything.

But as I got into, you know, using code and using computer science, I was like,

oh, I love this and I'm really good at it. And so I'm just gonna,

I think keep chasing the, you know, Ooh, that's cool. Um,

as long as that keeps working as a career choice,

I, I think that's really sound advice.

Attitudes follow what you enjoy and follow what you're good at. Mm-hmm.

Because obviously you enjoyed acting and singing mm-hmm. But maybe,

maybe not good at it, obviously I haven't mm-hmm. Heard you sing. Uh,

are you going to elmi or anything this year?

Maybe we can get you singing at that point. Uh, yeah.

I, I will be at Elmi. Uh, the last elmi I was at in Dublin in 2018,

I believe there was karaoke after, so

I did some karaoke s l i s this year, which was my first time since Covid.

So that was really fun.

That's cool. It'd be good to have you. I, because I, so I, we, I,

we met before Dublin, I think, but I, I certainly remember sitting,

it was at the, uh, it was a, it was a barbecue, wasn't it? Mm-hmm. Uh,

outside on the, on the wooden benches, I remember mm-hmm.

Talking to you and stuff. I,

and actually going back to my PhD students who were then teaching on a course at

Heidelberg beyond that,

Laura went on and you gave a load of really helpful advice,

I think always been in touch with you once or twice since by email asking

questions. Mm-hmm. Uh, yeah. I, I just so much influence,

you know, I, I I hope you appreciate, you know,

that you are appreciated. Uh, that's

Really nice to hear. Um, yeah, I mean, again, I love helping people.

I love that moment when, again,

sort of you help someone and they realize that there's this whole thing that

they could do before that they, they couldn't do. Now, I've, I've loved,

I did teaching in sort of high school and colleges, sort of like side jobs for,

you know, for coffee money and stuff like that. But, um,

with bio image analysis,

we have this sort of huge gap between people who really wanna do, um,

these analyses and really wanna find things in their data.

And the tools just are the best that we can make them with the time and money

that we have and the technology that exists. But there, there is a,

a gap for a lot of people in terms of like, here, here's what I need,

and then here's what exists.

And helping people over that gap and helping them get to the other side is just

incredibly rewarding.

So thinking of that side, how, how are, how,

how is the work funded and how easy is it to get funding?

Because I can't imagine it's the easiest area.

It's not solving a straight biological question, it's not mm-hmm. You know,

developing a whole new technology.

It's kind of sits almost in limbo land when it comes to this. Uh, where,

how do you find getting funding and different initiatives?

Yeah. So when I started, um, in the lab,

cell profiler did not actually have any dedicated funding at all.

It was basically funded by running a contract image analysis fee for service

thing,

and then hoping that that made enough money on the side that you could hire a

software engineer to do cell profiler maintenance. Um, in, oh gosh,

it was 2018 now. Um, when Chan Zuckerberg first got going and started bringing,

uh, giving out grants,

one of the first rounds of grants they gave out were these imaging software

fellows. Because it's,

there's a lot of grants out there for making a new piece of software.

There are very few things out there for taking a piece of software that

thousands of people depend on and use and keeping it working. Um,

'cause it's not, it's not flashy.

And grant organizations wanna have flashy things for the most part. Um,

and I give Chance Zuckerberg a lot of credit that one of the very first things

they did is fund image imagery, psych and image and cell profile,

or a software engineer each to maintain stuff and keep working. Um,

a lot of the rest of our labs funding comes from, uh,

this thing we call Center for Open Bio Image Analysis, which is a joint, um,

grant between myself and Anne and Kevin, er at University of Wisconsin, um,

which is an N I H grant where they have,

they have this mechanism that's about sort of taking technology and bringing it

to other people and making it easier to use. And so we do, um, tech development,

we make new tools. Um, we have a new tool called Pixie that, um,

I'm gonna be showing at elmi. And, um, that has been out for a while,

but we keep adding more functionality and is a really easy way to use deep

learning without needing to learn to code. Um, which we're very excited about.

Um, but we do it in the con, not in sort of vacuum,

but we do it with biological partners that we identify who have really difficult

to solve bio image analysis problems.

And the other thing with that grant funds is community engagement work.

So it funds us to be able to go out and do tutorials or write protocol papers

and things like that,

because there's a lot of problems that can be solved now with the right

knowledge. Um,

but we need to get the knowledge out there and we need to get it to the people

who need to read it. There's a lot of stuff that's trapped in computer,

computer vision journals, but that's not what biologists read.

So if you, if Anne had commercialized self profiler,

would that have enabled it to be develop developed faster or do you think it

would've hindered its uptake?

Um, I mean, there's a lot of models for commercialization. We do still sell a,

a sort of service plan for cell profiler that if you want sort of x number of

hours of help. Um, and we still do image analysis fee for service,

and that does help a little bit. Um, I,

I think it would have helped cell profiler get sort of slicker faster. But,

um, if you look at the,

the license of a lot of these commercial pieces of software,

it's beyond what an individual lab can afford. And certainly, you know,

even a beyond an individual lab,

a grad student who's trying to get started and wants to sort of push their lab

from, you know, representative image shown to, all right, let's,

let's start analyzing this quantitatively.

They're not gonna convince their PI to shell out thousands of dollars just

upfront. Um,

and so I think having it be free helps the adoption doesn't necessarily

help with the sort of, um, being able to pay to keep it used,

but we're lucky that we've found some groups over the years that think it's

valuable and we're gonna just keep trying to figure out how to do that.

So, so have, have you ever gone to the, like the, the, the big pharmaceuticals,

the likes of Pfizer, G s K, uh, and so forth and asked them mm-hmm.

For sponsorship, for, for supporting it because they use it. Mm-hmm.

They use it in a big way. Mm-hmm. So do, do they, do they help?

So some of them buy the buy the service plans? Um, but um,

I can, I can say we don't sell tons of those. Um, yeah, it's, I mean,

it's hard when someone's like, I could have this thing for free,

or I could give you $10,000. So it's like, well, but I could have it for free.

It's hard, it's hard for,

and the people who who work at the pharma companies have to then justify to

their bosses like, okay, I, here's why. I think putting, you know,

X thousand dollars into Sell Profiler is worth our time. Um,

it's one of those things that if everybody did it, um, you know,

it would be great for things,

but each individual company doing it isn't gonna see a lot of one-off value. So,

um, it's, it's hard in that sense.

Yeah. No, I, I can see that. I, I do wonder if,

if there was a specific new development that wanted it,

and it's going to certainly help pharma. Mm-hmm.

I can easily imagine you could approach three four big pharma and just say,

look, this is what we want to develop. It's in your interest,

you get free access. Mm-hmm.

Give us 50,000 each to sponsor that program

and mm-hmm. Yeah, I bet it could be done. But anyway,

Yeah. I'll say not a pharma,

but the Allen Institute really sponsored cell profiles move to three D 'cause

they wanted it for their stuff, and they were like, okay,

we see that there's this tool, we wanna not just build stuff internally,

but help it bring it to everyone. Um,

but that's a pretty one-off case for as long as I've been on, on the project,

uh, formally, which is seven years.

Right. And it, it, you know, asking the companies is hard work. Mm-hmm. Again,

then you have to have almost a project manager to go out selling mm-hmm.

And everything else, it, it's, ah, it's a complicated world, isn't it? Mm-hmm.

You touched on the machine learning side mm-hmm. The ai,

and obviously you've got a lot of the chat programs now coming through that can

mm-hmm. Well design for coding to start with. Mm-hmm. Do you,

are you starting to use that to start to write some of the scripts and then edit

them and correct them?

Or are you ignoring them or why they're threats even?

They're definitely not a threat. I mean, I think if,

if everyone knew how to code, that would be fantastic. Um, you know,

certainly we would pivot to doing different stuff.

'cause what we do is sort of bring code to people who don't know how to code.

Um,

but I think it would be amazing if more people wanted to and more people did it.

Um, I don't personally,

I don't personally get to do that much coding these days. Um,

it's a lot of paper writing and things like that. Um,

so I haven't used it much and I'm not that inclined to use it myself

personally because, um, I know what kind of mistakes I'm prone to make,

and therefore where to find 'em and fix them when stuff goes wrong.

I don't know as well yet when, um, you know,

the problem with all of these large language models is they write everything

very confidently. Um, and sometimes they're confidently wrong. Um,

and I know how to trace down my own mistakes.

I've been doing that for a long time.

I don't know as well yet how the kinds of mistakes it makes,

but people are allowed,

have been using it and have said that it's one of those things that 75% of the

time it works really well.

And then 25% of the time you then spend two hours trying to chase down, like,

where is the weird bug it introduced? So, um,

probably as these things get better, um, in the long run, it will, um, you know,

it will be faster than writing your own code. But I'm,

I'm not fully on fully onboard myself yet. Okay. Um,

I'll say,

Uh, reer put out yesterday this thing called Omega that's supposedly a bio image

analysis scripting thing that self-corrects errors.

And so if stuff like that becomes more common, that's gonna be really cool.

And if itno, it's made an error, I do, you know,

and I gonna say to your lab members, say if they're getting it 75% net,

at least they're asking it to do the right thing, which is also still in itself.

Mm-hmm. So, so it's that, that in itself is not trivial to get that close.

You mentioned your lab, you sent me some picture. Mm-hmm.

So a picture of your lab on a nice house. Mm-hmm. So how many,

how many are in your team currently?

Uh, at today we have six. We fluctuate between sort of six and 10.

We have a new post-op who's starting in about a week and a half. Um, and, uh,

one will be leaving in June and one starting in July. Um, yeah, uh, we,

we have a couple software engineers.

We have one staff scientist who does a lot of our high content screening stuff.

And the other folks are in my group, we have, we have a visiting, uh,

grad student for the year who's been fantastic.

And we're gonna miss her terribly when she leaves in June. Um,

but the postdocs in my lab are part of this postdoctoral training program that

we started in 2019,

which we call the postdoctoral training program in bio image analysis.

As far as we know, it's the only one. But, um,

it's modeled very heavily on Jennifer Waters's, um, microscopy Fellows program,

um,

where we take folks who have the same journey essentially that Anne and I have,

where they're molecular biologists, cell biologists, you know,

wet lab biologists who know a bit about, uh,

image analysis and a lot about imaging and wanna learn more and help,

we help them sort of become full-time computational biologists by coming to the

lab for two or three years and working on projects. And, um, I should say, yeah.

So we've, we've had, um, eight people start.

We have three more people sort of in the pipeline,

and we've had six people leave,

and it's been really successful and also really rewarding to see people come in

and develop these skills and then go off and get really cool jobs where they're

paid a lot of money.

Oh. So I, I've just noticed with this picture,

it makes it look like I've got a candle sticking out the top of my head.

So Yeah. I've one, and trust me, I'm, I, I'm a lot older than of age.

That's that. Uh, so the,

the funding who funds those pe those fellowships,

Um, so some of that's funded by the, by the center. Um,

some of that's funded by the, the contract image analysis work. Um,

and it's great because, you know, we, we do have a staff scientist who does,

you know, image analysis fee-for-service projects for us. But, you know,

she's seen everything at this point many times. But, um,

the variety of projects we get in for that are great because it allows the

post-ops for them.

All of these projects are theirs for the first time and they get to learn things

as they're trying to solve them. Um,

and so I don't know that we'd be able to do what we're doing as well as we could

if we didn't also have this commercial side where we sort of take in projects

and just sort of do small 10 to 20 hour, um,

image analysis tasks that then we give back to the collaborators. So, um,

it's a weird sort of hybrid funding model where some of it's, you know,

some of it's grant funded and some of it's just, um, core facility funded,

Which comes back again to, I, I think the whole community. Mm-hmm.

And I don't actually probably mattered very little what area of science you're

in. Mm-hmm. Uh, the importance of software for analysis of data is huge. Mm-hmm.

Mm-hmm. And yet still in the funding bodies mm-hmm. In general,

quite hard to get funding to do token bits. Yes.

Mm-hmm. Projects a lot harder still, it's still mm-hmm.

I'm not sure if that's the funders mm-hmm. Or whether it's the,

the panel that meets to discuss it that aren't so mm-hmm.

Into analysis and can't quite see the wider benefit straight away

mm-hmm. Until it's developed. I mean, image J was mm-hmm. You know,

really struggled. And now not only scientists use it, but home photographers.

Mm-hmm. You know, and, and it's had a huge impact. Mm-hmm. Yeah.

I don't think that was funded properly in the first instance.

Yeah. I,

I don't know too much about the original when Wayne wrote N i h and stuff, I, I,

I have the, uh, the impression that you're right,

that there have definitely been struggles, I mean,

for over the sort of 40 something years that it's been around. Um,

but yeah, it's, I,

one of the first slides I always put up when I give a talk is a picture,

a drawing by Ramon Kahal. Um, first of all, 'cause it's beautiful,

but second of all, because, um, I think in microscopy particularly,

we are hindered by the fact that most of our history is as a qualitative

technique. Um, and so I think, you know,

stuff like genome sequencing when it came out, people were immediately like,

okay, well we definitely need computers for this,

because it immediately came with sort of computers right alongside.

Whereas we used microscopes for hundreds of years without any computers and we

got along. Okay. Um, so I think, uh,

I think we're hampered a little bit by that history. Um,

and it's something that is, is slowly changing, you know, the,

the sort of beach heads are softening. But, um, you know, we've got a lot,

we've got 400 years of history to sort of make up for.

Yeah. Well, obviously we've done our app side of it,

so we get it. Mm-hmm. Uh, but it's me, if you go back, you know,

you say it's been around,

around for centuries and you go back to Robert Hook's pictures and they're

beautifully drawn pictures of a single inset,

lot of three or five of them just look at mm-hmm. What's different mm-hmm.

You know, and biology,

we looked down microscopes and you looked at a cell and many people still look

at a cell mm-hmm. And not the hundreds of cells mm-hmm.

To embrace the heterogeneity and mm-hmm.

Identify the subtle differences and why those subtle differences exist and

mm-hmm. What our profile is is out there to do and to,

and to enable us to make sense of it. Mm-hmm.

Uh,

Yeah. We've talked about funding, we've talked about getting funding setting.

How was it actually when you broke away from that? I,

I know you still work really closely with, it wasn't really, it wasn't almost,

I, I could be wrong. Mm-hmm. Probably wasn't the same way.

You had to start in a new institute to set up a group,

but you were still mm-hmm. Now you've got your own group and have to set it up.

Mm-hmm. How daunting was that?

Um, when Anne first approached me about, I think we should do this,

and she's written a whole article in eLife about like,

why her decisions around doing it and things like that. I was just sort of like,

I don't really know, I was not immediately on board with this idea. Um,

and it, it was more of a transition than I thought it would be. So I mean,

essentially, you know, pre-split, the lab had sort of three teams.

It had the bioinformatic team we call the image-based profiling team, um,

the image analysis team and the engineering team.

And I was already running the image analysis team.

Yeah. I should just say, just for clarity,

this wasn't a split between you and Anne. This was Anne. No. Yeah.

Divide up her lab to make it more manageable or, and to be more mm-hmm.

Mm-hmm. And she was like, so I want you to keep running the image analysis team,

but as your own lab and then also we'll fold the engineers in with you.

And so it was really in,

at the time only a head count increase of like two people. Um,

and I was already supervising like four people, so I was like,

how how much harder can it be? It's a lot harder. But,

um, yeah, just I think even if the day-to-day isn't harder, just the, um,

the knowing that these people are gonna rely on you that,

that they salaries are gonna get paid,

that you're gonna steer them in the right direction,

that you're gonna help them get the tools to where they need to be,

and also them as people where they need to be. Um, the, you know,

I take that responsibility very seriously. Um, 'cause I've,

I've had great mentors and I've had mentors, um,

and I want to be more like the great mentors that I've had,

and I wanna help people sort of reach their potential.

So throughout all that, and it's been a whirlwind 37 years you've been there.

Now you've got your own, this is whirlwind fast. Mm-hmm. How do you relax,

Figure it out? Um, yeah. Um, I, I have,

I have a wonderful husband who I love spending time with. Um, uh, we,

we go to some Red Sox games each year. Um, I do some gardening at home. Um,

Now you sent me a picture of your, is this mm-hmm. This isn't your garden.

This is just a garden box. Surely.

Yeah. So yeah, this is a garden box. This was the,

the first year that I was like, okay,

I'm gonna get a thing and actually gonna grow some vegetables. So there's some,

some on spring onions there and some carrots.

And the carrots didn't come out that well. Some potatoes in the back. And, um,

yeah, I mean, it's certainly not enough food to live on,

but just sort of in terms of like, oh, hey, I made this. I love to cook.

So being able to sort of take it one step further and be like,

I not only like put the food together, but, um, I made the food

This picture of a, a las mm-hmm. Yeah.

I swear that cheese needs more cooking on the top.

Yeah, it did need a bit. But, um, it was, uh, it was something I was,

my,

my sister had covid and I was giving it to my mom to bring to my sister and she

needed to get going. But, uh, I was part of the,

that picture also was funny 'cause was, um,

the most mad I've ever made someone on Twitter, um,

was I had made a giant thing of pasta sauce to make the lasagna.

And I had a very full pot of pasta sauce. And someone was like,

you don't deserve to run a lab if you keep food.

Containers that full on the stove. I'm

Like,

What is, like, what is wrong with you? Like, why is that?

I was like, oh,

well you're a computationalist so that's why I guess we won't take your lab

away, or something like that. And I was like,

a very strange reaction to someone posting a picture of food.

Was she, was that person serious?

I dunno.

I think so. I always try see the, I've take as tongue in cheek humor, but,

but no, I, I know some people are actually mm-hmm.

But I'll say, I can't believe I've criticized your cooking,

but if it wasn't cooked, you're gonna bet your sister even sicker than covid me.

Come on.

No, the las was cooked. It was just the,

the cheese needed a little more browning.

I'm teasing.

I know.

So, uh, what is your signature dish if you cook at home?

Um, the first thing I ever learned to do was sort of a pasta in, uh,

pasta with seafood. Usually like shrimp or scallops with, um,

sort of butter and garlic and white wine. 'cause it's really simple,

but it basically, you can't screw it up just about. It's, it's really nice. Um,

I do love cooking Italian, unfortunately,

my husband doesn't really like Italian. We end up cooking a lot more, um,

Asian and Mexican and sort of global flavors. Um,

just 'cause in Massachusetts we don't get as many of those.

There's a lot of pizza and, you know,

not a whole lot of other global flavors in Massachusetts.

I, I, I'm, I'm still taken aback that someone doesn't like Italian cooking.

That's amazing. He

Apparently grew up with a lot of bad Italian cooking, so, uh,

there's apparently some, some stuff related to that.

I, he, he, he may still get there. Yeah. Eat pizzas.

Surely he eat pizzas.

Oh yeah. Pizza. Yes. But sort of tomato sauce is, is, you know,

kind of hit or miss with him.

Wow. Actually, no, my, my son hated our pasta sauce for a long, long time.

Mm-hmm. Now he's growing into it, growing back into it.

Now he's in his twenties now,

either that is a lot more polite and eats it for us.

I'm not sure which. So also outside of work,

you sent me some other pictures. Actually,

you sent me a picture picture of Austin Red Sox.

Yeah. So this was a blanket that I actually made. Um, that's one of the other,

I, I don't sit still well at all. And so, um,

I picked up knitting in graduate school from a dear friend 'cause I was like,

oh, then at least when I keep my, my, my hands are fidgeting.

It does something useful. So yeah, that's like a full size, uh,

blanket with a, with a Red Sox logo on it. I made my husband a,

he's from San Francisco originally.

A San Francisco Giants one that matches that.

How big roughly are these?

Um, six feet by like three feet.

I knew you were gonna give me that in imperial measurements and not metric.

Alright.

About two meters by one meter. We'll go metric. Yeah. You're a scientist.

It's meant to be metric now.

Yeah, well, I mean, I, I can do the, like when folks, when, you know,

non-US folks in lab are talking about weather and stuff,

I'm usually the one who's like, when everyone else is like 70 degrees,

I'm like 20 C, um, I can do the conversion.

It just doesn't immediately pop to mind,

Oh God, could you please just,

just do a special American version of cell profile where we get a scale bar to a

fraction of an inch that drive everyone nuts.

That'd be quite entertaining though. Mm-hmm. So, coming back onto work, uh,

and beyond Cell Profiler mm-hmm. We, we mentioned new bias earlier. Mm-hmm.

New bias. Tell me a bit about it for those who, dunno what new bias is.

Yeah, so new BIAS stands for Network of European Union Bio Image Analysts.

Um, and original what

In the us So, so this is the network of European.

Mm-hmm. And your

Mm-hmm. So it was funded, the original version of it,

I started 20 15, 20 16 give or take. Um,

and was funded by a European Union cost grant mechanism. Um,

so it was originally designed for the eu,

but they could have a couple of members in other countries as well.

So it was like for, they could have other countries join,

they could have two members each.

And so the Carpenter lab was one of the two US official affiliates of New Bias.

Um, and the goal was just, you know, we have these bio image analysts,

these people who specialize in doing analysis.

They tend to be one person in like a biology lab or one person sitting

in a core alongside also teaching about microscopes and where the information

exists. But it's really scattered,

can we actually bring these people together to learn from each other,

to teach other people to sort of become a group as opposed to a lot of

individuals. And it did a fantastic job at that. It really sort of brought,

I think, analysis as a field together. Um,

and then the grant ended in 2020 at the same time Covid hit. And, you know,

folks still did some great stuff. Like, um, they set up an F 1000 gateway. They,

um, they did new Bias Academy and video tutorials when everyone was stuck inside

with their data. Um, but without funding,

it's hard to have a scientific society sort of like come together. And so, um,

starting this year, C z I has given us a grant to sort of reboot it.

We're probably also rebranding from new Bias to, so Bias, um,

society of Bio Image Analyst,

because our goal is not for it to be just in the EU anymore because of the

original grant mechanism. It had to start there.

But our goal is to make it global because we absolutely know that, um,

people are trying to solve these image analysis problems everywhere,

where people are using microscopes, which is everywhere. Um,

so our goal is to make bio image analyst bring bio image analyst together for

their career stuff to help make more educational material and sort of training

stuff so we're not all writing the same five tutorials over and over again,

but actually sort of sharing stuff. Um,

and then just sort of being recognized as a community and as a job path. Um,

you know, bio image analyst is still a very weird job for,

for most even scientists, but, um, we're getting there.

Yeah. And difficult career paths. Mm-hmm. I, you know,

yourself and are some of the few exceptions that who make a,

make a proper academic career outta it. Uh, and I, I've been at it a

Couple years. I dunno if I get, if I make the list yet, but, uh, we'll see if I

Succeed. You have your own group. Mm-hmm. And it's more around data analysis.

And I think, you know, there's a lot of people, as you say,

individuals in core facilities or in individual labs. Mm-hmm.

But they're not leading the labs. Mm-hmm. Uh,

and it's because of that funding model, it's so hard to develop at that,

that ethos and new bias and what was the new term for it?

So bias probably. So we, we still decided on the name

Bias, isn't it? So Bias, but you need to think that name through carefully.

New bias. Yes. So biased. I was just so biased. You know,

he's gonna come out wrong. Mm-hmm. But at least that creates a network.

Mm-hmm. That enables that.

But it'd be great to see those individual spots become

mm-hmm. Groups and, and have more of them and networks.

It's one of the most network communities. Mm-hmm.

Probably scientifically it's one of the most network groupings because it's so

international. Mm-hmm. It always works together. You mentioned Chan Zuckerberg.

Mm-hmm. Uh, Chan Zuckerberg Initiative, uh, a couple of times how these,

that I projects are you involved with?

Um, four at the moment? Um,

so they fund a self profile software engineer. Um,

I am one of their imaging scientists. Um, so, um, I've,

I've been funded by them for a couple years now and fully a couple more. Um, uh,

I'm on this, this grant for So Bias and also, um,

a grant that also started earlier this year. Uh,

head headed by Anna Jost over at Jennifer Waters' Group at h m s to make, uh,

online, uh, video courses for microscopy and image analysis.

'cause as I've alluded to a bunch of times, those of are in this call.

I love teaching. I love education.

I love helping people sort of find the knowledge that's already out there,

but get it to them specifically.

And so I'm super excited to be involved in that group too.

Uh, that was three, wasn't there a fourth?

Nope. Software, five software engineer me, my personal salary. And then, yeah,

those two new, uh, advancing collaborative imaging projects, uh, grants.

So, so there's so all based around the image analysis mm-hmm.

Like microscopy images. Mm-hmm. What about the electron microscopy side?

Mm-hmm. Yeah. We haven't done much with that just because in general,

other people are doing it really well. Um,

and our goal is not to sort of butt in where other people are doing things

really well.

Our goal is to sort of add value in places where we feel like we add value.

We're super happy to collaborate with those folks and to sort of learn from

them. And I think as deep learning becomes more popular,

a lot of the tools will become less specific to what the image exactly looks

like. Um, but it's, it's simply just something other people do better.

And I, so I'll, I'll give another shout out to Jan. I can break again. Mm-hmm.

Actually, do you know what I, I need, I need I need to podcast. Yeah. Mark,

that would be cool. Find out their, because the meetings,

we've just come off the back of a meeting, uh, c i

and they have brought together the, like mic cross image analyst mm-hmm.

The electron microscopy communities and image analysts. So they are talking,

they're there talking at the same meeting, they're talking together,

they're sharing ideas.

You've got those breakout rooms where different ideas can be hit back and forth.

Their problems can be raised and possible solutions from different sides of the

community.

I think it's before that those communities would've been

far more separated. Mm-hmm. And they're creating a genuine interface.

Uh, I think, you know, one thing, funding individuals mm-hmm.

I think something that is a bit different is they're bringing those individuals

together and those groups together so they can share across it.

And I think actually that's, sometimes they're very different.

You think these people are never gonna talk. Mm-hmm.

But because it's such a variety, it's sliding, sliding boxes.

Everyone's super cool.

Yeah. It's, it's one of the friendliest and again,

sort of the idea of somebody else is doing this better.

Let me just point them at that direction. I mean,

there's enough work to go around. So,

but rather than people trying to compete and one person trying to say,

I am the king or the queen of image analysis and all of you are below me.

Like everyone just is sort of like, yeah, let's work together. Um,

which I love about this community. I'm, yeah,

I'm super excited that em folks are getting bring, brought more in.

I'm excited about this, um,

communist initiative about bringing in more of the medical imaging stuff too,

because it's crazy how much sort of em,

light microscopy and medical imaging all are solving very similar problems and

don't talk to each other. Um, it's kind of wild.

And so I'm super excited to see people sort of trying to bring these groups

closer together and sort of stop trying to resolve the same problems over and

over again in different fields.

Yeah. And that's almost where you need international funding. And,

and they partly do fund international. We need even bigger.

Maybe that's where they dovetail into Bill and Melinda Gates. Mm-hmm. Mm-hmm.

Uh, you said you love training. Mm-hmm.

This is one of the training courses that you are Yeah,

So this is the training program that I mentioned where, um, yeah, I've got my,

my awesome folks. We've got, uh, like I said, three more starting and, um,

I particularly give shout outs to, um,

NAEM Jamal and David Sterling who were the first two who sort of took a chance

on this entirely unproven model of can we, you know,

we knew that we had had some people come through the lab and sort of, you know,

start off as co as biologists, sort of, uh, computational biologist, big B,

and then go back to sort of like balanced c and b. Um, but they were,

we started this new training program and they were just like, cool. Sign me up.

Um,

and they both did amazingly well and learned a ton and are now sort of impacting

things, um, elsewhere. And, um,

it's been absolutely my favorite part of the last five years has been having

those people come through my lab and sort of start their scientific careers and

push 'em in new directions and then, you know, get offered really great jobs at,

they're really excited about at the end of it.

And you mentioned how some of 'em go after. Mm-hmm.

I presume to industry as well in some cases,

'cause you mentioned they're getting paid very nicely or sort of handsome

mm-hmm. Going through to industry.

Do you feel that the amount of industry pay,

especially computer scientists that can go into, uh,

and not just into science, but can go into, uh,

banking industries mm-hmm. Market industries and so forth, and they,

they pay mega, but mm-hmm.

Do you see that as a prohibitive of attracting the best talent into the

scientific image analysis community?

Um, it certainly makes it harder, you know, when you know that a,

that a good software engineer could sort of walk across the street and in our

case it's literally across the street to Google and, um,

and go make twice as much money. Um, um,

I think for a lot of the problems we're trying to solve, um,

having a lot of domain expertise helps for many of the things, which is why,

again, we pull from biologists, um,

and then sort of give them the computational skills because there's a lot of

stuff that is of different between how microscopy works in theory and then how

microscopy actually works when your secondary antibody, you know, the,

the Fluor associated from the antibody and now you've got speckles everywhere or

you know, all of the gajillion things that your fixation, you know,

didn't quite work. There's a gajillion things that can go wrong, um,

in your sample on the way to the, to the detector. And so, um,

having that domain expertise helps. But, um,

we've been fortunate that we've had a couple of fantastic,

several fantastic software engineers in the group during my time. Um, but yeah,

absolutely,

it's a little bit hard sometimes to understand why they wanna work with us when

they go across the street and make a lot more money.

But I think some of it's for the love of making the world better and making,

making something that they know is gonna get used and is gonna improve sort of

all of science.

I, I remember talking to, I think it was Pearl Rider,

Pearl Rider and Mark Ray and I can't remember which went on,

wanted to do MD 'cause you know,

they wanted to help people and MD wasn't for them as it turned out. Mm-hmm.

And they turned two mm-hmm. Back into to research.

'cause there they can have a more profound effect. Mm-hmm. You, you,

you can't diagnose properly mm-hmm.

Without scientific tools and analyze actually what you're doing for the drug

developments. But also I think related on personalized medicine.

Sound profile is going to be behind a lot of the mm-hmm. Emergency, uh,

emergence through that side. Mm-hmm. I'm gonna ask you another question now.

I asked you what you want to be when you attend.

You want us to be an actress in, uh, you are now a very committed, uh,

ed scientist, I dunno what the right term is. Data analyst, BioAnalyt.

Uh, if you could do any other job though, for a day,

a week, a month, what type of job would you like to sample?

What sort of environments, what sort of job would you like to get a taste for?

Not to, not full time, but just go and try.

That's a really good question. Um, again, I I I,

I love sort of feeling like I have an impact. So, um, you know,

government would be something where, you know,

if I felt like I could do a lot of good there,

unfortunately from the outside it looks like a lot of government. I would,

I would probably not like the bureaucracy and the playing by the roles. And I,

I don't like playing by roles if I don't understand why they exist.

So I don't think I do well there.

But it would be really interesting to get a better sense of what could we do to

make a lot of things around us better and why aren't we doing it already?

You'd like to go to the political infrastructure and mm-hmm.

Try and try and enable something. That's a cool answer.

I've got some quick fire questions for you. Yeah. You must, this is coming.

Mm-hmm. What's your favorite color?

Blue.

Yeah. So Dappy? Yeah. Okay. I mean,

D Hooks, you know, either one. Yeah.

Either all work. Uh, are you an early bird or night owl?

Uh, left to my own devices. I am a night owl and so whenever, you know,

vacations and stuff I go that way. But early bird is, you know,

I have a couple hours where I can get work done before meetings start.

So I've become an early bird by necessity.

Are you an early bird and night, Al? Would you burn it both sides?

I try not to. I do my very best not to,

Uh, PC or Mac.

I prefer pc,

but I've been using Mac for a few years just 'cause they're better for a lot of

programming things. But all else being equal, I prefer pc.

Okay. McDonald's or Burger King?

Uh, McDonald's. Better fries. Yeah,

I agree. Much better. Fries. Mm-hmm. What's your go to at McDonald's then?

Um, oh gosh, I haven't been to a McDonald's. Chicken sandwiches are good.

Yeah. Okay. Good choice. Coffee or tea?

Coffee. Lots of coffee.

Long or short?

Long?

Ah, no short.

They each their own,

Uh, beer or wine?

Um, beer usually. I, I used to brew.

I haven't done it in a few years just 'cause the, the time it takes to do it.

But, um, you know what, whatever. Whatever's around.

Okay. That's fair enough. Chocolate or cheese?

Hmm? Chocolate. But there's no wrong answer there

If it's dark chocolate milk or dark.

Dark.

So who cooks at home? I

Cook and my husband cleans, cooks

Mostly.

Yeah. I cook and he cleans and that works great for both of us.

That sounds good. Uh, and what is your favorite food?

Um, um, I'm a,

I'm a sucker for desserts for, for chocolate peanut butter. Um,

the other thing is since, uh,

since this is a sort of light meeting week and I'm writing a lot this week,

popcorn is my, is my writing snack, you know,

sort of can have like write a couple sentences, get a piece of popcorn and,

you know, keeps me going.

Uh, a a true micro snack is a light meal, isn't it?

Mm-hmm. Such a bad joke. Sorry. I was, we was on a meeting yesterday,

so someone background had photon food.

I had a box behind there show Sean that said photon food. I couldn't,

I was just itching to know if it said a light meal. I just, just, I hope

So. I really hope so. Yeah,

I don't think he did, but they will do now. TV or book

On vacation. I love books, but I get really mad when I have to stop reading. Um,

so, uh, during the week when I don't have as much time, it's usually tv.

And do you get into serious in-depth programs or do you watch some trashy tv

Y More trashy TV than it used to be just to, to sort of wind down and relax.

But, uh, you know, again, when when I have free time,

I like to watch the serious stuff too. Okay.

So if you watch your watch your trashy TV confession,

what do you watch that you you are gonna confess to?

Um, I'm still, I don't know how,

I think they're like 20 seasons in now watching Grey's Anatomy and, uh,

I actually got my husband hooked on it with me. And so, uh,

that's one that we watch every week. Um, there again, there are,

there's like one character left from the beginning is how many, you know,

like assassination plots and, you know,

very questionable sexual harassment policy decisions. But, uh,

it's good fun. TVs

So bad. And if you, when you're reading books, what genre book do you like?

Um, I bounced back and forth between sort of nonfiction and sci-fi.

Um, you know, I love, uh, I recently,

now I'm not gonna be able to remember the guy's name on the podcast. Um,

it's a very cool guy, a who spoke at ss l a s at a,

gave a keynote and I don't remember his name,

but he does books about science and something.

And so he did a book about science and crime that I've been reading and is

really good. And now I'm really bummed. I can't remember the name. Um,

I love the Expanse series. Um, if the next Game of Thrones book ever comes out,

I will definitely read it. Um, and

So trashy tv, could you just say Game of Thrones?

Now I'm gonna get loads of comments. Having said that, I think

Thrones were trashy. I mean the last season, the last season, but

Wow. Uh, so you mentioned Sci-fi, star Trek or Star Wars. Serious question.

Star Trek. Even though we are recording this on May 4th, so it is Star Wars Day,

but, um, star Trek.

Yeah. Yeah. I hadn't teach you advice I hadn't even realized.

And what's your favorite film?

Uh,

Moulin Rouge is one that I loved in high school and is one I still find myself

coming back to. Um, it's just good, silly fun.

I love music and musicals, so it's, it's got that going for it too.

Yeah, my my wife loves that film too, and it is very good, isn't it? Mm-hmm.

In broadband. Mm-hmm. Awesome character in it. There's a really good, uh,

fat Boy slim version mm-hmm. Of the, with, uh, Jim Broadband. Uh mm-hmm.

Yes.

The, the, the new musical version of it,

like the Broadway musical is actually pretty good, so if you get a chance,

check it out.

Yeah. Yeah. London sells out really fast. I have tried. Uh,

what's what's your favorite Christmas film?

Oh, I gotta Google classical with, it's a Wonderful Life.

Okay. That's cool. And what sort of music do you listen to?

Um, I do a lot of sort of late nineties early Auts alternative.

Um, I have terrible taste in music honestly. So, um,

I also do a lot of like musical soundtracks. 'cause again,

I still somewhere deep inside and, uh, you know,

high school nerd who loved music and movies and just wanted to be part of it.

I think we've covered a lot of area, but this is where you are today.

Mm-hmm. Where do you see yourself in five and 10 years time?

Honestly, I kind of hope doing the same things.

Like I I've had a very weird strange path to get to where I am, but,

um, I love what I'm doing. I love that we're making tools that help people.

I think, think the tools might look very different in five to 10 years,

I imagine by then.

Yeah. Go. How do you think they'll look in five to 10 years?

How would you like 'em to look? Maybe that's a better answer.

How would you like 'em to look in five to 10 years time?

Um, I think there are some parts of image analysis,

so segmentation is coming close to being, um,

a almost solved problem with neural networks. Um, it's not fully there yet,

but compared to, you know, five years ago, um, you know,

self post two is so good. It is like ridiculously good. Um, uh,

we had Carson Stringer over for a visit recently and I was just totally

fangirling the whole time talking with her because it's so good. Um,

and there's a lot of things that right now,

especially if you care about objects that you need to know in order to do a

segmentation that just are about technique and not actually about

biology. And I would love to sort of strip most of that out. Um,

I think the tools will look very different.

I hope that we're still making them and still taking what's the best of what's

going on in the sort of worlds of deep learning and computer vision and bringing

it to people who can then use it to answer great biological questions. Um,

and I hope we're training more great computational biologists 'cause it's,

it's a lot of fun and there's a desperate need for them.

Like their companies are companies and universities are desperate to

have these sort of tech specialists.

People are desperate to become them and there's not really a bridge at the

moment to sort of like make them. So I,

I hope that we get to keep doing that too.

And one final question. It, it's a,

there's a lot of news about ai, uh mm-hmm.

There's scare mongering, maybe about AI warnings, about the dangers of ai.

You know, a lot of what you're doing is based around AI machine learning, it,

it mm-hmm. Part of the same mm-hmm. It's under the same umbrella.

Do you fear that it might be a risk of a public backlash against AI

that may then actually make funders be more risk averse and not,

not put funding in what the public are backlashing against?

Bit like gm, I think GM for a while got a GM is now back mm-hmm.

Had a very bad press. And if you're in the GM field, suddenly it's like,

oh my goodness,

funding has dropped through the floor because of public perception. Mm-hmm.

Do you fear there's a risk with that around ai?

That's a really good question. Um, I mean,

anything humans make and humans use has a,

has a possibility of going really well or really poorly. So I, I I think,

I think AI is, is just one of that class. Um,

I think that in general, the, even if the, the national funder,

like the grant funders stop funding this, um, the companies find it so useful.

Like, Google's not gonna stop. Facebook's not, or Meta now is not gonna stop.

Um, whether or not they're giving out grants to other people to work on stuff.

But, um, again, I think like any human tool, there are ways that people,

people are using it right now that are horrifying and that I can't stand and

there are ways that people are using it that make the world a better place. And,

uh,

we just have to sort of do whatever we can to try and make sure that the people

who are using it the right way win.

Yeah. I don't, it'll be interesting. I, I think it's so different to how the,

the chat things are one thing. Mm-hmm.

And that's the public are getting more scared.

I think using it as a scientific tool mm-hmm.

Is

A completely different objective. Mm-hmm. In my mind anyway. In my mind it's,

they are, you know, very, very different. But,

but the similarities are there that, that I, mm-hmm.

Yeah. Um, there, there was a, a paper not too long ago that, um,

someone was saying they were trying to train AI to make,

to figure out if drugs are gonna be toxic so that you make fewer toxic drugs and

everybody wants that. You want, you know, more drugs and cheaper and stuff.

But then they realized that, oh crap, we can't put this out.

'cause if you run it in reverse, you invent all sorts of chemical weapons,

the combat. Um, and I was just like, oh my gosh,

I never would've thought about running it in reverse. Um,

so we have to be careful. We, like, if we're making really powerful tools,

whenever you make something really powerful,

you have to think about how it could be used for harm.

And even the sorts of things that again,

like is not like a large language model,

but something that's just predicting drug safety could be misused in the wrong

hands. But, um, I think at this point, the genie's outta the bottle.

Well, absolutely. Right. Because now we know how to do it. Mm-hmm.

People use it for bad, so you might as well use it for good. Mm-hmm.

And hopefully the good outweighs the bad. I guess. Beth, we,

we are just over the hour. Thank you so much for joining me today. I,

I can't wait to see Elmi, so I'll see you soon. I'll be buying beer, uh,

don't get chocolate peanut butter, but who knows? I be over there in Amsterdam,

or near, sorry, Netherlands. Uh,

it's been super cool to chat and get to know you. Yeah, no,

It, this has been fantastic.

Just keep going with what you're doing and I hope that in 10 years time you are

still there, still making sound profile up all the next iteration,

whatever it may be. Mm-hmm. Uh, for us all to use and, and for all of us,

everyone mm-hmm. The benefits as, as that develops, its, its U utilization.

Thank you everyone for listening to the Microscopists. Uh,

don't forget best talked a lot about Anne Carpenter.

You can go listen to Anne Carpenter, mark, Ray and Roger also out of Ann's lab.

I think it's most people I've had out at one lab ever. Uh,

but that just shows the impact and the importance of the research.

And it's great to have Beth here today. You are the future for this,

and it's marvelous to have you here. Thank you.

Thank you for listening to the Microscopists,

a bite-sized bio podcast sponsored by Zeiss Microscopy.

To view all audio and video recordings from this series,

please visit bite-size bio.com/the-microscopists.