“What we’re still seeing is that some of these models either pick up on the wrong signals. So if you were to change context, they’re not able to do what’s called transfer learning very effectively. In other cases, it’s just too narrowly defined based off of the data that was initially given to them.
Or the third category, which we’re really excited about, is that models haven’t been trained yet with this human, you could say, chain-of-thought reasoning, which is kind of a word that I’m taking from how large language models were trained, but they weren’t trained to think in the step-by-step way that humans kind of begin to understand the world around them.”
I had the pleasure of talking with Claire Swadling, whose path toward founding an artificial intelligence company began with a rather unsettling observation in a college research lab. The undergraduate interns were spending hours labeling images, drawing boxes around objects, and performing painstaking visual tasks. That was expected. What surprised her was that the postdoctoral researchers, people who had invested years in advanced scientific training, were doing much the same thing. Swadling began asking herself an uncomfortable question: Why spend another seven or eight years pursuing an academic career if so much of the work would remain stubbornly manual?
That frustration eventually became the foundation of Visual Intuition, the AI company she co-founded to teach machines something considerably more difficult than recognizing pictures: how to look at visual information and reason through it the way a trained human expert might. Yet there is an interesting contradiction at the center of her ambitions. Swadling wants to automate some of the most tedious work performed by scientists, while remaining deeply convinced that human judgment, expertise, and what she calls “taste” are qualities machines still struggle to reproduce. Her vision for AI depends, in some respects, on understanding precisely what makes human intelligence different.
Growing up in southeastern Michigan, Swadling was already pursuing scientific research while most of her peers were navigating high school. She worked in computational biology laboratories analyzing DNA and RNA sequences, opportunities she credits partly to an early willingness to contact professors without introductions. Looking back, she describes herself as ambitious and perhaps insufficiently conscious of the possibility of embarrassment. Her attitude was simple: “Who cares if I embarrass myself in front of one more professor?” The willingness to send an unsolicited email would become one of the habits she carried into entrepreneurship.
At Harvard, where she studied physics and economics, Swadling initially hoped physics would give her a broader intellectual foundation. Yet she found herself returning to computational biology, where her growing dissatisfaction with repetitive research tasks began to take shape. Around the end of her undergraduate years, she met Ishaan, her future co-founder, who was serving as a teaching assistant in her wave mechanics course. They shared a physics adviser, who encouraged them to talk. Ishaan had encountered similar frustrations in connectomics, a branch of neuroscience concerned with mapping the brain’s intricate network of neurons. Their conversations gradually suggested a possibility neither could pursue as effectively from inside a single laboratory: building technology that might help many researchers at once.
They began forming Visual Intuition in May 2026. By the time of our conversation, the venture had grown into a three-person team with an office in Cambridge, Massachusetts, and had raised $3.2 million in funding that summer. Swadling also reported that the company’s machine-learning research had received an award at ICML, a major conference in the field. Those are meaningful early milestones, although she is careful to acknowledge that the company is still validating its approach. Its first customers will likely be academic laboratories, with potential expansion into medicine, industrial quality control, and agriculture.
The underlying problem is remarkably concrete. In connectomics, a researcher may have to follow a single neuron through thousands of microscopic images, identifying its boundaries and determining where it continues from one image to the next. A mistake can compromise the accuracy of the reconstruction. Visual Intuition’s approach involves learning from the behavior of human specialists, including where they point, what they click, and how long they hesitate before making a decision. Even a momentary pause may contain information. An expert who slows down upon encountering an unusual white area, perhaps an imaging artifact or air bubble, is demonstrating a judgment that an AI system might learn to recognize.
For Swadling, this is part of the difference between pattern recognition and genuine visual reasoning. She remembers earlier computer-vision models that appeared capable of distinguishing dogs from wolves, only for researchers to discover that the systems had learned to rely on environmental clues rather than the animals themselves. Modern models have improved enormously, but the problem of recognizing the right signals remains. Her goal is to develop systems that can learn from the way experts approach unfamiliar situations, rather than depend exclusively on patterns in their original training data.
Turning that research ambition into a business introduces another challenge. Swadling has to move between the language of neuroscientists and physicists and the priorities of investors and prospective customers. She recalls attempting ordinary small talk with a scientist, asking about the weather and local sports, only to encounter what she describes as a “blank wall.” Some people, she learned, would rather skip the pleasantries and get straight to the science. The experience offered a small but useful lesson in adapting her communication to the person across from her.
More consequential is the daily tension between pursuing intellectually interesting research and delivering a practical product. An unexpected technical question might deserve weeks of investigation, but a potential customer may need a working result sooner. Swadling describes that balancing act as one of the company’s central challenges. For someone drawn to science by curiosity, the discipline of deciding when research is sufficient may prove as important as the research itself.
Her enthusiasm for increasingly capable AI is accompanied by concerns about what happens when machines take over the work traditionally assigned to beginners. She believes AI has already reached extraordinary levels of performance in certain narrow fields, yet worries about how professions will train their next generation of experts if entry-level assignments disappear. She raises the example of junior attorneys learning the meticulous habits of senior lawyers. It is a striking concern coming from someone whose own business emerged from dissatisfaction with the repetitive tasks given to young researchers.
Beyond the laboratory and the company, Swadling speaks openly about the importance of faith, humility, and human connection. Prayer is a regular part of her mornings and evenings, providing time for gratitude and reflection. She does not pretend to have perfected the routines of a busy founder, acknowledging that there are days when she neglects her health more than she should. And when asked what kind of movement she would most want to inspire, she chooses empathy, the ability to understand another person’s needs and perspective before trying to move forward together.
There is something fitting about that answer. Swadling is building machines that learn by paying closer attention to what human beings do, while arguing that the human capacity for discernment and understanding remains essential. The question running through her work is how much intelligence can be transferred to a machine without losing the qualities that make expertise valuable in the first place. For a young founder attempting to give scientists more time to think, it is a question with consequences far beyond the laboratory.
Yitzi Weiner: Claire, it’s a delight to meet you. Before we dive in deep and talk about your amazing work, our readers would love to learn about your personal origin story. Can you share with us the story of your childhood and how you grew up and the genesis of all the amazing, we could say, genius that has come since then?
Claire Swadling: Yeah, I’d be happy to talk about kind of my upbringing and how it led me to where I am. So, I grew up in southeast Michigan and I was always interested in science and had some inklings of entrepreneurship before. So in high school I started working for a few labs doing computational biology research mostly at the gene sequencing level, which meant that I was looking at lots of like DNA or RNA sequences and then also doing some data analysis.
But when I came to college I thought, you know, I want to learn something that’s more broadly applicable. So I studied physics, but somewhere along the way I still found myself back in a computational biology lab. Something I realized kind of halfway through college is that not only were the undergraduate interns being assigned some really manual tasks of, you know, here’s a bunch of images and go through and, you know, label them or segment them or draw boxes around them, not only were the undergrad interns doing that, but the postdocs were also.
So at some point I thought, you know, why would I want to invest another seven to eight years in an academic career when I’d be doing the same thing? So I became a little frustrated with a lot of the manual tasks that aren’t able to be automated in science. Towards the end of my undergrad, I also met my co-founder Ishaan, who had a similar frustration coming from connectomics, which is a field of neuroscience that I’m happy to go in more depth on in a little bit. Ultimately, that’s how I kind of discovered science and then realized that you can probably do more for science by automating some of those more tedious steps along the way.
Yitzi Weiner: Can you clarify a bit what you mean by automating science?
Claire Swadling: Yeah. So, what we hope to do is put productive hours back into the hands of scientists. Nowadays scientists are doing a lot of data processing, especially when they work in a visual field, which means that there aren’t models right now that currently are able to do some of these annotation tasks.
An example is a radiation oncologist looking at a tumor and being precise about the boundaries of that tumor so that, of course, healthy tissue isn’t being shot with chemotherapy. Similarly, in more basic fields of biology, if you have to classify cells or track them through different images, a scientist has to do that. It’s very tedious for them and they’re just the only person that is able to do that, and it takes away from the time from designing experiments or thinking about the future of science and more productive tasks like that.
Yitzi Weiner: Are you saying, so for example, let’s say a radiologist can’t upload an X-ray scan to an AI model and that can’t read if that’s cancerous or healthy tissue because that doesn’t exist?
Claire Swadling: It depends on the field, but right now there still is a very significant human in the loop just because, especially in fields where you are dealing with human data, the threshold for accuracy is incredibly high. You want to be close to 100% certain before you take a human out of the loop there. So that’s why some fields are slower to adopt, and also why pixel precision can be very important.
Yitzi Weiner: So, let’s jump for a minute. Let’s jump to the present. Just tell us about the amazing things you’re doing today. But then I’d like to go back to the story that led to all this amazing work. So, tell us what you’re doing today.
Claire Swadling: Yeah. So, Visual Intuition is an AI company that aims to teach machine learning models how to visually reason. When I say visually reason, in the same way that you and I, when we see an image, we aren’t just going, you know, looking left to right across say like the New York skyline and trying to categorize every pixel. We actually are looking for heuristics, or like let’s say there’s the Empire State Building. That’s how I know I’m in New York. In the same way, we want to teach machine learning models to actually think about images like a human, to catch on to different hallmarks in order to be more efficient.
In the same way that you would, you know, teach someone, maybe your grandma, to navigate a web browser and kind of sit next to them and point things out to them and correct them as you go, that’s also how we want to train models to be able to do things that they can’t do already, which is, like I mentioned before, doing very pixel-precise scientific tasks.
We are coming from a field of neuroscience called connectomics, which is broadly concerned with mapping the human brain neuron by neuron, which of course has a very high degree of precision that’s needed. What was traditionally done in this field is you would have a graduate student go image by image and track, you know, maybe a neuron starts on slide one and it spans all the way to slide, you know, 4,556. And it would be a real person who’d have to make that decision.
Because we’re coming from this field, what we realize is that when you have this really long time horizon, humans actually do a lot of thinking in that with their eyes, where they are pointing on a website, and the timestamps between the clicks that they’re making. So we train our models on this sort of holistic approach of how do human experts actually think about these visual processes and how do we teach our models to do something similar to that. We’re starting in neuroscience and automating some of those pesky tasks and hoping to expand to other biological sciences and also other industries. So for example, quality control or agriculture and some of these other industries.
Yitzi Weiner: That’s fascinating. So you’re saying your technology could help instead of having a scientist go slide by slide to map the human brain, you could have an AI do that. Am I understanding it correctly?
Claire Swadling: Yeah, that’s correct.
Yitzi Weiner: That’s amazing. Wow. So can you explain to our readers what you mean that AI doesn’t see things the same way we do? Because a person could argue, and again of course I agree with you, people argue that if I upload a picture, it sees the picture, you know, even if I upload a whole video, it could kind of understand the video, so maybe it sees it. So what do you mean that it sees differently than that?
Claire Swadling: Oh yeah, this is a fun question because computer vision models, the models that are able to comprehend images and understand them the way we hope to, have come a very long way. I still remember a time where you would try to teach a machine learning model, you know, is this a dog or a wolf? And then over time you would figure out that the model was just training on if the background was sunlight or if it was the dark.
We’ve come a very long way from that. What we’re still seeing is that some of these models either pick up on the wrong signals. So if you were to change context, they’re not able to do what’s called transfer learning very effectively. In other cases, it’s just too narrowly defined based off of the data that was initially given to them.
Or the third category, which we’re really excited about, is that models haven’t been trained yet with this human, you could say, chain-of-thought reasoning, which is kind of a word that I’m taking from how large language models were trained, but they weren’t trained to think in the step-by-step way that humans kind of begin to understand the world around them.
Yitzi Weiner: So you mentioned that human beings, our brains work with heuristics, right? It sounds like you’re saying there are certain mental shortcuts. Am I understanding you correctly?
Claire Swadling: Yeah, that’s correct.
Yitzi Weiner: So what kind of heuristics would, let’s say, be programmed for an AI to have like a shortcut and be able to see and process visual information quicker?
Claire Swadling: Yeah. So, this kind of relates again to our method of training models based off of the actual human processes. This would actually depend on the context of a task. Let’s go back to say labeling neurons. A heuristic that a model could learn is let’s say this human annotator is realizing that whenever there is this large white area, that represents an artifact, or like an air bubble. A human, whenever they encounter that, spends extra time labeling that image.
That actually teaches the model in a way to be careful and to slow down and to be more precise when they’re dealing with a similar situation. I would say it’s mostly context-dependent, but it can take signals based off of the data that you feed it. And the more times you give it examples of humans thinking through a similar problem, it’s able to reproduce that.
Yitzi Weiner: Okay. So, let’s talk a little bit more about your work. What’s your business model? Like how do you make money? Who are your customers? Who are your clients?
Claire Swadling: Yeah. So, right now we are a venture-backed company. We raised $3.2 million over the summer and are still in the process of, you know, validating our approach. We have a machine learning paper that won an award at ICML, which is a large machine learning conference. Our first customers will likely be academic labs that are in these immediate spaces of neuroscience and biology, but we’re hoping to expand to larger enterprises again in these potential areas of medicine and quality control.
What we hope to do is essentially provide our model on a subscription basis. People would be able to give our model a chance to see their data. Then we can learn from their workflows and then fully automate those parts that are most tedious.
Yitzi Weiner: Are you competing with the massive LLMs like Google AI? Are they also doing research in this area?
Claire Swadling: Yes, that’s a good question. A lot of the frontier labs are focused on making products more generally available for consumers. Because of that, the average consumer typically has more need for let’s say like an agent or an LLM to be able to simply navigate a web browser, be able to point things out without this, you could say again, like this level of pixel precision. We would differentiate ourselves by saying we are made by and for scientists and to deal with these tasks where the human burden of review is just that much higher.
Yitzi Weiner: That’s amazing. I’ve experienced this. It’s not pixel by pixel, but it’s reminiscent of it. I tried translating a 700-page book with Argon. Argon is a very strong, you know, very powerful model, and it did a decent job, but it wasn’t perfect. Every time I asked, “Is it perfect now?” Oh, I found another mistake and another mistake. I didn’t realize that. And it says, “I’m reading it with a human eye,” you use those terms. It was eye-opening to me how it reads differently than we do. As much as you try to validate it, it’s never 100%, you know, so I need like, that’s not pixel by pixel, but I need a letter-by-letter validation. And it’s hard.
Claire Swadling: Yeah. I think LLMs especially are very good at having really strong working memories. You can tell it something in like hour one and it’ll remember it, you know, three or 100 days later, but it still doesn’t quite have the taste that humans do. And I think they’re still working out ways in their reward architecture to make sure that they have taste. So for now, I still think humans are probably the best person to do that final review.
Yitzi Weiner: So can you tell us a story? Tell us the story of how you started this company. Tell us how it started.
Claire Swadling: Oh yeah. So as I mentioned a bit before, around the time that I got frustrated with a lot of the manual work in science, I met my co-founder, who was actually the teaching assistant for a wave mechanics course I was taking. We also found out that we had the same physics advisor, who is now the chair of the department, who actually encouraged us to talk more. He knew that I was unsure about my next career steps and that I was exploring more areas of business. I had done a banking internship and wasn’t sure if that was where I should end up since I was also really interested in science.
He knew that Ishaan, my co-founder, had also been seeing these similar problems in science, and again this field of neuroscience. The more Ishaan and I started talking, the more that we realized that, you know, at some point we can impact a lot more than just our immediate labs if we actually turned it into a business. We started formation of the company around May of this year, and since then it’s grown. We’re a team of three, rapidly expanding. We now have an office in Cambridge, and yeah, that’s kind of more about the origins and how we met.
Yitzi Weiner: So there’s a saying that sometimes our mistakes can be our greatest teachers. Do you have a story about a humorous mistake that you made when you were first starting and the lesson that you took away from that?
Claire Swadling: Yeah, that’s a good question. Let me think for a second. I will say something that I’m still learning is how much I have to kind of change how you communicate with different people. Because of the nature of my job, I communicate with a lot of neuroscientists and physicists, but also a lot of business stakeholders, so investors or people that are say business leaders at companies.
Because of that, I’ve been learning over time how to better switch contexts. I remember at the beginning of this journey, I was talking to a scientist and I kind of made the mistake of going in with some of the classic like small talk questions, like asking about the weather and, you know, like the sports team in the area, and I was kind of met with just like a blank wall. So I was like, okay, note to self, you know, some people prefer me to go like straight to the science. I think that’s something that’s been part of the learning process.
Yitzi Weiner: That’s great. Amazing. What’s been the most challenging role or project or sub-project you’ve taken on so far and why?
Claire Swadling: Let me think about this as well. Okay, I’m taking a moment just because we have a few things that are confidential. I think one thing that I can say is that we’ve been thinking a lot how to intentionally balance our research aims along with our product focus. Our ultimate goal, of course, is to put useful time back into the hands of scientists and be able to speed up productive research, and hopefully especially in this area of biology that we’re coming from as well.
Because of that, we make a lot of decisions on the day-to-day of, you know, should we go down this extra research rabbit hole because maybe it’s interesting for X, Y, or Z, but on the other hand, maybe it actually could lead to a product that’s like more accurate or faster. We try to balance that with, okay, how quickly do we want to say deliver the initial test result so we can deliver this quote.
I think that balance between research, but also being product-driven, is something that we’re very intentional about and try to balance also depending on the potential client that we’re working with, if they are more of like a research lab or if they are a company, which informs the speed that we work and also how much we can afford to think more about these research-oriented milestones.
Yitzi Weiner: Let’s talk a bit about AI safety. I’ve heard the debate between the different camps and different stakeholders in AI safety described as a debate between the zoomers and the doomers, you know, are we all doomed or no, we should just accelerate, accelerate, achieve AGI and then that will solve the problem. What’s your perspective on AI safety? Yeah, what’s your perspective on it?
Claire Swadling: That’s a very timely question, especially since OpenAI recently released a list of, you know, they had the 722 math problems that they were able to solve, or claim that they solved, and then more recently the advancement in the standard model in physics and then computer science, and then also another area of progress in math.
I’m of the firm belief I think we already have narrow superintelligence. All of these models, from what we’ve been told, have had humans in the loop, human academics that are orchestrating them. I think that will remain to be important, again coming back to this idea of taste, of you know, human experts are the ones that are still being the shepherds to these models, as they’re calling them.
However, as we move into other domains where I think the stakes can be higher for potential human harm, as we move into, you know, robotics in the real world, that’s where I think some form of regulation I expect to crop up. You know, I’d be lying if I said I had a complete picture and therefore a complete opinion on things moving forward. But I do think as we give models more free rein in the real world that we will need a little bit more regulation to keep up.
Yitzi Weiner: So if I’m understanding you correctly, you’re saying right now when it’s not embodied AI, then we don’t have to worry too much. But once we start having robots and having cars drive by themselves, then we have to have more regulation. Am I understanding it correctly?
Claire Swadling: I think that’s correct in terms of worrying about physical safety, but at the same time I think the greater danger with some of the areas that have already been automated is this danger of how are we going to train human juniors, you know, people entering the workforce. There are so many firms now that are automating, for example, parts of the law profession, but then how do you train new attorneys to be able to, you know, be as detail-oriented as current seniors are? I think there’s still a salient danger in that sense, but I would say the physical danger is, in my opinion, closer to the embodied AI in the physical world.
Yitzi Weiner: Got it. You’re saying right now the main danger is job displacement. Amazing.
Claire Swadling: Yeah, and in training juniors. Yeah.
Yitzi Weiner: Amazing. Now that it’s political season, a big debate is about data centers. What’s your case for or against data centers?
Claire Swadling: Okay. Interesting question. You know, I’m from Michigan and there’s a very big data center that is trying to be built there. I’ve also recently talked to someone that works for the US government in their public lands. I think it is a very delicate balance because on one hand, I do believe in American exceptionalism and that we have to make sure that we are supporting our economy and especially the tech sector as robustly as we can.
At the same time, we have a responsibility with some of the natural lands we’ve been given, you know, by the Constitution. I know I’m kind of waffling on this point here, but I think it really is something that’s going to have to be found on a case-by-case basis, and I really do think that there are some communities that are probably just not good candidates for these types of data centers. I don’t think that they should be unilaterally stopped. I do think we need to move forward, but I think community voices should be taken into account at each of these steps because definitely I think some counties are better candidates than others for them.
Yitzi Weiner: That’s great. You mentioned that you think that we’ve achieved or are almost achieving superintelligence. Can you articulate to our readers how superintelligence or AGI is different than what we have now? Because right now it’s very intelligent, but how is AGI different than AI?
Claire Swadling: Yeah, that’s another timely question because I almost feel like anytime a new innovation is made, people kind of push the definition of AGI further just because maybe there are some truths that they don’t want to acknowledge. I would define AGI or superintelligence, which I kind of put in the same category, because I think functionally it’s when AI is able to complete the whole cohesive task that you expect a human expert to do.
I think a large part of that is proactiveness. As we get closer to AI that does have, you could call again, taste, which in this example of math it’s been getting closer to because the results it’s coming up with are actually towards proving these mathematical problems that were put in place by humans who determined that they represented something meaningful.
I think as AI gets closer to not only being able to properly solve or execute the problems in the field, but actually start to make judgments about it and proactively find directions that human experts in that group would agree are the logical next steps, that’s when we get closer to the labels of AGI or SI. I think that the labels for that are also changing every day, but that’s how I think about things broadly.
Yitzi Weiner: That’s great. Amazing. Okay, so this is our signature question. Claire, you’ve been blessed with a lot of success now and you must have learned a lot from your experience. Looking back all the way to the beginning when you first started, can you share five things that you’ve learned over the years that someone starting out today can really benefit from? Five things you’ve learned over the years that someone starting out today can benefit from.
Claire Swadling: Okay, this is a great question. I think the first thing, which is probably the thing that helped me the most growing up, is never be afraid to send that first cold email. In high school when I started doing research, that was also through me just being ambitious and not socially aware enough just to think, you know, who cares if I embarrass myself in front of one more professor? I think that helped me into my professional life in terms of not being afraid to knock on the first door. So I think that’s number one.
I would say a second one is also to have some practice of humility. I think for me that happens through my faith, but I think other people are able to achieve that other ways, whether through meditation or reflection. I think that’s been helpful in terms of not letting yourself go to any extreme, but also to be able to process both the highs and lows well.
Let’s see. A third piece of advice I would say is anytime you have a major life event that happens and you are trying to process it, to really give yourself the time to let your emotions run their course, but then afterwards be honest about if there’s something you could have changed, and if there wasn’t, to really do your best to be forgiving with yourself and let it go.
Another thing I would note is that at the end of the day, what I believe machines will never be able to take away from us are our human relationships. So, that’s something to really prioritize. Also, I’d probably go back and tell myself in college just to, you know, not have watched the extra episode of Hulu and to have gone outside and talked to more people.
And then I guess my last piece of advice would be really to surround yourself with the good that you want to be. I think we often talk about this in terms of the people that we surround ourselves with, but for me this also means the type of media that you consume. If you’re, let’s say, always scrolling on one particular negative side of X, or if you’re always listening to sad music, you’re kind of setting yourself up for the type of day that you want to have. Throughout college I realized that if I start the day on the right note, that sets me up for success rather than, you know, listening to the catchiest song that’s out even if it doesn’t have a good message, just to think about what effect that will actually have on me. So, those are some just very broad things I would say.
Yitzi Weiner: That’s beautiful. Can you share some of the self-care routines that you do to help your body, mind, and heart to thrive?
Claire Swadling: Okay, that’s interesting. I’d say probably the number one thing that comes to mind is I do make prayer a large portion of my life. In the morning, taking time to be thankful, and in the evening, taking time to kind of account for some of my actions and think about them, as well as again to have a gratefulness practice. Then in terms of the rest of my routine as a founder, it’s kind of up in the air. There are some days I prioritize my health less than I should, but I would say a spiritual practice is probably the only one that I’m routinely doing.
Yitzi Weiner: Beautiful. This is our final aspirational question. Claire, because of your amazing work and the platform that you’ve built and the company that you lead, you’re honestly a person of great influence. If you could spread an idea or inspire a movement that will bring the most amount of good to the most amount of people, what would that be? Because you never know what your idea can spread.
Claire Swadling: Yeah, I’ve never really thought of myself as a person of influence, but if I had to say something, the movement I would want to start is one probably centered on empathy, and really just getting a chance to think through not only your immediate needs, but those of others. I think that’s something that continues to help you whether you think about your personal or your professional life, because whether you’re trying to like sell a product or even just like talk to a friend about something, you can’t really go anywhere until you start on kind of the ground that each of you are on. So that’s my message.
Yitzi Weiner: Amazing. How can readers continue to follow your work? How can they reach out if they want to learn more or engage your services? How can our readers support your work in any possible way?
Claire Swadling: The best way to reach us now is through my LinkedIn, and I’m happy to send you that link afterwards.
Yitzi Weiner: Claire, it’s been an honor to meet you. I wish you continued success and good health and blessings, and I’d love to do this again next year.
Claire Swadling: Yeah, that was wonderful. Thank you for the wonderful interview. It’s my first one of this kind, so I appreciate how great you’ve been about this and it’s not been intimidating, so I’ve appreciated that.
Visual Intuition’s Claire Swadling on Teaching AI to ‘Visually Reason’ Like Humans was originally published in Authority Magazine on Medium, where people are continuing the conversation by highlighting and responding to this story.