WEBVTT - BioRender CEO on Anthropic Deal, AI for Complex Imagery

0:00:02.520 --> 0:00:07.000
<v Speaker 1>Bloomberg Audio Studios, podcasts, radio news.

0:00:07.960 --> 0:00:11.680
<v Speaker 2>You're listening to Bloomberg Business Week with Carol Masser and

0:00:11.760 --> 0:00:14.160
<v Speaker 2>Tim Stenovek on Bloomberg Radio.

0:00:14.400 --> 0:00:16.439
<v Speaker 1>Garbage in, garbage out.

0:00:16.560 --> 0:00:18.480
<v Speaker 3>What could you be talking about, mister Stenevak.

0:00:18.600 --> 0:00:21.600
<v Speaker 1>Well, it's like you know computer science term from back

0:00:21.640 --> 0:00:23.040
<v Speaker 1>in the day, but we talk about it a lot

0:00:23.079 --> 0:00:26.720
<v Speaker 1>with AAI, like if data that goes into something is flawed,

0:00:26.720 --> 0:00:28.640
<v Speaker 1>the data that comes out is flawed. It's only as

0:00:28.680 --> 0:00:29.680
<v Speaker 1>good as the data that goes in.

0:00:29.800 --> 0:00:31.639
<v Speaker 3>Yeah, we do talk about this in the context of

0:00:31.800 --> 0:00:35.760
<v Speaker 3>artificial intelligence hallucinations. It's like when an LM gives you

0:00:35.760 --> 0:00:37.800
<v Speaker 3>an answer that you know is totally off base, or

0:00:38.440 --> 0:00:40.960
<v Speaker 3>gives you a picture image that's way off the mark

0:00:41.000 --> 0:00:41.320
<v Speaker 3>as well.

0:00:41.720 --> 0:00:44.479
<v Speaker 1>Sorry, I was just thinking to my friend, like, actually

0:00:44.600 --> 0:00:46.600
<v Speaker 1>asked for an invitation to want you go on to

0:00:46.600 --> 0:00:48.680
<v Speaker 1>create an invitation for a Halloween party, and he was

0:00:48.720 --> 0:00:51.960
<v Speaker 1>like he wanted the image generator to be to really

0:00:52.000 --> 0:00:53.280
<v Speaker 1>mess it up, like.

0:00:53.240 --> 0:00:53.880
<v Speaker 2>Oh totally.

0:00:54.120 --> 0:00:56.280
<v Speaker 1>He's like, no, make it more messed up, mess it

0:00:56.360 --> 0:00:56.960
<v Speaker 1>up even more.

0:00:57.520 --> 0:00:58.200
<v Speaker 2>Oh yeah it did.

0:00:58.280 --> 0:00:59.920
<v Speaker 1>Yeah, sometimes it does what it tells you to do.

0:01:00.080 --> 0:01:01.520
<v Speaker 3>Oh yeah, it worked.

0:01:01.560 --> 0:01:04.360
<v Speaker 2>Now I'm curious. Tvd'll find it one. I bet she's

0:01:04.400 --> 0:01:06.319
<v Speaker 2>AOKI thinks about this stuff a lot. Not the halloween

0:01:06.360 --> 0:01:08.720
<v Speaker 2>part of this, but you know, the hallucination stuff. She's

0:01:08.720 --> 0:01:10.520
<v Speaker 2>the founder and CEO of bio Render. It's a company

0:01:10.520 --> 0:01:13.800
<v Speaker 2>that creates software to generate scientific images and illustrations. It's

0:01:13.840 --> 0:01:16.920
<v Speaker 2>important that those are accurate. She was formerly the lead

0:01:17.000 --> 0:01:19.959
<v Speaker 2>medical illustrator for National Geographic. She did that for a decade.

0:01:20.040 --> 0:01:23.920
<v Speaker 2>She joins US from Toronto. She is welcome to the program.

0:01:23.959 --> 0:01:27.520
<v Speaker 2>I'm by a very interesting company. We don't talk about

0:01:27.680 --> 0:01:31.000
<v Speaker 2>this part of AI a lot. You've got this recent

0:01:31.000 --> 0:01:35.360
<v Speaker 2>partnership with Anthropic, the maker of the Claude LM. You

0:01:35.400 --> 0:01:38.840
<v Speaker 2>have this incredible background Johns Hopkins School of Medicine, a

0:01:38.920 --> 0:01:42.080
<v Speaker 2>dual BFA and fine Art and pre med from Queen's University.

0:01:42.760 --> 0:01:46.399
<v Speaker 2>Why is scientific imaging important in the context of AI.

0:01:48.440 --> 0:01:50.480
<v Speaker 4>Yeah, Firs, thank you for having me, Carly and Tim,

0:01:50.960 --> 0:01:56.880
<v Speaker 4>and you know, for science in particular, accuracy is really paramount,

0:01:56.920 --> 0:02:00.640
<v Speaker 4>and so we're really excited to partner with companies like

0:02:00.680 --> 0:02:04.520
<v Speaker 4>Anthropic because this is just actually one big step in

0:02:04.600 --> 0:02:08.880
<v Speaker 4>making AI safer and more reliable for science. You said

0:02:08.919 --> 0:02:11.880
<v Speaker 4>that you know your friend made some Halloween invites or

0:02:11.960 --> 0:02:13.920
<v Speaker 4>garbage in garbage oe and that, and that truly is

0:02:13.960 --> 0:02:18.160
<v Speaker 4>the case right now. It really doesn't do scientists work justice.

0:02:18.360 --> 0:02:23.440
<v Speaker 4>How bad the state of bioimages are today and actually

0:02:23.440 --> 0:02:27.040
<v Speaker 4>the you know, Dario Nthropic CEO said that AI could

0:02:27.080 --> 0:02:30.080
<v Speaker 4>compress the next fifty to one hundred years of biological

0:02:30.080 --> 0:02:33.600
<v Speaker 4>progress into just five to ten, which means that the

0:02:33.639 --> 0:02:38.040
<v Speaker 4>bottleneck really ships from computation to human comprehension.

0:02:38.280 --> 0:02:42.359
<v Speaker 1>Do you talk about Dario Amiday, I'm sixty minutes last night. Yeah,

0:02:42.400 --> 0:02:45.600
<v Speaker 1>I've interviewed him before over in the in March in

0:02:45.840 --> 0:02:49.359
<v Speaker 1>San Francisco. Do you believe him? Do you believe Do

0:02:49.360 --> 0:02:50.919
<v Speaker 1>you agree with him?

0:02:52.080 --> 0:02:55.240
<v Speaker 4>I do? I do, and I agree with him in

0:02:55.360 --> 0:02:59.080
<v Speaker 4>that we know that AI is great at helping with

0:02:59.240 --> 0:03:03.280
<v Speaker 4>discoveries written form, and I think you'll agree that we're

0:03:03.280 --> 0:03:06.920
<v Speaker 4>still not quite there with making visuals, particularly in science

0:03:06.960 --> 0:03:09.640
<v Speaker 4>where it matters so much to be accurate. We've seen

0:03:09.680 --> 0:03:13.240
<v Speaker 4>horrific diagrams of you know, legs with extra bones or

0:03:13.600 --> 0:03:16.600
<v Speaker 4>maybe less obvious mistakes, but you know, more dangerous ones

0:03:16.639 --> 0:03:20.320
<v Speaker 4>are like protein folding the wrong way. It can actually

0:03:20.360 --> 0:03:23.840
<v Speaker 4>mean the difference between a functioning cell in Alzheimer's disease,

0:03:24.320 --> 0:03:26.840
<v Speaker 4>or even like one arrow moving in the wrong direction

0:03:26.919 --> 0:03:30.200
<v Speaker 4>and about metabolic pathway could mean that, you know, we're

0:03:30.200 --> 0:03:34.120
<v Speaker 4>feeding the tumor instead of starving it. And for buyer renders,

0:03:34.160 --> 0:03:36.560
<v Speaker 4>you know, it turns out that this is exactly what

0:03:36.680 --> 0:03:39.360
<v Speaker 4>AI needed. We've been spending the last eight nine years

0:03:39.400 --> 0:03:44.960
<v Speaker 4>building the world's largest library of expert expert vetted biovisuals,

0:03:45.480 --> 0:03:47.840
<v Speaker 4>and we're very proud to be sort of the trusted

0:03:47.880 --> 0:03:51.160
<v Speaker 4>source now, the foundational layer for going forward. You know,

0:03:51.240 --> 0:03:54.840
<v Speaker 4>all the frontier models where accuracy is really non negotiable

0:03:54.840 --> 0:03:55.680
<v Speaker 4>in science.

0:03:55.440 --> 0:03:58.080
<v Speaker 3>Well, tell us a little bit about your company in

0:03:58.160 --> 0:04:01.960
<v Speaker 3>particular what you're doing and this idea of again, these

0:04:01.960 --> 0:04:03.280
<v Speaker 3>are some of these things I feel like we just

0:04:03.320 --> 0:04:07.840
<v Speaker 3>take for granted images within science, right or medicine. But

0:04:07.960 --> 0:04:13.040
<v Speaker 3>tell us about the need to have these images make

0:04:13.360 --> 0:04:16.120
<v Speaker 3>you know, and are they constantly being updated, Like, give

0:04:16.200 --> 0:04:18.320
<v Speaker 3>us an idea of what you guys are doing and

0:04:18.320 --> 0:04:22.000
<v Speaker 3>what we need to understand about medical imaging, not MRIs,

0:04:22.040 --> 0:04:24.839
<v Speaker 3>but just understanding everything and anything that's in the body.

0:04:25.960 --> 0:04:28.279
<v Speaker 4>Yeah, and that's a great point. You MRIs can take

0:04:28.360 --> 0:04:30.560
<v Speaker 4>us to a certain point, but at the end of

0:04:30.560 --> 0:04:33.000
<v Speaker 4>the day. The body is very complex, surgical fields are

0:04:33.040 --> 0:04:36.600
<v Speaker 4>very bloody. Photos don't cut it, and so we have

0:04:36.640 --> 0:04:38.920
<v Speaker 4>to rely on a little bit more of I guess

0:04:39.000 --> 0:04:43.600
<v Speaker 4>artistically licensed rendered graphics and bio renders, a software that

0:04:44.200 --> 0:04:48.600
<v Speaker 4>help scientists create beautiful biological diagrams really quickly. You can

0:04:48.640 --> 0:04:50.640
<v Speaker 4>think of us as sort of the canva or figma

0:04:50.720 --> 0:04:53.520
<v Speaker 4>for biology for those folks that aren't familiar with the

0:04:53.560 --> 0:04:57.479
<v Speaker 4>scientific graphics world. So you can imagine a scientist coming in.

0:04:57.520 --> 0:05:00.599
<v Speaker 4>They drag and drop like an anatomically correct brain, for example,

0:05:00.680 --> 0:05:03.360
<v Speaker 4>onto the canvas. You can zoom in and then look

0:05:03.360 --> 0:05:06.039
<v Speaker 4>at the details of a brain cell interacting with a

0:05:06.080 --> 0:05:08.360
<v Speaker 4>tumor cell, and then you can zoom in even closer

0:05:08.800 --> 0:05:11.880
<v Speaker 4>and show the biochemical reactions happening within the brain cell.

0:05:12.480 --> 0:05:15.240
<v Speaker 4>But you can see how quickly it's getting very technical

0:05:15.279 --> 0:05:18.760
<v Speaker 4>and complex, and before bio under scientists we're actually using

0:05:19.640 --> 0:05:22.360
<v Speaker 4>and they still do believe it or not, primarily PowerPoint

0:05:22.440 --> 0:05:25.799
<v Speaker 4>of all tools. So they spend you know, days, sometimes

0:05:25.880 --> 0:05:30.279
<v Speaker 4>weeks just trying to create one image with you know,

0:05:30.320 --> 0:05:33.320
<v Speaker 4>like the circles and squares and the lines and arrow shapes,

0:05:33.760 --> 0:05:35.120
<v Speaker 4>and so we thought, you know, there's got to be

0:05:35.160 --> 0:05:38.560
<v Speaker 4>a better way. So my co founders and I, Ryan

0:05:38.600 --> 0:05:41.159
<v Speaker 4>Katty and I launched by a Under about eight years ago,

0:05:41.560 --> 0:05:44.040
<v Speaker 4>hoping to solve this communication gap for scientists.

0:05:44.440 --> 0:05:49.760
<v Speaker 3>Who are your customers and how do you guys make money?

0:05:50.120 --> 0:05:51.240
<v Speaker 2>We do make money.

0:05:52.080 --> 0:05:55.440
<v Speaker 4>Yeah, yeah, we've been lucky to be profitable pretty early. Actually,

0:05:55.640 --> 0:05:58.320
<v Speaker 4>it was very fortunate that the scientific community saw the

0:05:58.440 --> 0:06:04.719
<v Speaker 4>need and just up to and our customers are primarily researchers,

0:06:05.000 --> 0:06:09.039
<v Speaker 4>including leaders. Actually they're in there making PowerPoint diagrams. Hopefully

0:06:09.040 --> 0:06:13.360
<v Speaker 4>not much longer if these biorender in pharmaceutical companies, biotechs,

0:06:14.440 --> 0:06:16.799
<v Speaker 4>academic institutions, and even the publishing world.

0:06:17.480 --> 0:06:21.680
<v Speaker 1>Yeah, I'm curious about copyright with this. There's an article

0:06:21.720 --> 0:06:24.799
<v Speaker 1>in Chemistry World from just about a year ago about

0:06:24.839 --> 0:06:27.720
<v Speaker 1>how thousands of published studies could contain images within correct

0:06:27.760 --> 0:06:32.760
<v Speaker 1>copyright license licenses, and they mentioned some biorender images created

0:06:32.800 --> 0:06:36.000
<v Speaker 1>by biorender. Have you guys figured out the copyright issues

0:06:36.040 --> 0:06:37.919
<v Speaker 1>in terms of what you're looking at and then what

0:06:37.920 --> 0:06:38.560
<v Speaker 1>you're producing.

0:06:40.160 --> 0:06:43.360
<v Speaker 4>Yeah, it's a great question. I think. You know, at

0:06:43.360 --> 0:06:46.080
<v Speaker 4>the end of the day, we have to shift the

0:06:46.160 --> 0:06:51.400
<v Speaker 4>narrative to away from sort of the copyright or it's

0:06:51.440 --> 0:06:54.440
<v Speaker 4>sort of the antiquated era of you know kind of

0:06:54.480 --> 0:06:57.560
<v Speaker 4>who published it first. We're really trying. Our mission is

0:06:57.600 --> 0:07:00.760
<v Speaker 4>to empower the world to communicate science and understand it

0:07:00.839 --> 0:07:04.560
<v Speaker 4>faster through visuals. And so what's happening actually now in

0:07:04.680 --> 0:07:10.880
<v Speaker 4>our library is scientists. We have four million plus users

0:07:10.880 --> 0:07:14.239
<v Speaker 4>on our platform. They've been knocking down our doors wanting

0:07:14.240 --> 0:07:18.160
<v Speaker 4>to share their diagrams in our community library. They say

0:07:18.280 --> 0:07:20.640
<v Speaker 4>they don't care if scientists you know, rip it apart

0:07:20.720 --> 0:07:24.760
<v Speaker 4>and use it. They just feel so compelled to change

0:07:24.880 --> 0:07:27.640
<v Speaker 4>the narrative and change the status quo of how science

0:07:27.640 --> 0:07:31.360
<v Speaker 4>is communicated today that they are willing to share the

0:07:31.440 --> 0:07:33.920
<v Speaker 4>work that they do environ render back into the community.

0:07:33.960 --> 0:07:36.480
<v Speaker 4>So it's almost looking more like instead of the Canva

0:07:36.600 --> 0:07:39.840
<v Speaker 4>or Figma, more of the GitHub or even Wikipedia. Because

0:07:39.880 --> 0:07:44.600
<v Speaker 4>they share their work back into our library and either

0:07:44.640 --> 0:07:47.120
<v Speaker 4>we spot errors or the communities quick to jump on

0:07:47.160 --> 0:07:50.080
<v Speaker 4>those errors, we update those and then get back upload

0:07:50.120 --> 0:07:53.160
<v Speaker 4>into the repository. So we're kind of shifting the focus

0:07:53.160 --> 0:07:56.400
<v Speaker 4>away from you know who said it first, to let's

0:07:56.640 --> 0:07:59.920
<v Speaker 4>be open and share our work and really further neck

0:08:00.000 --> 0:08:00.840
<v Speaker 4>celebrate science.

0:08:00.960 --> 0:08:04.840
<v Speaker 3>Yeh, super interesting. Hopefully we can come back and continue

0:08:05.000 --> 0:08:08.160
<v Speaker 3>this in the future. She is AOKI founder and CEO

0:08:08.280 --> 0:08:14.320
<v Speaker 3>bio Render, joining us from Toronto, Ontario, Canada.